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_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 /* BPF syscall commands, see bpf(2) man-page for details. */ 85 enum bpf_cmd { 86 BPF_MAP_CREATE, 87 BPF_MAP_LOOKUP_ELEM, 88 BPF_MAP_UPDATE_ELEM, 89 BPF_MAP_DELETE_ELEM, 90 BPF_MAP_GET_NEXT_KEY, 91 BPF_PROG_LOAD, 92 BPF_OBJ_PIN, 93 BPF_OBJ_GET, 94 BPF_PROG_ATTACH, 95 BPF_PROG_DETACH, 96 BPF_PROG_TEST_RUN, 97 BPF_PROG_GET_NEXT_ID, 98 BPF_MAP_GET_NEXT_ID, 99 BPF_PROG_GET_FD_BY_ID, 100 BPF_MAP_GET_FD_BY_ID, 101 BPF_OBJ_GET_INFO_BY_FD, 102 BPF_PROG_QUERY, 103 BPF_RAW_TRACEPOINT_OPEN, 104 BPF_BTF_LOAD, 105 BPF_BTF_GET_FD_BY_ID, 106 BPF_TASK_FD_QUERY, 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 108 BPF_MAP_FREEZE, 109 }; 110 111 enum bpf_map_type { 112 BPF_MAP_TYPE_UNSPEC, 113 BPF_MAP_TYPE_HASH, 114 BPF_MAP_TYPE_ARRAY, 115 BPF_MAP_TYPE_PROG_ARRAY, 116 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 117 BPF_MAP_TYPE_PERCPU_HASH, 118 BPF_MAP_TYPE_PERCPU_ARRAY, 119 BPF_MAP_TYPE_STACK_TRACE, 120 BPF_MAP_TYPE_CGROUP_ARRAY, 121 BPF_MAP_TYPE_LRU_HASH, 122 BPF_MAP_TYPE_LRU_PERCPU_HASH, 123 BPF_MAP_TYPE_LPM_TRIE, 124 BPF_MAP_TYPE_ARRAY_OF_MAPS, 125 BPF_MAP_TYPE_HASH_OF_MAPS, 126 BPF_MAP_TYPE_DEVMAP, 127 BPF_MAP_TYPE_SOCKMAP, 128 BPF_MAP_TYPE_CPUMAP, 129 BPF_MAP_TYPE_XSKMAP, 130 BPF_MAP_TYPE_SOCKHASH, 131 BPF_MAP_TYPE_CGROUP_STORAGE, 132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 134 BPF_MAP_TYPE_QUEUE, 135 BPF_MAP_TYPE_STACK, 136 BPF_MAP_TYPE_SK_STORAGE, 137 }; 138 139 /* Note that tracing related programs such as 140 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 141 * are not subject to a stable API since kernel internal data 142 * structures can change from release to release and may 143 * therefore break existing tracing BPF programs. Tracing BPF 144 * programs correspond to /a/ specific kernel which is to be 145 * analyzed, and not /a/ specific kernel /and/ all future ones. 146 */ 147 enum bpf_prog_type { 148 BPF_PROG_TYPE_UNSPEC, 149 BPF_PROG_TYPE_SOCKET_FILTER, 150 BPF_PROG_TYPE_KPROBE, 151 BPF_PROG_TYPE_SCHED_CLS, 152 BPF_PROG_TYPE_SCHED_ACT, 153 BPF_PROG_TYPE_TRACEPOINT, 154 BPF_PROG_TYPE_XDP, 155 BPF_PROG_TYPE_PERF_EVENT, 156 BPF_PROG_TYPE_CGROUP_SKB, 157 BPF_PROG_TYPE_CGROUP_SOCK, 158 BPF_PROG_TYPE_LWT_IN, 159 BPF_PROG_TYPE_LWT_OUT, 160 BPF_PROG_TYPE_LWT_XMIT, 161 BPF_PROG_TYPE_SOCK_OPS, 162 BPF_PROG_TYPE_SK_SKB, 163 BPF_PROG_TYPE_CGROUP_DEVICE, 164 BPF_PROG_TYPE_SK_MSG, 165 BPF_PROG_TYPE_RAW_TRACEPOINT, 166 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 167 BPF_PROG_TYPE_LWT_SEG6LOCAL, 168 BPF_PROG_TYPE_LIRC_MODE2, 169 BPF_PROG_TYPE_SK_REUSEPORT, 170 BPF_PROG_TYPE_FLOW_DISSECTOR, 171 BPF_PROG_TYPE_CGROUP_SYSCTL, 172 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 173 }; 174 175 enum bpf_attach_type { 176 BPF_CGROUP_INET_INGRESS, 177 BPF_CGROUP_INET_EGRESS, 178 BPF_CGROUP_INET_SOCK_CREATE, 179 BPF_CGROUP_SOCK_OPS, 180 BPF_SK_SKB_STREAM_PARSER, 181 BPF_SK_SKB_STREAM_VERDICT, 182 BPF_CGROUP_DEVICE, 183 BPF_SK_MSG_VERDICT, 184 BPF_CGROUP_INET4_BIND, 185 BPF_CGROUP_INET6_BIND, 186 BPF_CGROUP_INET4_CONNECT, 187 BPF_CGROUP_INET6_CONNECT, 188 BPF_CGROUP_INET4_POST_BIND, 189 BPF_CGROUP_INET6_POST_BIND, 190 BPF_CGROUP_UDP4_SENDMSG, 191 BPF_CGROUP_UDP6_SENDMSG, 192 BPF_LIRC_MODE2, 193 BPF_FLOW_DISSECTOR, 194 BPF_CGROUP_SYSCTL, 195 __MAX_BPF_ATTACH_TYPE 196 }; 197 198 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 199 200 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 201 * 202 * NONE(default): No further bpf programs allowed in the subtree. 203 * 204 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 205 * the program in this cgroup yields to sub-cgroup program. 206 * 207 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 208 * that cgroup program gets run in addition to the program in this cgroup. 209 * 210 * Only one program is allowed to be attached to a cgroup with 211 * NONE or BPF_F_ALLOW_OVERRIDE flag. 212 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 213 * release old program and attach the new one. Attach flags has to match. 214 * 215 * Multiple programs are allowed to be attached to a cgroup with 216 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 217 * (those that were attached first, run first) 218 * The programs of sub-cgroup are executed first, then programs of 219 * this cgroup and then programs of parent cgroup. 220 * When children program makes decision (like picking TCP CA or sock bind) 221 * parent program has a chance to override it. 222 * 223 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 224 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 225 * Ex1: 226 * cgrp1 (MULTI progs A, B) -> 227 * cgrp2 (OVERRIDE prog C) -> 228 * cgrp3 (MULTI prog D) -> 229 * cgrp4 (OVERRIDE prog E) -> 230 * cgrp5 (NONE prog F) 231 * the event in cgrp5 triggers execution of F,D,A,B in that order. 232 * if prog F is detached, the execution is E,D,A,B 233 * if prog F and D are detached, the execution is E,A,B 234 * if prog F, E and D are detached, the execution is C,A,B 235 * 236 * All eligible programs are executed regardless of return code from 237 * earlier programs. 238 */ 239 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 240 #define BPF_F_ALLOW_MULTI (1U << 1) 241 242 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 243 * verifier will perform strict alignment checking as if the kernel 244 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 245 * and NET_IP_ALIGN defined to 2. 246 */ 247 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 248 249 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 250 * verifier will allow any alignment whatsoever. On platforms 251 * with strict alignment requirements for loads ands stores (such 252 * as sparc and mips) the verifier validates that all loads and 253 * stores provably follow this requirement. This flag turns that 254 * checking and enforcement off. 255 * 256 * It is mostly used for testing when we want to validate the 257 * context and memory access aspects of the verifier, but because 258 * of an unaligned access the alignment check would trigger before 259 * the one we are interested in. 260 */ 261 #define BPF_F_ANY_ALIGNMENT (1U << 1) 262 263 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 264 * Verifier does sub-register def/use analysis and identifies instructions whose 265 * def only matters for low 32-bit, high 32-bit is never referenced later 266 * through implicit zero extension. Therefore verifier notifies JIT back-ends 267 * that it is safe to ignore clearing high 32-bit for these instructions. This 268 * saves some back-ends a lot of code-gen. However such optimization is not 269 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 270 * hence hasn't used verifier's analysis result. But, we really want to have a 271 * way to be able to verify the correctness of the described optimization on 272 * x86_64 on which testsuites are frequently exercised. 273 * 274 * So, this flag is introduced. Once it is set, verifier will randomize high 275 * 32-bit for those instructions who has been identified as safe to ignore them. 276 * Then, if verifier is not doing correct analysis, such randomization will 277 * regress tests to expose bugs. 278 */ 279 #define BPF_F_TEST_RND_HI32 (1U << 2) 280 281 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 282 * two extensions: 283 * 284 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 285 * insn[0].imm: map fd map fd 286 * insn[1].imm: 0 offset into value 287 * insn[0].off: 0 0 288 * insn[1].off: 0 0 289 * ldimm64 rewrite: address of map address of map[0]+offset 290 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 291 */ 292 #define BPF_PSEUDO_MAP_FD 1 293 #define BPF_PSEUDO_MAP_VALUE 2 294 295 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 296 * offset to another bpf function 297 */ 298 #define BPF_PSEUDO_CALL 1 299 300 /* flags for BPF_MAP_UPDATE_ELEM command */ 301 #define BPF_ANY 0 /* create new element or update existing */ 302 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 303 #define BPF_EXIST 2 /* update existing element */ 304 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 305 306 /* flags for BPF_MAP_CREATE command */ 307 #define BPF_F_NO_PREALLOC (1U << 0) 308 /* Instead of having one common LRU list in the 309 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 310 * which can scale and perform better. 311 * Note, the LRU nodes (including free nodes) cannot be moved 312 * across different LRU lists. 313 */ 314 #define BPF_F_NO_COMMON_LRU (1U << 1) 315 /* Specify numa node during map creation */ 316 #define BPF_F_NUMA_NODE (1U << 2) 317 318 #define BPF_OBJ_NAME_LEN 16U 319 320 /* Flags for accessing BPF object from syscall side. */ 321 #define BPF_F_RDONLY (1U << 3) 322 #define BPF_F_WRONLY (1U << 4) 323 324 /* Flag for stack_map, store build_id+offset instead of pointer */ 325 #define BPF_F_STACK_BUILD_ID (1U << 5) 326 327 /* Zero-initialize hash function seed. This should only be used for testing. */ 328 #define BPF_F_ZERO_SEED (1U << 6) 329 330 /* Flags for accessing BPF object from program side. */ 331 #define BPF_F_RDONLY_PROG (1U << 7) 332 #define BPF_F_WRONLY_PROG (1U << 8) 333 334 /* flags for BPF_PROG_QUERY */ 335 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 336 337 enum bpf_stack_build_id_status { 338 /* user space need an empty entry to identify end of a trace */ 339 BPF_STACK_BUILD_ID_EMPTY = 0, 340 /* with valid build_id and offset */ 341 BPF_STACK_BUILD_ID_VALID = 1, 342 /* couldn't get build_id, fallback to ip */ 343 BPF_STACK_BUILD_ID_IP = 2, 344 }; 345 346 #define BPF_BUILD_ID_SIZE 20 347 struct bpf_stack_build_id { 348 __s32 status; 349 unsigned char build_id[BPF_BUILD_ID_SIZE]; 350 union { 351 __u64 offset; 352 __u64 ip; 353 }; 354 }; 355 356 union bpf_attr { 357 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 358 __u32 map_type; /* one of enum bpf_map_type */ 359 __u32 key_size; /* size of key in bytes */ 360 __u32 value_size; /* size of value in bytes */ 361 __u32 max_entries; /* max number of entries in a map */ 362 __u32 map_flags; /* BPF_MAP_CREATE related 363 * flags defined above. 364 */ 365 __u32 inner_map_fd; /* fd pointing to the inner map */ 366 __u32 numa_node; /* numa node (effective only if 367 * BPF_F_NUMA_NODE is set). 368 */ 369 char map_name[BPF_OBJ_NAME_LEN]; 370 __u32 map_ifindex; /* ifindex of netdev to create on */ 371 __u32 btf_fd; /* fd pointing to a BTF type data */ 372 __u32 btf_key_type_id; /* BTF type_id of the key */ 373 __u32 btf_value_type_id; /* BTF type_id of the value */ 374 }; 375 376 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 377 __u32 map_fd; 378 __aligned_u64 key; 379 union { 380 __aligned_u64 value; 381 __aligned_u64 next_key; 382 }; 383 __u64 flags; 384 }; 385 386 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 387 __u32 prog_type; /* one of enum bpf_prog_type */ 388 __u32 insn_cnt; 389 __aligned_u64 insns; 390 __aligned_u64 license; 391 __u32 log_level; /* verbosity level of verifier */ 392 __u32 log_size; /* size of user buffer */ 393 __aligned_u64 log_buf; /* user supplied buffer */ 394 __u32 kern_version; /* not used */ 395 __u32 prog_flags; 396 char prog_name[BPF_OBJ_NAME_LEN]; 397 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 398 /* For some prog types expected attach type must be known at 399 * load time to verify attach type specific parts of prog 400 * (context accesses, allowed helpers, etc). 401 */ 402 __u32 expected_attach_type; 403 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 404 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 405 __aligned_u64 func_info; /* func info */ 406 __u32 func_info_cnt; /* number of bpf_func_info records */ 407 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 408 __aligned_u64 line_info; /* line info */ 409 __u32 line_info_cnt; /* number of bpf_line_info records */ 410 }; 411 412 struct { /* anonymous struct used by BPF_OBJ_* commands */ 413 __aligned_u64 pathname; 414 __u32 bpf_fd; 415 __u32 file_flags; 416 }; 417 418 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 419 __u32 target_fd; /* container object to attach to */ 420 __u32 attach_bpf_fd; /* eBPF program to attach */ 421 __u32 attach_type; 422 __u32 attach_flags; 423 }; 424 425 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 426 __u32 prog_fd; 427 __u32 retval; 428 __u32 data_size_in; /* input: len of data_in */ 429 __u32 data_size_out; /* input/output: len of data_out 430 * returns ENOSPC if data_out 431 * is too small. 432 */ 433 __aligned_u64 data_in; 434 __aligned_u64 data_out; 435 __u32 repeat; 436 __u32 duration; 437 __u32 ctx_size_in; /* input: len of ctx_in */ 438 __u32 ctx_size_out; /* input/output: len of ctx_out 439 * returns ENOSPC if ctx_out 440 * is too small. 441 */ 442 __aligned_u64 ctx_in; 443 __aligned_u64 ctx_out; 444 } test; 445 446 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 447 union { 448 __u32 start_id; 449 __u32 prog_id; 450 __u32 map_id; 451 __u32 btf_id; 452 }; 453 __u32 next_id; 454 __u32 open_flags; 455 }; 456 457 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 458 __u32 bpf_fd; 459 __u32 info_len; 460 __aligned_u64 info; 461 } info; 462 463 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 464 __u32 target_fd; /* container object to query */ 465 __u32 attach_type; 466 __u32 query_flags; 467 __u32 attach_flags; 468 __aligned_u64 prog_ids; 469 __u32 prog_cnt; 470 } query; 471 472 struct { 473 __u64 name; 474 __u32 prog_fd; 475 } raw_tracepoint; 476 477 struct { /* anonymous struct for BPF_BTF_LOAD */ 478 __aligned_u64 btf; 479 __aligned_u64 btf_log_buf; 480 __u32 btf_size; 481 __u32 btf_log_size; 482 __u32 btf_log_level; 483 }; 484 485 struct { 486 __u32 pid; /* input: pid */ 487 __u32 fd; /* input: fd */ 488 __u32 flags; /* input: flags */ 489 __u32 buf_len; /* input/output: buf len */ 490 __aligned_u64 buf; /* input/output: 491 * tp_name for tracepoint 492 * symbol for kprobe 493 * filename for uprobe 494 */ 495 __u32 prog_id; /* output: prod_id */ 496 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 497 __u64 probe_offset; /* output: probe_offset */ 498 __u64 probe_addr; /* output: probe_addr */ 499 } task_fd_query; 500 } __attribute__((aligned(8))); 501 502 /* The description below is an attempt at providing documentation to eBPF 503 * developers about the multiple available eBPF helper functions. It can be 504 * parsed and used to produce a manual page. The workflow is the following, 505 * and requires the rst2man utility: 506 * 507 * $ ./scripts/bpf_helpers_doc.py \ 508 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 509 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 510 * $ man /tmp/bpf-helpers.7 511 * 512 * Note that in order to produce this external documentation, some RST 513 * formatting is used in the descriptions to get "bold" and "italics" in 514 * manual pages. Also note that the few trailing white spaces are 515 * intentional, removing them would break paragraphs for rst2man. 516 * 517 * Start of BPF helper function descriptions: 518 * 519 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 520 * Description 521 * Perform a lookup in *map* for an entry associated to *key*. 522 * Return 523 * Map value associated to *key*, or **NULL** if no entry was 524 * found. 525 * 526 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 527 * Description 528 * Add or update the value of the entry associated to *key* in 529 * *map* with *value*. *flags* is one of: 530 * 531 * **BPF_NOEXIST** 532 * The entry for *key* must not exist in the map. 533 * **BPF_EXIST** 534 * The entry for *key* must already exist in the map. 535 * **BPF_ANY** 536 * No condition on the existence of the entry for *key*. 537 * 538 * Flag value **BPF_NOEXIST** cannot be used for maps of types 539 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 540 * elements always exist), the helper would return an error. 541 * Return 542 * 0 on success, or a negative error in case of failure. 543 * 544 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 545 * Description 546 * Delete entry with *key* from *map*. 547 * Return 548 * 0 on success, or a negative error in case of failure. 549 * 550 * int bpf_probe_read(void *dst, u32 size, const void *src) 551 * Description 552 * For tracing programs, safely attempt to read *size* bytes from 553 * address *src* and store the data in *dst*. 554 * Return 555 * 0 on success, or a negative error in case of failure. 556 * 557 * u64 bpf_ktime_get_ns(void) 558 * Description 559 * Return the time elapsed since system boot, in nanoseconds. 560 * Return 561 * Current *ktime*. 562 * 563 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 564 * Description 565 * This helper is a "printk()-like" facility for debugging. It 566 * prints a message defined by format *fmt* (of size *fmt_size*) 567 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 568 * available. It can take up to three additional **u64** 569 * arguments (as an eBPF helpers, the total number of arguments is 570 * limited to five). 571 * 572 * Each time the helper is called, it appends a line to the trace. 573 * The format of the trace is customizable, and the exact output 574 * one will get depends on the options set in 575 * *\/sys/kernel/debug/tracing/trace_options* (see also the 576 * *README* file under the same directory). However, it usually 577 * defaults to something like: 578 * 579 * :: 580 * 581 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 582 * 583 * In the above: 584 * 585 * * ``telnet`` is the name of the current task. 586 * * ``470`` is the PID of the current task. 587 * * ``001`` is the CPU number on which the task is 588 * running. 589 * * In ``.N..``, each character refers to a set of 590 * options (whether irqs are enabled, scheduling 591 * options, whether hard/softirqs are running, level of 592 * preempt_disabled respectively). **N** means that 593 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 594 * are set. 595 * * ``419421.045894`` is a timestamp. 596 * * ``0x00000001`` is a fake value used by BPF for the 597 * instruction pointer register. 598 * * ``<formatted msg>`` is the message formatted with 599 * *fmt*. 600 * 601 * The conversion specifiers supported by *fmt* are similar, but 602 * more limited than for printk(). They are **%d**, **%i**, 603 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 604 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 605 * of field, padding with zeroes, etc.) is available, and the 606 * helper will return **-EINVAL** (but print nothing) if it 607 * encounters an unknown specifier. 608 * 609 * Also, note that **bpf_trace_printk**\ () is slow, and should 610 * only be used for debugging purposes. For this reason, a notice 611 * bloc (spanning several lines) is printed to kernel logs and 612 * states that the helper should not be used "for production use" 613 * the first time this helper is used (or more precisely, when 614 * **trace_printk**\ () buffers are allocated). For passing values 615 * to user space, perf events should be preferred. 616 * Return 617 * The number of bytes written to the buffer, or a negative error 618 * in case of failure. 619 * 620 * u32 bpf_get_prandom_u32(void) 621 * Description 622 * Get a pseudo-random number. 623 * 624 * From a security point of view, this helper uses its own 625 * pseudo-random internal state, and cannot be used to infer the 626 * seed of other random functions in the kernel. However, it is 627 * essential to note that the generator used by the helper is not 628 * cryptographically secure. 629 * Return 630 * A random 32-bit unsigned value. 631 * 632 * u32 bpf_get_smp_processor_id(void) 633 * Description 634 * Get the SMP (symmetric multiprocessing) processor id. Note that 635 * all programs run with preemption disabled, which means that the 636 * SMP processor id is stable during all the execution of the 637 * program. 638 * Return 639 * The SMP id of the processor running the program. 640 * 641 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 642 * Description 643 * Store *len* bytes from address *from* into the packet 644 * associated to *skb*, at *offset*. *flags* are a combination of 645 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 646 * checksum for the packet after storing the bytes) and 647 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 648 * **->swhash** and *skb*\ **->l4hash** to 0). 649 * 650 * A call to this helper is susceptible to change the underlying 651 * packet buffer. Therefore, at load time, all checks on pointers 652 * previously done by the verifier are invalidated and must be 653 * performed again, if the helper is used in combination with 654 * direct packet access. 655 * Return 656 * 0 on success, or a negative error in case of failure. 657 * 658 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 659 * Description 660 * Recompute the layer 3 (e.g. IP) checksum for the packet 661 * associated to *skb*. Computation is incremental, so the helper 662 * must know the former value of the header field that was 663 * modified (*from*), the new value of this field (*to*), and the 664 * number of bytes (2 or 4) for this field, stored in *size*. 665 * Alternatively, it is possible to store the difference between 666 * the previous and the new values of the header field in *to*, by 667 * setting *from* and *size* to 0. For both methods, *offset* 668 * indicates the location of the IP checksum within the packet. 669 * 670 * This helper works in combination with **bpf_csum_diff**\ (), 671 * which does not update the checksum in-place, but offers more 672 * flexibility and can handle sizes larger than 2 or 4 for the 673 * checksum to update. 674 * 675 * A call to this helper is susceptible to change the underlying 676 * packet buffer. Therefore, at load time, all checks on pointers 677 * previously done by the verifier are invalidated and must be 678 * performed again, if the helper is used in combination with 679 * direct packet access. 680 * Return 681 * 0 on success, or a negative error in case of failure. 682 * 683 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 684 * Description 685 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 686 * packet associated to *skb*. Computation is incremental, so the 687 * helper must know the former value of the header field that was 688 * modified (*from*), the new value of this field (*to*), and the 689 * number of bytes (2 or 4) for this field, stored on the lowest 690 * four bits of *flags*. Alternatively, it is possible to store 691 * the difference between the previous and the new values of the 692 * header field in *to*, by setting *from* and the four lowest 693 * bits of *flags* to 0. For both methods, *offset* indicates the 694 * location of the IP checksum within the packet. In addition to 695 * the size of the field, *flags* can be added (bitwise OR) actual 696 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 697 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 698 * for updates resulting in a null checksum the value is set to 699 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 700 * the checksum is to be computed against a pseudo-header. 701 * 702 * This helper works in combination with **bpf_csum_diff**\ (), 703 * which does not update the checksum in-place, but offers more 704 * flexibility and can handle sizes larger than 2 or 4 for the 705 * checksum to update. 706 * 707 * A call to this helper is susceptible to change the underlying 708 * packet buffer. Therefore, at load time, all checks on pointers 709 * previously done by the verifier are invalidated and must be 710 * performed again, if the helper is used in combination with 711 * direct packet access. 712 * Return 713 * 0 on success, or a negative error in case of failure. 714 * 715 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 716 * Description 717 * This special helper is used to trigger a "tail call", or in 718 * other words, to jump into another eBPF program. The same stack 719 * frame is used (but values on stack and in registers for the 720 * caller are not accessible to the callee). This mechanism allows 721 * for program chaining, either for raising the maximum number of 722 * available eBPF instructions, or to execute given programs in 723 * conditional blocks. For security reasons, there is an upper 724 * limit to the number of successive tail calls that can be 725 * performed. 726 * 727 * Upon call of this helper, the program attempts to jump into a 728 * program referenced at index *index* in *prog_array_map*, a 729 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 730 * *ctx*, a pointer to the context. 731 * 732 * If the call succeeds, the kernel immediately runs the first 733 * instruction of the new program. This is not a function call, 734 * and it never returns to the previous program. If the call 735 * fails, then the helper has no effect, and the caller continues 736 * to run its subsequent instructions. A call can fail if the 737 * destination program for the jump does not exist (i.e. *index* 738 * is superior to the number of entries in *prog_array_map*), or 739 * if the maximum number of tail calls has been reached for this 740 * chain of programs. This limit is defined in the kernel by the 741 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 742 * which is currently set to 32. 743 * Return 744 * 0 on success, or a negative error in case of failure. 745 * 746 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 747 * Description 748 * Clone and redirect the packet associated to *skb* to another 749 * net device of index *ifindex*. Both ingress and egress 750 * interfaces can be used for redirection. The **BPF_F_INGRESS** 751 * value in *flags* is used to make the distinction (ingress path 752 * is selected if the flag is present, egress path otherwise). 753 * This is the only flag supported for now. 754 * 755 * In comparison with **bpf_redirect**\ () helper, 756 * **bpf_clone_redirect**\ () has the associated cost of 757 * duplicating the packet buffer, but this can be executed out of 758 * the eBPF program. Conversely, **bpf_redirect**\ () is more 759 * efficient, but it is handled through an action code where the 760 * redirection happens only after the eBPF program has returned. 761 * 762 * A call to this helper is susceptible to change the underlying 763 * packet buffer. Therefore, at load time, all checks on pointers 764 * previously done by the verifier are invalidated and must be 765 * performed again, if the helper is used in combination with 766 * direct packet access. 767 * Return 768 * 0 on success, or a negative error in case of failure. 769 * 770 * u64 bpf_get_current_pid_tgid(void) 771 * Return 772 * A 64-bit integer containing the current tgid and pid, and 773 * created as such: 774 * *current_task*\ **->tgid << 32 \|** 775 * *current_task*\ **->pid**. 776 * 777 * u64 bpf_get_current_uid_gid(void) 778 * Return 779 * A 64-bit integer containing the current GID and UID, and 780 * created as such: *current_gid* **<< 32 \|** *current_uid*. 781 * 782 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 783 * Description 784 * Copy the **comm** attribute of the current task into *buf* of 785 * *size_of_buf*. The **comm** attribute contains the name of 786 * the executable (excluding the path) for the current task. The 787 * *size_of_buf* must be strictly positive. On success, the 788 * helper makes sure that the *buf* is NUL-terminated. On failure, 789 * it is filled with zeroes. 790 * Return 791 * 0 on success, or a negative error in case of failure. 792 * 793 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 794 * Description 795 * Retrieve the classid for the current task, i.e. for the net_cls 796 * cgroup to which *skb* belongs. 797 * 798 * This helper can be used on TC egress path, but not on ingress. 799 * 800 * The net_cls cgroup provides an interface to tag network packets 801 * based on a user-provided identifier for all traffic coming from 802 * the tasks belonging to the related cgroup. See also the related 803 * kernel documentation, available from the Linux sources in file 804 * *Documentation/cgroup-v1/net_cls.txt*. 805 * 806 * The Linux kernel has two versions for cgroups: there are 807 * cgroups v1 and cgroups v2. Both are available to users, who can 808 * use a mixture of them, but note that the net_cls cgroup is for 809 * cgroup v1 only. This makes it incompatible with BPF programs 810 * run on cgroups, which is a cgroup-v2-only feature (a socket can 811 * only hold data for one version of cgroups at a time). 812 * 813 * This helper is only available is the kernel was compiled with 814 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 815 * "**y**" or to "**m**". 816 * Return 817 * The classid, or 0 for the default unconfigured classid. 818 * 819 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 820 * Description 821 * Push a *vlan_tci* (VLAN tag control information) of protocol 822 * *vlan_proto* to the packet associated to *skb*, then update 823 * the checksum. Note that if *vlan_proto* is different from 824 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 825 * be **ETH_P_8021Q**. 826 * 827 * A call to this helper is susceptible to change the underlying 828 * packet buffer. Therefore, at load time, all checks on pointers 829 * previously done by the verifier are invalidated and must be 830 * performed again, if the helper is used in combination with 831 * direct packet access. 832 * Return 833 * 0 on success, or a negative error in case of failure. 834 * 835 * int bpf_skb_vlan_pop(struct sk_buff *skb) 836 * Description 837 * Pop a VLAN header from the packet associated to *skb*. 838 * 839 * A call to this helper is susceptible to change the underlying 840 * packet buffer. Therefore, at load time, all checks on pointers 841 * previously done by the verifier are invalidated and must be 842 * performed again, if the helper is used in combination with 843 * direct packet access. 844 * Return 845 * 0 on success, or a negative error in case of failure. 846 * 847 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 848 * Description 849 * Get tunnel metadata. This helper takes a pointer *key* to an 850 * empty **struct bpf_tunnel_key** of **size**, that will be 851 * filled with tunnel metadata for the packet associated to *skb*. 852 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 853 * indicates that the tunnel is based on IPv6 protocol instead of 854 * IPv4. 855 * 856 * The **struct bpf_tunnel_key** is an object that generalizes the 857 * principal parameters used by various tunneling protocols into a 858 * single struct. This way, it can be used to easily make a 859 * decision based on the contents of the encapsulation header, 860 * "summarized" in this struct. In particular, it holds the IP 861 * address of the remote end (IPv4 or IPv6, depending on the case) 862 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 863 * this struct exposes the *key*\ **->tunnel_id**, which is 864 * generally mapped to a VNI (Virtual Network Identifier), making 865 * it programmable together with the **bpf_skb_set_tunnel_key**\ 866 * () helper. 867 * 868 * Let's imagine that the following code is part of a program 869 * attached to the TC ingress interface, on one end of a GRE 870 * tunnel, and is supposed to filter out all messages coming from 871 * remote ends with IPv4 address other than 10.0.0.1: 872 * 873 * :: 874 * 875 * int ret; 876 * struct bpf_tunnel_key key = {}; 877 * 878 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 879 * if (ret < 0) 880 * return TC_ACT_SHOT; // drop packet 881 * 882 * if (key.remote_ipv4 != 0x0a000001) 883 * return TC_ACT_SHOT; // drop packet 884 * 885 * return TC_ACT_OK; // accept packet 886 * 887 * This interface can also be used with all encapsulation devices 888 * that can operate in "collect metadata" mode: instead of having 889 * one network device per specific configuration, the "collect 890 * metadata" mode only requires a single device where the 891 * configuration can be extracted from this helper. 892 * 893 * This can be used together with various tunnels such as VXLan, 894 * Geneve, GRE or IP in IP (IPIP). 895 * Return 896 * 0 on success, or a negative error in case of failure. 897 * 898 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 899 * Description 900 * Populate tunnel metadata for packet associated to *skb.* The 901 * tunnel metadata is set to the contents of *key*, of *size*. The 902 * *flags* can be set to a combination of the following values: 903 * 904 * **BPF_F_TUNINFO_IPV6** 905 * Indicate that the tunnel is based on IPv6 protocol 906 * instead of IPv4. 907 * **BPF_F_ZERO_CSUM_TX** 908 * For IPv4 packets, add a flag to tunnel metadata 909 * indicating that checksum computation should be skipped 910 * and checksum set to zeroes. 911 * **BPF_F_DONT_FRAGMENT** 912 * Add a flag to tunnel metadata indicating that the 913 * packet should not be fragmented. 914 * **BPF_F_SEQ_NUMBER** 915 * Add a flag to tunnel metadata indicating that a 916 * sequence number should be added to tunnel header before 917 * sending the packet. This flag was added for GRE 918 * encapsulation, but might be used with other protocols 919 * as well in the future. 920 * 921 * Here is a typical usage on the transmit path: 922 * 923 * :: 924 * 925 * struct bpf_tunnel_key key; 926 * populate key ... 927 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 928 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 929 * 930 * See also the description of the **bpf_skb_get_tunnel_key**\ () 931 * helper for additional information. 932 * Return 933 * 0 on success, or a negative error in case of failure. 934 * 935 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 936 * Description 937 * Read the value of a perf event counter. This helper relies on a 938 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 939 * the perf event counter is selected when *map* is updated with 940 * perf event file descriptors. The *map* is an array whose size 941 * is the number of available CPUs, and each cell contains a value 942 * relative to one CPU. The value to retrieve is indicated by 943 * *flags*, that contains the index of the CPU to look up, masked 944 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 945 * **BPF_F_CURRENT_CPU** to indicate that the value for the 946 * current CPU should be retrieved. 947 * 948 * Note that before Linux 4.13, only hardware perf event can be 949 * retrieved. 950 * 951 * Also, be aware that the newer helper 952 * **bpf_perf_event_read_value**\ () is recommended over 953 * **bpf_perf_event_read**\ () in general. The latter has some ABI 954 * quirks where error and counter value are used as a return code 955 * (which is wrong to do since ranges may overlap). This issue is 956 * fixed with **bpf_perf_event_read_value**\ (), which at the same 957 * time provides more features over the **bpf_perf_event_read**\ 958 * () interface. Please refer to the description of 959 * **bpf_perf_event_read_value**\ () for details. 960 * Return 961 * The value of the perf event counter read from the map, or a 962 * negative error code in case of failure. 963 * 964 * int bpf_redirect(u32 ifindex, u64 flags) 965 * Description 966 * Redirect the packet to another net device of index *ifindex*. 967 * This helper is somewhat similar to **bpf_clone_redirect**\ 968 * (), except that the packet is not cloned, which provides 969 * increased performance. 970 * 971 * Except for XDP, both ingress and egress interfaces can be used 972 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 973 * to make the distinction (ingress path is selected if the flag 974 * is present, egress path otherwise). Currently, XDP only 975 * supports redirection to the egress interface, and accepts no 976 * flag at all. 977 * 978 * The same effect can be attained with the more generic 979 * **bpf_redirect_map**\ (), which requires specific maps to be 980 * used but offers better performance. 981 * Return 982 * For XDP, the helper returns **XDP_REDIRECT** on success or 983 * **XDP_ABORTED** on error. For other program types, the values 984 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 985 * error. 986 * 987 * u32 bpf_get_route_realm(struct sk_buff *skb) 988 * Description 989 * Retrieve the realm or the route, that is to say the 990 * **tclassid** field of the destination for the *skb*. The 991 * indentifier retrieved is a user-provided tag, similar to the 992 * one used with the net_cls cgroup (see description for 993 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 994 * held by a route (a destination entry), not by a task. 995 * 996 * Retrieving this identifier works with the clsact TC egress hook 997 * (see also **tc-bpf(8)**), or alternatively on conventional 998 * classful egress qdiscs, but not on TC ingress path. In case of 999 * clsact TC egress hook, this has the advantage that, internally, 1000 * the destination entry has not been dropped yet in the transmit 1001 * path. Therefore, the destination entry does not need to be 1002 * artificially held via **netif_keep_dst**\ () for a classful 1003 * qdisc until the *skb* is freed. 1004 * 1005 * This helper is available only if the kernel was compiled with 1006 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1007 * Return 1008 * The realm of the route for the packet associated to *skb*, or 0 1009 * if none was found. 1010 * 1011 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1012 * Description 1013 * Write raw *data* blob into a special BPF perf event held by 1014 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1015 * event must have the following attributes: **PERF_SAMPLE_RAW** 1016 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1017 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1018 * 1019 * The *flags* are used to indicate the index in *map* for which 1020 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1021 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1022 * to indicate that the index of the current CPU core should be 1023 * used. 1024 * 1025 * The value to write, of *size*, is passed through eBPF stack and 1026 * pointed by *data*. 1027 * 1028 * The context of the program *ctx* needs also be passed to the 1029 * helper. 1030 * 1031 * On user space, a program willing to read the values needs to 1032 * call **perf_event_open**\ () on the perf event (either for 1033 * one or for all CPUs) and to store the file descriptor into the 1034 * *map*. This must be done before the eBPF program can send data 1035 * into it. An example is available in file 1036 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1037 * tree (the eBPF program counterpart is in 1038 * *samples/bpf/trace_output_kern.c*). 1039 * 1040 * **bpf_perf_event_output**\ () achieves better performance 1041 * than **bpf_trace_printk**\ () for sharing data with user 1042 * space, and is much better suitable for streaming data from eBPF 1043 * programs. 1044 * 1045 * Note that this helper is not restricted to tracing use cases 1046 * and can be used with programs attached to TC or XDP as well, 1047 * where it allows for passing data to user space listeners. Data 1048 * can be: 1049 * 1050 * * Only custom structs, 1051 * * Only the packet payload, or 1052 * * A combination of both. 1053 * Return 1054 * 0 on success, or a negative error in case of failure. 1055 * 1056 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 1057 * Description 1058 * This helper was provided as an easy way to load data from a 1059 * packet. It can be used to load *len* bytes from *offset* from 1060 * the packet associated to *skb*, into the buffer pointed by 1061 * *to*. 1062 * 1063 * Since Linux 4.7, usage of this helper has mostly been replaced 1064 * by "direct packet access", enabling packet data to be 1065 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1066 * pointing respectively to the first byte of packet data and to 1067 * the byte after the last byte of packet data. However, it 1068 * remains useful if one wishes to read large quantities of data 1069 * at once from a packet into the eBPF stack. 1070 * Return 1071 * 0 on success, or a negative error in case of failure. 1072 * 1073 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 1074 * Description 1075 * Walk a user or a kernel stack and return its id. To achieve 1076 * this, the helper needs *ctx*, which is a pointer to the context 1077 * on which the tracing program is executed, and a pointer to a 1078 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1079 * 1080 * The last argument, *flags*, holds the number of stack frames to 1081 * skip (from 0 to 255), masked with 1082 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1083 * a combination of the following flags: 1084 * 1085 * **BPF_F_USER_STACK** 1086 * Collect a user space stack instead of a kernel stack. 1087 * **BPF_F_FAST_STACK_CMP** 1088 * Compare stacks by hash only. 1089 * **BPF_F_REUSE_STACKID** 1090 * If two different stacks hash into the same *stackid*, 1091 * discard the old one. 1092 * 1093 * The stack id retrieved is a 32 bit long integer handle which 1094 * can be further combined with other data (including other stack 1095 * ids) and used as a key into maps. This can be useful for 1096 * generating a variety of graphs (such as flame graphs or off-cpu 1097 * graphs). 1098 * 1099 * For walking a stack, this helper is an improvement over 1100 * **bpf_probe_read**\ (), which can be used with unrolled loops 1101 * but is not efficient and consumes a lot of eBPF instructions. 1102 * Instead, **bpf_get_stackid**\ () can collect up to 1103 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1104 * this limit can be controlled with the **sysctl** program, and 1105 * that it should be manually increased in order to profile long 1106 * user stacks (such as stacks for Java programs). To do so, use: 1107 * 1108 * :: 1109 * 1110 * # sysctl kernel.perf_event_max_stack=<new value> 1111 * Return 1112 * The positive or null stack id on success, or a negative error 1113 * in case of failure. 1114 * 1115 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1116 * Description 1117 * Compute a checksum difference, from the raw buffer pointed by 1118 * *from*, of length *from_size* (that must be a multiple of 4), 1119 * towards the raw buffer pointed by *to*, of size *to_size* 1120 * (same remark). An optional *seed* can be added to the value 1121 * (this can be cascaded, the seed may come from a previous call 1122 * to the helper). 1123 * 1124 * This is flexible enough to be used in several ways: 1125 * 1126 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1127 * checksum, it can be used when pushing new data. 1128 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1129 * checksum, it can be used when removing data from a packet. 1130 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1131 * can be used to compute a diff. Note that *from_size* and 1132 * *to_size* do not need to be equal. 1133 * 1134 * This helper can be used in combination with 1135 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1136 * which one can feed in the difference computed with 1137 * **bpf_csum_diff**\ (). 1138 * Return 1139 * The checksum result, or a negative error code in case of 1140 * failure. 1141 * 1142 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1143 * Description 1144 * Retrieve tunnel options metadata for the packet associated to 1145 * *skb*, and store the raw tunnel option data to the buffer *opt* 1146 * of *size*. 1147 * 1148 * This helper can be used with encapsulation devices that can 1149 * operate in "collect metadata" mode (please refer to the related 1150 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1151 * more details). A particular example where this can be used is 1152 * in combination with the Geneve encapsulation protocol, where it 1153 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1154 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1155 * the eBPF program. This allows for full customization of these 1156 * headers. 1157 * Return 1158 * The size of the option data retrieved. 1159 * 1160 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1161 * Description 1162 * Set tunnel options metadata for the packet associated to *skb* 1163 * to the option data contained in the raw buffer *opt* of *size*. 1164 * 1165 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1166 * helper for additional information. 1167 * Return 1168 * 0 on success, or a negative error in case of failure. 1169 * 1170 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1171 * Description 1172 * Change the protocol of the *skb* to *proto*. Currently 1173 * supported are transition from IPv4 to IPv6, and from IPv6 to 1174 * IPv4. The helper takes care of the groundwork for the 1175 * transition, including resizing the socket buffer. The eBPF 1176 * program is expected to fill the new headers, if any, via 1177 * **skb_store_bytes**\ () and to recompute the checksums with 1178 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1179 * (). The main case for this helper is to perform NAT64 1180 * operations out of an eBPF program. 1181 * 1182 * Internally, the GSO type is marked as dodgy so that headers are 1183 * checked and segments are recalculated by the GSO/GRO engine. 1184 * The size for GSO target is adapted as well. 1185 * 1186 * All values for *flags* are reserved for future usage, and must 1187 * be left at zero. 1188 * 1189 * A call to this helper is susceptible to change the underlying 1190 * packet buffer. Therefore, at load time, all checks on pointers 1191 * previously done by the verifier are invalidated and must be 1192 * performed again, if the helper is used in combination with 1193 * direct packet access. 1194 * Return 1195 * 0 on success, or a negative error in case of failure. 1196 * 1197 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1198 * Description 1199 * Change the packet type for the packet associated to *skb*. This 1200 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1201 * the eBPF program does not have a write access to *skb*\ 1202 * **->pkt_type** beside this helper. Using a helper here allows 1203 * for graceful handling of errors. 1204 * 1205 * The major use case is to change incoming *skb*s to 1206 * **PACKET_HOST** in a programmatic way instead of having to 1207 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1208 * example. 1209 * 1210 * Note that *type* only allows certain values. At this time, they 1211 * are: 1212 * 1213 * **PACKET_HOST** 1214 * Packet is for us. 1215 * **PACKET_BROADCAST** 1216 * Send packet to all. 1217 * **PACKET_MULTICAST** 1218 * Send packet to group. 1219 * **PACKET_OTHERHOST** 1220 * Send packet to someone else. 1221 * Return 1222 * 0 on success, or a negative error in case of failure. 1223 * 1224 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1225 * Description 1226 * Check whether *skb* is a descendant of the cgroup2 held by 1227 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1228 * Return 1229 * The return value depends on the result of the test, and can be: 1230 * 1231 * * 0, if the *skb* failed the cgroup2 descendant test. 1232 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1233 * * A negative error code, if an error occurred. 1234 * 1235 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1236 * Description 1237 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1238 * not set, in particular if the hash was cleared due to mangling, 1239 * recompute this hash. Later accesses to the hash can be done 1240 * directly with *skb*\ **->hash**. 1241 * 1242 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1243 * prototype with **bpf_skb_change_proto**\ (), or calling 1244 * **bpf_skb_store_bytes**\ () with the 1245 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1246 * the hash and to trigger a new computation for the next call to 1247 * **bpf_get_hash_recalc**\ (). 1248 * Return 1249 * The 32-bit hash. 1250 * 1251 * u64 bpf_get_current_task(void) 1252 * Return 1253 * A pointer to the current task struct. 1254 * 1255 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1256 * Description 1257 * Attempt in a safe way to write *len* bytes from the buffer 1258 * *src* to *dst* in memory. It only works for threads that are in 1259 * user context, and *dst* must be a valid user space address. 1260 * 1261 * This helper should not be used to implement any kind of 1262 * security mechanism because of TOC-TOU attacks, but rather to 1263 * debug, divert, and manipulate execution of semi-cooperative 1264 * processes. 1265 * 1266 * Keep in mind that this feature is meant for experiments, and it 1267 * has a risk of crashing the system and running programs. 1268 * Therefore, when an eBPF program using this helper is attached, 1269 * a warning including PID and process name is printed to kernel 1270 * logs. 1271 * Return 1272 * 0 on success, or a negative error in case of failure. 1273 * 1274 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1275 * Description 1276 * Check whether the probe is being run is the context of a given 1277 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1278 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1279 * Return 1280 * The return value depends on the result of the test, and can be: 1281 * 1282 * * 0, if the *skb* task belongs to the cgroup2. 1283 * * 1, if the *skb* task does not belong to the cgroup2. 1284 * * A negative error code, if an error occurred. 1285 * 1286 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1287 * Description 1288 * Resize (trim or grow) the packet associated to *skb* to the 1289 * new *len*. The *flags* are reserved for future usage, and must 1290 * be left at zero. 1291 * 1292 * The basic idea is that the helper performs the needed work to 1293 * change the size of the packet, then the eBPF program rewrites 1294 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1295 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1296 * and others. This helper is a slow path utility intended for 1297 * replies with control messages. And because it is targeted for 1298 * slow path, the helper itself can afford to be slow: it 1299 * implicitly linearizes, unclones and drops offloads from the 1300 * *skb*. 1301 * 1302 * A call to this helper is susceptible to change the underlying 1303 * packet buffer. Therefore, at load time, all checks on pointers 1304 * previously done by the verifier are invalidated and must be 1305 * performed again, if the helper is used in combination with 1306 * direct packet access. 1307 * Return 1308 * 0 on success, or a negative error in case of failure. 1309 * 1310 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1311 * Description 1312 * Pull in non-linear data in case the *skb* is non-linear and not 1313 * all of *len* are part of the linear section. Make *len* bytes 1314 * from *skb* readable and writable. If a zero value is passed for 1315 * *len*, then the whole length of the *skb* is pulled. 1316 * 1317 * This helper is only needed for reading and writing with direct 1318 * packet access. 1319 * 1320 * For direct packet access, testing that offsets to access 1321 * are within packet boundaries (test on *skb*\ **->data_end**) is 1322 * susceptible to fail if offsets are invalid, or if the requested 1323 * data is in non-linear parts of the *skb*. On failure the 1324 * program can just bail out, or in the case of a non-linear 1325 * buffer, use a helper to make the data available. The 1326 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1327 * the data. Another one consists in using **bpf_skb_pull_data** 1328 * to pull in once the non-linear parts, then retesting and 1329 * eventually access the data. 1330 * 1331 * At the same time, this also makes sure the *skb* is uncloned, 1332 * which is a necessary condition for direct write. As this needs 1333 * to be an invariant for the write part only, the verifier 1334 * detects writes and adds a prologue that is calling 1335 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1336 * the very beginning in case it is indeed cloned. 1337 * 1338 * A call to this helper is susceptible to change the underlying 1339 * packet buffer. Therefore, at load time, all checks on pointers 1340 * previously done by the verifier are invalidated and must be 1341 * performed again, if the helper is used in combination with 1342 * direct packet access. 1343 * Return 1344 * 0 on success, or a negative error in case of failure. 1345 * 1346 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1347 * Description 1348 * Add the checksum *csum* into *skb*\ **->csum** in case the 1349 * driver has supplied a checksum for the entire packet into that 1350 * field. Return an error otherwise. This helper is intended to be 1351 * used in combination with **bpf_csum_diff**\ (), in particular 1352 * when the checksum needs to be updated after data has been 1353 * written into the packet through direct packet access. 1354 * Return 1355 * The checksum on success, or a negative error code in case of 1356 * failure. 1357 * 1358 * void bpf_set_hash_invalid(struct sk_buff *skb) 1359 * Description 1360 * Invalidate the current *skb*\ **->hash**. It can be used after 1361 * mangling on headers through direct packet access, in order to 1362 * indicate that the hash is outdated and to trigger a 1363 * recalculation the next time the kernel tries to access this 1364 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1365 * 1366 * int bpf_get_numa_node_id(void) 1367 * Description 1368 * Return the id of the current NUMA node. The primary use case 1369 * for this helper is the selection of sockets for the local NUMA 1370 * node, when the program is attached to sockets using the 1371 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1372 * but the helper is also available to other eBPF program types, 1373 * similarly to **bpf_get_smp_processor_id**\ (). 1374 * Return 1375 * The id of current NUMA node. 1376 * 1377 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1378 * Description 1379 * Grows headroom of packet associated to *skb* and adjusts the 1380 * offset of the MAC header accordingly, adding *len* bytes of 1381 * space. It automatically extends and reallocates memory as 1382 * required. 1383 * 1384 * This helper can be used on a layer 3 *skb* to push a MAC header 1385 * for redirection into a layer 2 device. 1386 * 1387 * All values for *flags* are reserved for future usage, and must 1388 * be left at zero. 1389 * 1390 * A call to this helper is susceptible to change the underlying 1391 * packet buffer. Therefore, at load time, all checks on pointers 1392 * previously done by the verifier are invalidated and must be 1393 * performed again, if the helper is used in combination with 1394 * direct packet access. 1395 * Return 1396 * 0 on success, or a negative error in case of failure. 1397 * 1398 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1399 * Description 1400 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1401 * it is possible to use a negative value for *delta*. This helper 1402 * can be used to prepare the packet for pushing or popping 1403 * headers. 1404 * 1405 * A call to this helper is susceptible to change the underlying 1406 * packet buffer. Therefore, at load time, all checks on pointers 1407 * previously done by the verifier are invalidated and must be 1408 * performed again, if the helper is used in combination with 1409 * direct packet access. 1410 * Return 1411 * 0 on success, or a negative error in case of failure. 1412 * 1413 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1414 * Description 1415 * Copy a NUL terminated string from an unsafe address 1416 * *unsafe_ptr* to *dst*. The *size* should include the 1417 * terminating NUL byte. In case the string length is smaller than 1418 * *size*, the target is not padded with further NUL bytes. If the 1419 * string length is larger than *size*, just *size*-1 bytes are 1420 * copied and the last byte is set to NUL. 1421 * 1422 * On success, the length of the copied string is returned. This 1423 * makes this helper useful in tracing programs for reading 1424 * strings, and more importantly to get its length at runtime. See 1425 * the following snippet: 1426 * 1427 * :: 1428 * 1429 * SEC("kprobe/sys_open") 1430 * void bpf_sys_open(struct pt_regs *ctx) 1431 * { 1432 * char buf[PATHLEN]; // PATHLEN is defined to 256 1433 * int res = bpf_probe_read_str(buf, sizeof(buf), 1434 * ctx->di); 1435 * 1436 * // Consume buf, for example push it to 1437 * // userspace via bpf_perf_event_output(); we 1438 * // can use res (the string length) as event 1439 * // size, after checking its boundaries. 1440 * } 1441 * 1442 * In comparison, using **bpf_probe_read()** helper here instead 1443 * to read the string would require to estimate the length at 1444 * compile time, and would often result in copying more memory 1445 * than necessary. 1446 * 1447 * Another useful use case is when parsing individual process 1448 * arguments or individual environment variables navigating 1449 * *current*\ **->mm->arg_start** and *current*\ 1450 * **->mm->env_start**: using this helper and the return value, 1451 * one can quickly iterate at the right offset of the memory area. 1452 * Return 1453 * On success, the strictly positive length of the string, 1454 * including the trailing NUL character. On error, a negative 1455 * value. 1456 * 1457 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1458 * Description 1459 * If the **struct sk_buff** pointed by *skb* has a known socket, 1460 * retrieve the cookie (generated by the kernel) of this socket. 1461 * If no cookie has been set yet, generate a new cookie. Once 1462 * generated, the socket cookie remains stable for the life of the 1463 * socket. This helper can be useful for monitoring per socket 1464 * networking traffic statistics as it provides a unique socket 1465 * identifier per namespace. 1466 * Return 1467 * A 8-byte long non-decreasing number on success, or 0 if the 1468 * socket field is missing inside *skb*. 1469 * 1470 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1471 * Description 1472 * Equivalent to bpf_get_socket_cookie() helper that accepts 1473 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1474 * Return 1475 * A 8-byte long non-decreasing number. 1476 * 1477 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1478 * Description 1479 * Equivalent to bpf_get_socket_cookie() helper that accepts 1480 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1481 * Return 1482 * A 8-byte long non-decreasing number. 1483 * 1484 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1485 * Return 1486 * The owner UID of the socket associated to *skb*. If the socket 1487 * is **NULL**, or if it is not a full socket (i.e. if it is a 1488 * time-wait or a request socket instead), **overflowuid** value 1489 * is returned (note that **overflowuid** might also be the actual 1490 * UID value for the socket). 1491 * 1492 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1493 * Description 1494 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1495 * to value *hash*. 1496 * Return 1497 * 0 1498 * 1499 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1500 * Description 1501 * Emulate a call to **setsockopt()** on the socket associated to 1502 * *bpf_socket*, which must be a full socket. The *level* at 1503 * which the option resides and the name *optname* of the option 1504 * must be specified, see **setsockopt(2)** for more information. 1505 * The option value of length *optlen* is pointed by *optval*. 1506 * 1507 * This helper actually implements a subset of **setsockopt()**. 1508 * It supports the following *level*\ s: 1509 * 1510 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1511 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1512 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1513 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1514 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1515 * **TCP_BPF_SNDCWND_CLAMP**. 1516 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1517 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1518 * Return 1519 * 0 on success, or a negative error in case of failure. 1520 * 1521 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1522 * Description 1523 * Grow or shrink the room for data in the packet associated to 1524 * *skb* by *len_diff*, and according to the selected *mode*. 1525 * 1526 * There are two supported modes at this time: 1527 * 1528 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1529 * (room space is added or removed below the layer 2 header). 1530 * 1531 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1532 * (room space is added or removed below the layer 3 header). 1533 * 1534 * The following flags are supported at this time: 1535 * 1536 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1537 * Adjusting mss in this way is not allowed for datagrams. 1538 * 1539 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1540 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1541 * Any new space is reserved to hold a tunnel header. 1542 * Configure skb offsets and other fields accordingly. 1543 * 1544 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1545 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1546 * Use with ENCAP_L3 flags to further specify the tunnel type. 1547 * 1548 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1549 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1550 * type; *len* is the length of the inner MAC header. 1551 * 1552 * A call to this helper is susceptible to change the underlying 1553 * packet buffer. Therefore, at load time, all checks on pointers 1554 * previously done by the verifier are invalidated and must be 1555 * performed again, if the helper is used in combination with 1556 * direct packet access. 1557 * Return 1558 * 0 on success, or a negative error in case of failure. 1559 * 1560 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1561 * Description 1562 * Redirect the packet to the endpoint referenced by *map* at 1563 * index *key*. Depending on its type, this *map* can contain 1564 * references to net devices (for forwarding packets through other 1565 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1566 * but this is only implemented for native XDP (with driver 1567 * support) as of this writing). 1568 * 1569 * All values for *flags* are reserved for future usage, and must 1570 * be left at zero. 1571 * 1572 * When used to redirect packets to net devices, this helper 1573 * provides a high performance increase over **bpf_redirect**\ (). 1574 * This is due to various implementation details of the underlying 1575 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1576 * () tries to send packet as a "bulk" to the device. 1577 * Return 1578 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1579 * 1580 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1581 * Description 1582 * Redirect the packet to the socket referenced by *map* (of type 1583 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1584 * egress interfaces can be used for redirection. The 1585 * **BPF_F_INGRESS** value in *flags* is used to make the 1586 * distinction (ingress path is selected if the flag is present, 1587 * egress path otherwise). This is the only flag supported for now. 1588 * Return 1589 * **SK_PASS** on success, or **SK_DROP** on error. 1590 * 1591 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1592 * Description 1593 * Add an entry to, or update a *map* referencing sockets. The 1594 * *skops* is used as a new value for the entry associated to 1595 * *key*. *flags* is one of: 1596 * 1597 * **BPF_NOEXIST** 1598 * The entry for *key* must not exist in the map. 1599 * **BPF_EXIST** 1600 * The entry for *key* must already exist in the map. 1601 * **BPF_ANY** 1602 * No condition on the existence of the entry for *key*. 1603 * 1604 * If the *map* has eBPF programs (parser and verdict), those will 1605 * be inherited by the socket being added. If the socket is 1606 * already attached to eBPF programs, this results in an error. 1607 * Return 1608 * 0 on success, or a negative error in case of failure. 1609 * 1610 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1611 * Description 1612 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1613 * *delta* (which can be positive or negative). Note that this 1614 * operation modifies the address stored in *xdp_md*\ **->data**, 1615 * so the latter must be loaded only after the helper has been 1616 * called. 1617 * 1618 * The use of *xdp_md*\ **->data_meta** is optional and programs 1619 * are not required to use it. The rationale is that when the 1620 * packet is processed with XDP (e.g. as DoS filter), it is 1621 * possible to push further meta data along with it before passing 1622 * to the stack, and to give the guarantee that an ingress eBPF 1623 * program attached as a TC classifier on the same device can pick 1624 * this up for further post-processing. Since TC works with socket 1625 * buffers, it remains possible to set from XDP the **mark** or 1626 * **priority** pointers, or other pointers for the socket buffer. 1627 * Having this scratch space generic and programmable allows for 1628 * more flexibility as the user is free to store whatever meta 1629 * data they need. 1630 * 1631 * A call to this helper is susceptible to change the underlying 1632 * packet buffer. Therefore, at load time, all checks on pointers 1633 * previously done by the verifier are invalidated and must be 1634 * performed again, if the helper is used in combination with 1635 * direct packet access. 1636 * Return 1637 * 0 on success, or a negative error in case of failure. 1638 * 1639 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1640 * Description 1641 * Read the value of a perf event counter, and store it into *buf* 1642 * of size *buf_size*. This helper relies on a *map* of type 1643 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1644 * counter is selected when *map* is updated with perf event file 1645 * descriptors. The *map* is an array whose size is the number of 1646 * available CPUs, and each cell contains a value relative to one 1647 * CPU. The value to retrieve is indicated by *flags*, that 1648 * contains the index of the CPU to look up, masked with 1649 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1650 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1651 * current CPU should be retrieved. 1652 * 1653 * This helper behaves in a way close to 1654 * **bpf_perf_event_read**\ () helper, save that instead of 1655 * just returning the value observed, it fills the *buf* 1656 * structure. This allows for additional data to be retrieved: in 1657 * particular, the enabled and running times (in *buf*\ 1658 * **->enabled** and *buf*\ **->running**, respectively) are 1659 * copied. In general, **bpf_perf_event_read_value**\ () is 1660 * recommended over **bpf_perf_event_read**\ (), which has some 1661 * ABI issues and provides fewer functionalities. 1662 * 1663 * These values are interesting, because hardware PMU (Performance 1664 * Monitoring Unit) counters are limited resources. When there are 1665 * more PMU based perf events opened than available counters, 1666 * kernel will multiplex these events so each event gets certain 1667 * percentage (but not all) of the PMU time. In case that 1668 * multiplexing happens, the number of samples or counter value 1669 * will not reflect the case compared to when no multiplexing 1670 * occurs. This makes comparison between different runs difficult. 1671 * Typically, the counter value should be normalized before 1672 * comparing to other experiments. The usual normalization is done 1673 * as follows. 1674 * 1675 * :: 1676 * 1677 * normalized_counter = counter * t_enabled / t_running 1678 * 1679 * Where t_enabled is the time enabled for event and t_running is 1680 * the time running for event since last normalization. The 1681 * enabled and running times are accumulated since the perf event 1682 * open. To achieve scaling factor between two invocations of an 1683 * eBPF program, users can can use CPU id as the key (which is 1684 * typical for perf array usage model) to remember the previous 1685 * value and do the calculation inside the eBPF program. 1686 * Return 1687 * 0 on success, or a negative error in case of failure. 1688 * 1689 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1690 * Description 1691 * For en eBPF program attached to a perf event, retrieve the 1692 * value of the event counter associated to *ctx* and store it in 1693 * the structure pointed by *buf* and of size *buf_size*. Enabled 1694 * and running times are also stored in the structure (see 1695 * description of helper **bpf_perf_event_read_value**\ () for 1696 * more details). 1697 * Return 1698 * 0 on success, or a negative error in case of failure. 1699 * 1700 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1701 * Description 1702 * Emulate a call to **getsockopt()** on the socket associated to 1703 * *bpf_socket*, which must be a full socket. The *level* at 1704 * which the option resides and the name *optname* of the option 1705 * must be specified, see **getsockopt(2)** for more information. 1706 * The retrieved value is stored in the structure pointed by 1707 * *opval* and of length *optlen*. 1708 * 1709 * This helper actually implements a subset of **getsockopt()**. 1710 * It supports the following *level*\ s: 1711 * 1712 * * **IPPROTO_TCP**, which supports *optname* 1713 * **TCP_CONGESTION**. 1714 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1715 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1716 * Return 1717 * 0 on success, or a negative error in case of failure. 1718 * 1719 * int bpf_override_return(struct pt_reg *regs, u64 rc) 1720 * Description 1721 * Used for error injection, this helper uses kprobes to override 1722 * the return value of the probed function, and to set it to *rc*. 1723 * The first argument is the context *regs* on which the kprobe 1724 * works. 1725 * 1726 * This helper works by setting setting the PC (program counter) 1727 * to an override function which is run in place of the original 1728 * probed function. This means the probed function is not run at 1729 * all. The replacement function just returns with the required 1730 * value. 1731 * 1732 * This helper has security implications, and thus is subject to 1733 * restrictions. It is only available if the kernel was compiled 1734 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1735 * option, and in this case it only works on functions tagged with 1736 * **ALLOW_ERROR_INJECTION** in the kernel code. 1737 * 1738 * Also, the helper is only available for the architectures having 1739 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1740 * x86 architecture is the only one to support this feature. 1741 * Return 1742 * 0 1743 * 1744 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1745 * Description 1746 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1747 * for the full TCP socket associated to *bpf_sock_ops* to 1748 * *argval*. 1749 * 1750 * The primary use of this field is to determine if there should 1751 * be calls to eBPF programs of type 1752 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1753 * code. A program of the same type can change its value, per 1754 * connection and as necessary, when the connection is 1755 * established. This field is directly accessible for reading, but 1756 * this helper must be used for updates in order to return an 1757 * error if an eBPF program tries to set a callback that is not 1758 * supported in the current kernel. 1759 * 1760 * *argval* is a flag array which can combine these flags: 1761 * 1762 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1763 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1764 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1765 * 1766 * Therefore, this function can be used to clear a callback flag by 1767 * setting the appropriate bit to zero. e.g. to disable the RTO 1768 * callback: 1769 * 1770 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1771 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1772 * 1773 * Here are some examples of where one could call such eBPF 1774 * program: 1775 * 1776 * * When RTO fires. 1777 * * When a packet is retransmitted. 1778 * * When the connection terminates. 1779 * * When a packet is sent. 1780 * * When a packet is received. 1781 * Return 1782 * Code **-EINVAL** if the socket is not a full TCP socket; 1783 * otherwise, a positive number containing the bits that could not 1784 * be set is returned (which comes down to 0 if all bits were set 1785 * as required). 1786 * 1787 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1788 * Description 1789 * This helper is used in programs implementing policies at the 1790 * socket level. If the message *msg* is allowed to pass (i.e. if 1791 * the verdict eBPF program returns **SK_PASS**), redirect it to 1792 * the socket referenced by *map* (of type 1793 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1794 * egress interfaces can be used for redirection. The 1795 * **BPF_F_INGRESS** value in *flags* is used to make the 1796 * distinction (ingress path is selected if the flag is present, 1797 * egress path otherwise). This is the only flag supported for now. 1798 * Return 1799 * **SK_PASS** on success, or **SK_DROP** on error. 1800 * 1801 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1802 * Description 1803 * For socket policies, apply the verdict of the eBPF program to 1804 * the next *bytes* (number of bytes) of message *msg*. 1805 * 1806 * For example, this helper can be used in the following cases: 1807 * 1808 * * A single **sendmsg**\ () or **sendfile**\ () system call 1809 * contains multiple logical messages that the eBPF program is 1810 * supposed to read and for which it should apply a verdict. 1811 * * An eBPF program only cares to read the first *bytes* of a 1812 * *msg*. If the message has a large payload, then setting up 1813 * and calling the eBPF program repeatedly for all bytes, even 1814 * though the verdict is already known, would create unnecessary 1815 * overhead. 1816 * 1817 * When called from within an eBPF program, the helper sets a 1818 * counter internal to the BPF infrastructure, that is used to 1819 * apply the last verdict to the next *bytes*. If *bytes* is 1820 * smaller than the current data being processed from a 1821 * **sendmsg**\ () or **sendfile**\ () system call, the first 1822 * *bytes* will be sent and the eBPF program will be re-run with 1823 * the pointer for start of data pointing to byte number *bytes* 1824 * **+ 1**. If *bytes* is larger than the current data being 1825 * processed, then the eBPF verdict will be applied to multiple 1826 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1827 * consumed. 1828 * 1829 * Note that if a socket closes with the internal counter holding 1830 * a non-zero value, this is not a problem because data is not 1831 * being buffered for *bytes* and is sent as it is received. 1832 * Return 1833 * 0 1834 * 1835 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1836 * Description 1837 * For socket policies, prevent the execution of the verdict eBPF 1838 * program for message *msg* until *bytes* (byte number) have been 1839 * accumulated. 1840 * 1841 * This can be used when one needs a specific number of bytes 1842 * before a verdict can be assigned, even if the data spans 1843 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1844 * case would be a user calling **sendmsg**\ () repeatedly with 1845 * 1-byte long message segments. Obviously, this is bad for 1846 * performance, but it is still valid. If the eBPF program needs 1847 * *bytes* bytes to validate a header, this helper can be used to 1848 * prevent the eBPF program to be called again until *bytes* have 1849 * been accumulated. 1850 * Return 1851 * 0 1852 * 1853 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1854 * Description 1855 * For socket policies, pull in non-linear data from user space 1856 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1857 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1858 * respectively. 1859 * 1860 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1861 * *msg* it can only parse data that the (**data**, **data_end**) 1862 * pointers have already consumed. For **sendmsg**\ () hooks this 1863 * is likely the first scatterlist element. But for calls relying 1864 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1865 * be the range (**0**, **0**) because the data is shared with 1866 * user space and by default the objective is to avoid allowing 1867 * user space to modify data while (or after) eBPF verdict is 1868 * being decided. This helper can be used to pull in data and to 1869 * set the start and end pointer to given values. Data will be 1870 * copied if necessary (i.e. if data was not linear and if start 1871 * and end pointers do not point to the same chunk). 1872 * 1873 * A call to this helper is susceptible to change the underlying 1874 * packet buffer. Therefore, at load time, all checks on pointers 1875 * previously done by the verifier are invalidated and must be 1876 * performed again, if the helper is used in combination with 1877 * direct packet access. 1878 * 1879 * All values for *flags* are reserved for future usage, and must 1880 * be left at zero. 1881 * Return 1882 * 0 on success, or a negative error in case of failure. 1883 * 1884 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1885 * Description 1886 * Bind the socket associated to *ctx* to the address pointed by 1887 * *addr*, of length *addr_len*. This allows for making outgoing 1888 * connection from the desired IP address, which can be useful for 1889 * example when all processes inside a cgroup should use one 1890 * single IP address on a host that has multiple IP configured. 1891 * 1892 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1893 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1894 * **AF_INET6**). Looking for a free port to bind to can be 1895 * expensive, therefore binding to port is not permitted by the 1896 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1897 * must be set to zero. 1898 * Return 1899 * 0 on success, or a negative error in case of failure. 1900 * 1901 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1902 * Description 1903 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1904 * only possible to shrink the packet as of this writing, 1905 * therefore *delta* must be a negative integer. 1906 * 1907 * A call to this helper is susceptible to change the underlying 1908 * packet buffer. Therefore, at load time, all checks on pointers 1909 * previously done by the verifier are invalidated and must be 1910 * performed again, if the helper is used in combination with 1911 * direct packet access. 1912 * Return 1913 * 0 on success, or a negative error in case of failure. 1914 * 1915 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1916 * Description 1917 * Retrieve the XFRM state (IP transform framework, see also 1918 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1919 * 1920 * The retrieved value is stored in the **struct bpf_xfrm_state** 1921 * pointed by *xfrm_state* and of length *size*. 1922 * 1923 * All values for *flags* are reserved for future usage, and must 1924 * be left at zero. 1925 * 1926 * This helper is available only if the kernel was compiled with 1927 * **CONFIG_XFRM** configuration option. 1928 * Return 1929 * 0 on success, or a negative error in case of failure. 1930 * 1931 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1932 * Description 1933 * Return a user or a kernel stack in bpf program provided buffer. 1934 * To achieve this, the helper needs *ctx*, which is a pointer 1935 * to the context on which the tracing program is executed. 1936 * To store the stacktrace, the bpf program provides *buf* with 1937 * a nonnegative *size*. 1938 * 1939 * The last argument, *flags*, holds the number of stack frames to 1940 * skip (from 0 to 255), masked with 1941 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1942 * the following flags: 1943 * 1944 * **BPF_F_USER_STACK** 1945 * Collect a user space stack instead of a kernel stack. 1946 * **BPF_F_USER_BUILD_ID** 1947 * Collect buildid+offset instead of ips for user stack, 1948 * only valid if **BPF_F_USER_STACK** is also specified. 1949 * 1950 * **bpf_get_stack**\ () can collect up to 1951 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1952 * to sufficient large buffer size. Note that 1953 * this limit can be controlled with the **sysctl** program, and 1954 * that it should be manually increased in order to profile long 1955 * user stacks (such as stacks for Java programs). To do so, use: 1956 * 1957 * :: 1958 * 1959 * # sysctl kernel.perf_event_max_stack=<new value> 1960 * Return 1961 * A non-negative value equal to or less than *size* on success, 1962 * or a negative error in case of failure. 1963 * 1964 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1965 * Description 1966 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1967 * it provides an easy way to load *len* bytes from *offset* 1968 * from the packet associated to *skb*, into the buffer pointed 1969 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1970 * a fifth argument *start_header* exists in order to select a 1971 * base offset to start from. *start_header* can be one of: 1972 * 1973 * **BPF_HDR_START_MAC** 1974 * Base offset to load data from is *skb*'s mac header. 1975 * **BPF_HDR_START_NET** 1976 * Base offset to load data from is *skb*'s network header. 1977 * 1978 * In general, "direct packet access" is the preferred method to 1979 * access packet data, however, this helper is in particular useful 1980 * in socket filters where *skb*\ **->data** does not always point 1981 * to the start of the mac header and where "direct packet access" 1982 * is not available. 1983 * Return 1984 * 0 on success, or a negative error in case of failure. 1985 * 1986 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 1987 * Description 1988 * Do FIB lookup in kernel tables using parameters in *params*. 1989 * If lookup is successful and result shows packet is to be 1990 * forwarded, the neighbor tables are searched for the nexthop. 1991 * If successful (ie., FIB lookup shows forwarding and nexthop 1992 * is resolved), the nexthop address is returned in ipv4_dst 1993 * or ipv6_dst based on family, smac is set to mac address of 1994 * egress device, dmac is set to nexthop mac address, rt_metric 1995 * is set to metric from route (IPv4/IPv6 only), and ifindex 1996 * is set to the device index of the nexthop from the FIB lookup. 1997 * 1998 * *plen* argument is the size of the passed in struct. 1999 * *flags* argument can be a combination of one or more of the 2000 * following values: 2001 * 2002 * **BPF_FIB_LOOKUP_DIRECT** 2003 * Do a direct table lookup vs full lookup using FIB 2004 * rules. 2005 * **BPF_FIB_LOOKUP_OUTPUT** 2006 * Perform lookup from an egress perspective (default is 2007 * ingress). 2008 * 2009 * *ctx* is either **struct xdp_md** for XDP programs or 2010 * **struct sk_buff** tc cls_act programs. 2011 * Return 2012 * * < 0 if any input argument is invalid 2013 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2014 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2015 * packet is not forwarded or needs assist from full stack 2016 * 2017 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 2018 * Description 2019 * Add an entry to, or update a sockhash *map* referencing sockets. 2020 * The *skops* is used as a new value for the entry associated to 2021 * *key*. *flags* is one of: 2022 * 2023 * **BPF_NOEXIST** 2024 * The entry for *key* must not exist in the map. 2025 * **BPF_EXIST** 2026 * The entry for *key* must already exist in the map. 2027 * **BPF_ANY** 2028 * No condition on the existence of the entry for *key*. 2029 * 2030 * If the *map* has eBPF programs (parser and verdict), those will 2031 * be inherited by the socket being added. If the socket is 2032 * already attached to eBPF programs, this results in an error. 2033 * Return 2034 * 0 on success, or a negative error in case of failure. 2035 * 2036 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2037 * Description 2038 * This helper is used in programs implementing policies at the 2039 * socket level. If the message *msg* is allowed to pass (i.e. if 2040 * the verdict eBPF program returns **SK_PASS**), redirect it to 2041 * the socket referenced by *map* (of type 2042 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2043 * egress interfaces can be used for redirection. The 2044 * **BPF_F_INGRESS** value in *flags* is used to make the 2045 * distinction (ingress path is selected if the flag is present, 2046 * egress path otherwise). This is the only flag supported for now. 2047 * Return 2048 * **SK_PASS** on success, or **SK_DROP** on error. 2049 * 2050 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2051 * Description 2052 * This helper is used in programs implementing policies at the 2053 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2054 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2055 * to the socket referenced by *map* (of type 2056 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2057 * egress interfaces can be used for redirection. The 2058 * **BPF_F_INGRESS** value in *flags* is used to make the 2059 * distinction (ingress path is selected if the flag is present, 2060 * egress otherwise). This is the only flag supported for now. 2061 * Return 2062 * **SK_PASS** on success, or **SK_DROP** on error. 2063 * 2064 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2065 * Description 2066 * Encapsulate the packet associated to *skb* within a Layer 3 2067 * protocol header. This header is provided in the buffer at 2068 * address *hdr*, with *len* its size in bytes. *type* indicates 2069 * the protocol of the header and can be one of: 2070 * 2071 * **BPF_LWT_ENCAP_SEG6** 2072 * IPv6 encapsulation with Segment Routing Header 2073 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2074 * the IPv6 header is computed by the kernel. 2075 * **BPF_LWT_ENCAP_SEG6_INLINE** 2076 * Only works if *skb* contains an IPv6 packet. Insert a 2077 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2078 * the IPv6 header. 2079 * **BPF_LWT_ENCAP_IP** 2080 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2081 * must be IPv4 or IPv6, followed by zero or more 2082 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2083 * total bytes in all prepended headers. Please note that 2084 * if **skb_is_gso**\ (*skb*) is true, no more than two 2085 * headers can be prepended, and the inner header, if 2086 * present, should be either GRE or UDP/GUE. 2087 * 2088 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2089 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2090 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2091 * **BPF_PROG_TYPE_LWT_XMIT**. 2092 * 2093 * A call to this helper is susceptible to change the underlying 2094 * packet buffer. Therefore, at load time, all checks on pointers 2095 * previously done by the verifier are invalidated and must be 2096 * performed again, if the helper is used in combination with 2097 * direct packet access. 2098 * Return 2099 * 0 on success, or a negative error in case of failure. 2100 * 2101 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2102 * Description 2103 * Store *len* bytes from address *from* into the packet 2104 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2105 * inside the outermost IPv6 Segment Routing Header can be 2106 * modified through this helper. 2107 * 2108 * A call to this helper is susceptible to change the underlying 2109 * packet buffer. Therefore, at load time, all checks on pointers 2110 * previously done by the verifier are invalidated and must be 2111 * performed again, if the helper is used in combination with 2112 * direct packet access. 2113 * Return 2114 * 0 on success, or a negative error in case of failure. 2115 * 2116 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2117 * Description 2118 * Adjust the size allocated to TLVs in the outermost IPv6 2119 * Segment Routing Header contained in the packet associated to 2120 * *skb*, at position *offset* by *delta* bytes. Only offsets 2121 * after the segments are accepted. *delta* can be as well 2122 * positive (growing) as negative (shrinking). 2123 * 2124 * A call to this helper is susceptible to change the underlying 2125 * packet buffer. Therefore, at load time, all checks on pointers 2126 * previously done by the verifier are invalidated and must be 2127 * performed again, if the helper is used in combination with 2128 * direct packet access. 2129 * Return 2130 * 0 on success, or a negative error in case of failure. 2131 * 2132 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2133 * Description 2134 * Apply an IPv6 Segment Routing action of type *action* to the 2135 * packet associated to *skb*. Each action takes a parameter 2136 * contained at address *param*, and of length *param_len* bytes. 2137 * *action* can be one of: 2138 * 2139 * **SEG6_LOCAL_ACTION_END_X** 2140 * End.X action: Endpoint with Layer-3 cross-connect. 2141 * Type of *param*: **struct in6_addr**. 2142 * **SEG6_LOCAL_ACTION_END_T** 2143 * End.T action: Endpoint with specific IPv6 table lookup. 2144 * Type of *param*: **int**. 2145 * **SEG6_LOCAL_ACTION_END_B6** 2146 * End.B6 action: Endpoint bound to an SRv6 policy. 2147 * Type of *param*: **struct ipv6_sr_hdr**. 2148 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2149 * End.B6.Encap action: Endpoint bound to an SRv6 2150 * encapsulation policy. 2151 * Type of *param*: **struct ipv6_sr_hdr**. 2152 * 2153 * A call to this helper is susceptible to change the underlying 2154 * packet buffer. Therefore, at load time, all checks on pointers 2155 * previously done by the verifier are invalidated and must be 2156 * performed again, if the helper is used in combination with 2157 * direct packet access. 2158 * Return 2159 * 0 on success, or a negative error in case of failure. 2160 * 2161 * int bpf_rc_repeat(void *ctx) 2162 * Description 2163 * This helper is used in programs implementing IR decoding, to 2164 * report a successfully decoded repeat key message. This delays 2165 * the generation of a key up event for previously generated 2166 * key down event. 2167 * 2168 * Some IR protocols like NEC have a special IR message for 2169 * repeating last button, for when a button is held down. 2170 * 2171 * The *ctx* should point to the lirc sample as passed into 2172 * the program. 2173 * 2174 * This helper is only available is the kernel was compiled with 2175 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2176 * "**y**". 2177 * Return 2178 * 0 2179 * 2180 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2181 * Description 2182 * This helper is used in programs implementing IR decoding, to 2183 * report a successfully decoded key press with *scancode*, 2184 * *toggle* value in the given *protocol*. The scancode will be 2185 * translated to a keycode using the rc keymap, and reported as 2186 * an input key down event. After a period a key up event is 2187 * generated. This period can be extended by calling either 2188 * **bpf_rc_keydown**\ () again with the same values, or calling 2189 * **bpf_rc_repeat**\ (). 2190 * 2191 * Some protocols include a toggle bit, in case the button was 2192 * released and pressed again between consecutive scancodes. 2193 * 2194 * The *ctx* should point to the lirc sample as passed into 2195 * the program. 2196 * 2197 * The *protocol* is the decoded protocol number (see 2198 * **enum rc_proto** for some predefined values). 2199 * 2200 * This helper is only available is the kernel was compiled with 2201 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2202 * "**y**". 2203 * Return 2204 * 0 2205 * 2206 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2207 * Description 2208 * Return the cgroup v2 id of the socket associated with the *skb*. 2209 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2210 * helper for cgroup v1 by providing a tag resp. identifier that 2211 * can be matched on or used for map lookups e.g. to implement 2212 * policy. The cgroup v2 id of a given path in the hierarchy is 2213 * exposed in user space through the f_handle API in order to get 2214 * to the same 64-bit id. 2215 * 2216 * This helper can be used on TC egress path, but not on ingress, 2217 * and is available only if the kernel was compiled with the 2218 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2219 * Return 2220 * The id is returned or 0 in case the id could not be retrieved. 2221 * 2222 * u64 bpf_get_current_cgroup_id(void) 2223 * Return 2224 * A 64-bit integer containing the current cgroup id based 2225 * on the cgroup within which the current task is running. 2226 * 2227 * void *bpf_get_local_storage(void *map, u64 flags) 2228 * Description 2229 * Get the pointer to the local storage area. 2230 * The type and the size of the local storage is defined 2231 * by the *map* argument. 2232 * The *flags* meaning is specific for each map type, 2233 * and has to be 0 for cgroup local storage. 2234 * 2235 * Depending on the BPF program type, a local storage area 2236 * can be shared between multiple instances of the BPF program, 2237 * running simultaneously. 2238 * 2239 * A user should care about the synchronization by himself. 2240 * For example, by using the **BPF_STX_XADD** instruction to alter 2241 * the shared data. 2242 * Return 2243 * A pointer to the local storage area. 2244 * 2245 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2246 * Description 2247 * Select a **SO_REUSEPORT** socket from a 2248 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2249 * It checks the selected socket is matching the incoming 2250 * request in the socket buffer. 2251 * Return 2252 * 0 on success, or a negative error in case of failure. 2253 * 2254 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2255 * Description 2256 * Return id of cgroup v2 that is ancestor of cgroup associated 2257 * with the *skb* at the *ancestor_level*. The root cgroup is at 2258 * *ancestor_level* zero and each step down the hierarchy 2259 * increments the level. If *ancestor_level* == level of cgroup 2260 * associated with *skb*, then return value will be same as that 2261 * of **bpf_skb_cgroup_id**\ (). 2262 * 2263 * The helper is useful to implement policies based on cgroups 2264 * that are upper in hierarchy than immediate cgroup associated 2265 * with *skb*. 2266 * 2267 * The format of returned id and helper limitations are same as in 2268 * **bpf_skb_cgroup_id**\ (). 2269 * Return 2270 * The id is returned or 0 in case the id could not be retrieved. 2271 * 2272 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2273 * Description 2274 * Look for TCP socket matching *tuple*, optionally in a child 2275 * network namespace *netns*. The return value must be checked, 2276 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2277 * 2278 * The *ctx* should point to the context of the program, such as 2279 * the skb or socket (depending on the hook in use). This is used 2280 * to determine the base network namespace for the lookup. 2281 * 2282 * *tuple_size* must be one of: 2283 * 2284 * **sizeof**\ (*tuple*\ **->ipv4**) 2285 * Look for an IPv4 socket. 2286 * **sizeof**\ (*tuple*\ **->ipv6**) 2287 * Look for an IPv6 socket. 2288 * 2289 * If the *netns* is a negative signed 32-bit integer, then the 2290 * socket lookup table in the netns associated with the *ctx* will 2291 * will be used. For the TC hooks, this is the netns of the device 2292 * in the skb. For socket hooks, this is the netns of the socket. 2293 * If *netns* is any other signed 32-bit value greater than or 2294 * equal to zero then it specifies the ID of the netns relative to 2295 * the netns associated with the *ctx*. *netns* values beyond the 2296 * range of 32-bit integers are reserved for future use. 2297 * 2298 * All values for *flags* are reserved for future usage, and must 2299 * be left at zero. 2300 * 2301 * This helper is available only if the kernel was compiled with 2302 * **CONFIG_NET** configuration option. 2303 * Return 2304 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2305 * For sockets with reuseport option, the **struct bpf_sock** 2306 * result is from *reuse*\ **->socks**\ [] using the hash of the 2307 * tuple. 2308 * 2309 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2310 * Description 2311 * Look for UDP socket matching *tuple*, optionally in a child 2312 * network namespace *netns*. The return value must be checked, 2313 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2314 * 2315 * The *ctx* should point to the context of the program, such as 2316 * the skb or socket (depending on the hook in use). This is used 2317 * to determine the base network namespace for the lookup. 2318 * 2319 * *tuple_size* must be one of: 2320 * 2321 * **sizeof**\ (*tuple*\ **->ipv4**) 2322 * Look for an IPv4 socket. 2323 * **sizeof**\ (*tuple*\ **->ipv6**) 2324 * Look for an IPv6 socket. 2325 * 2326 * If the *netns* is a negative signed 32-bit integer, then the 2327 * socket lookup table in the netns associated with the *ctx* will 2328 * will be used. For the TC hooks, this is the netns of the device 2329 * in the skb. For socket hooks, this is the netns of the socket. 2330 * If *netns* is any other signed 32-bit value greater than or 2331 * equal to zero then it specifies the ID of the netns relative to 2332 * the netns associated with the *ctx*. *netns* values beyond the 2333 * range of 32-bit integers are reserved for future use. 2334 * 2335 * All values for *flags* are reserved for future usage, and must 2336 * be left at zero. 2337 * 2338 * This helper is available only if the kernel was compiled with 2339 * **CONFIG_NET** configuration option. 2340 * Return 2341 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2342 * For sockets with reuseport option, the **struct bpf_sock** 2343 * result is from *reuse*\ **->socks**\ [] using the hash of the 2344 * tuple. 2345 * 2346 * int bpf_sk_release(struct bpf_sock *sock) 2347 * Description 2348 * Release the reference held by *sock*. *sock* must be a 2349 * non-**NULL** pointer that was returned from 2350 * **bpf_sk_lookup_xxx**\ (). 2351 * Return 2352 * 0 on success, or a negative error in case of failure. 2353 * 2354 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2355 * Description 2356 * Push an element *value* in *map*. *flags* is one of: 2357 * 2358 * **BPF_EXIST** 2359 * If the queue/stack is full, the oldest element is 2360 * removed to make room for this. 2361 * Return 2362 * 0 on success, or a negative error in case of failure. 2363 * 2364 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2365 * Description 2366 * Pop an element from *map*. 2367 * Return 2368 * 0 on success, or a negative error in case of failure. 2369 * 2370 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2371 * Description 2372 * Get an element from *map* without removing it. 2373 * Return 2374 * 0 on success, or a negative error in case of failure. 2375 * 2376 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 2377 * Description 2378 * For socket policies, insert *len* bytes into *msg* at offset 2379 * *start*. 2380 * 2381 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2382 * *msg* it may want to insert metadata or options into the *msg*. 2383 * This can later be read and used by any of the lower layer BPF 2384 * hooks. 2385 * 2386 * This helper may fail if under memory pressure (a malloc 2387 * fails) in these cases BPF programs will get an appropriate 2388 * error and BPF programs will need to handle them. 2389 * Return 2390 * 0 on success, or a negative error in case of failure. 2391 * 2392 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags) 2393 * Description 2394 * Will remove *pop* bytes from a *msg* starting at byte *start*. 2395 * This may result in **ENOMEM** errors under certain situations if 2396 * an allocation and copy are required due to a full ring buffer. 2397 * However, the helper will try to avoid doing the allocation 2398 * if possible. Other errors can occur if input parameters are 2399 * invalid either due to *start* byte not being valid part of *msg* 2400 * payload and/or *pop* value being to large. 2401 * Return 2402 * 0 on success, or a negative error in case of failure. 2403 * 2404 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2405 * Description 2406 * This helper is used in programs implementing IR decoding, to 2407 * report a successfully decoded pointer movement. 2408 * 2409 * The *ctx* should point to the lirc sample as passed into 2410 * the program. 2411 * 2412 * This helper is only available is the kernel was compiled with 2413 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2414 * "**y**". 2415 * Return 2416 * 0 2417 * 2418 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2419 * Description 2420 * Acquire a spinlock represented by the pointer *lock*, which is 2421 * stored as part of a value of a map. Taking the lock allows to 2422 * safely update the rest of the fields in that value. The 2423 * spinlock can (and must) later be released with a call to 2424 * **bpf_spin_unlock**\ (\ *lock*\ ). 2425 * 2426 * Spinlocks in BPF programs come with a number of restrictions 2427 * and constraints: 2428 * 2429 * * **bpf_spin_lock** objects are only allowed inside maps of 2430 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2431 * list could be extended in the future). 2432 * * BTF description of the map is mandatory. 2433 * * The BPF program can take ONE lock at a time, since taking two 2434 * or more could cause dead locks. 2435 * * Only one **struct bpf_spin_lock** is allowed per map element. 2436 * * When the lock is taken, calls (either BPF to BPF or helpers) 2437 * are not allowed. 2438 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2439 * allowed inside a spinlock-ed region. 2440 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2441 * the lock, on all execution paths, before it returns. 2442 * * The BPF program can access **struct bpf_spin_lock** only via 2443 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2444 * helpers. Loading or storing data into the **struct 2445 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2446 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2447 * of the map value must be a struct and have **struct 2448 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2449 * Nested lock inside another struct is not allowed. 2450 * * The **struct bpf_spin_lock** *lock* field in a map value must 2451 * be aligned on a multiple of 4 bytes in that value. 2452 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2453 * the **bpf_spin_lock** field to user space. 2454 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2455 * a BPF program, do not update the **bpf_spin_lock** field. 2456 * * **bpf_spin_lock** cannot be on the stack or inside a 2457 * networking packet (it can only be inside of a map values). 2458 * * **bpf_spin_lock** is available to root only. 2459 * * Tracing programs and socket filter programs cannot use 2460 * **bpf_spin_lock**\ () due to insufficient preemption checks 2461 * (but this may change in the future). 2462 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2463 * Return 2464 * 0 2465 * 2466 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2467 * Description 2468 * Release the *lock* previously locked by a call to 2469 * **bpf_spin_lock**\ (\ *lock*\ ). 2470 * Return 2471 * 0 2472 * 2473 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2474 * Description 2475 * This helper gets a **struct bpf_sock** pointer such 2476 * that all the fields in this **bpf_sock** can be accessed. 2477 * Return 2478 * A **struct bpf_sock** pointer on success, or **NULL** in 2479 * case of failure. 2480 * 2481 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2482 * Description 2483 * This helper gets a **struct bpf_tcp_sock** pointer from a 2484 * **struct bpf_sock** pointer. 2485 * Return 2486 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2487 * case of failure. 2488 * 2489 * int bpf_skb_ecn_set_ce(struct sk_buf *skb) 2490 * Description 2491 * Set ECN (Explicit Congestion Notification) field of IP header 2492 * to **CE** (Congestion Encountered) if current value is **ECT** 2493 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2494 * and IPv4. 2495 * Return 2496 * 1 if the **CE** flag is set (either by the current helper call 2497 * or because it was already present), 0 if it is not set. 2498 * 2499 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2500 * Description 2501 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2502 * **bpf_sk_release**\ () is unnecessary and not allowed. 2503 * Return 2504 * A **struct bpf_sock** pointer on success, or **NULL** in 2505 * case of failure. 2506 * 2507 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2508 * Description 2509 * Look for TCP socket matching *tuple*, optionally in a child 2510 * network namespace *netns*. The return value must be checked, 2511 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2512 * 2513 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2514 * that it also returns timewait or request sockets. Use 2515 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2516 * full structure. 2517 * 2518 * This helper is available only if the kernel was compiled with 2519 * **CONFIG_NET** configuration option. 2520 * Return 2521 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2522 * For sockets with reuseport option, the **struct bpf_sock** 2523 * result is from *reuse*\ **->socks**\ [] using the hash of the 2524 * tuple. 2525 * 2526 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2527 * Description 2528 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2529 * the listening socket in *sk*. 2530 * 2531 * *iph* points to the start of the IPv4 or IPv6 header, while 2532 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2533 * **sizeof**\ (**struct ip6hdr**). 2534 * 2535 * *th* points to the start of the TCP header, while *th_len* 2536 * contains **sizeof**\ (**struct tcphdr**). 2537 * 2538 * Return 2539 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2540 * error otherwise. 2541 * 2542 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2543 * Description 2544 * Get name of sysctl in /proc/sys/ and copy it into provided by 2545 * program buffer *buf* of size *buf_len*. 2546 * 2547 * The buffer is always NUL terminated, unless it's zero-sized. 2548 * 2549 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2550 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2551 * only (e.g. "tcp_mem"). 2552 * Return 2553 * Number of character copied (not including the trailing NUL). 2554 * 2555 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2556 * truncated name in this case). 2557 * 2558 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2559 * Description 2560 * Get current value of sysctl as it is presented in /proc/sys 2561 * (incl. newline, etc), and copy it as a string into provided 2562 * by program buffer *buf* of size *buf_len*. 2563 * 2564 * The whole value is copied, no matter what file position user 2565 * space issued e.g. sys_read at. 2566 * 2567 * The buffer is always NUL terminated, unless it's zero-sized. 2568 * Return 2569 * Number of character copied (not including the trailing NUL). 2570 * 2571 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2572 * truncated name in this case). 2573 * 2574 * **-EINVAL** if current value was unavailable, e.g. because 2575 * sysctl is uninitialized and read returns -EIO for it. 2576 * 2577 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2578 * Description 2579 * Get new value being written by user space to sysctl (before 2580 * the actual write happens) and copy it as a string into 2581 * provided by program buffer *buf* of size *buf_len*. 2582 * 2583 * User space may write new value at file position > 0. 2584 * 2585 * The buffer is always NUL terminated, unless it's zero-sized. 2586 * Return 2587 * Number of character copied (not including the trailing NUL). 2588 * 2589 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2590 * truncated name in this case). 2591 * 2592 * **-EINVAL** if sysctl is being read. 2593 * 2594 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2595 * Description 2596 * Override new value being written by user space to sysctl with 2597 * value provided by program in buffer *buf* of size *buf_len*. 2598 * 2599 * *buf* should contain a string in same form as provided by user 2600 * space on sysctl write. 2601 * 2602 * User space may write new value at file position > 0. To override 2603 * the whole sysctl value file position should be set to zero. 2604 * Return 2605 * 0 on success. 2606 * 2607 * **-E2BIG** if the *buf_len* is too big. 2608 * 2609 * **-EINVAL** if sysctl is being read. 2610 * 2611 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2612 * Description 2613 * Convert the initial part of the string from buffer *buf* of 2614 * size *buf_len* to a long integer according to the given base 2615 * and save the result in *res*. 2616 * 2617 * The string may begin with an arbitrary amount of white space 2618 * (as determined by **isspace**\ (3)) followed by a single 2619 * optional '**-**' sign. 2620 * 2621 * Five least significant bits of *flags* encode base, other bits 2622 * are currently unused. 2623 * 2624 * Base must be either 8, 10, 16 or 0 to detect it automatically 2625 * similar to user space **strtol**\ (3). 2626 * Return 2627 * Number of characters consumed on success. Must be positive but 2628 * no more than *buf_len*. 2629 * 2630 * **-EINVAL** if no valid digits were found or unsupported base 2631 * was provided. 2632 * 2633 * **-ERANGE** if resulting value was out of range. 2634 * 2635 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2636 * Description 2637 * Convert the initial part of the string from buffer *buf* of 2638 * size *buf_len* to an unsigned long integer according to the 2639 * given base and save the result in *res*. 2640 * 2641 * The string may begin with an arbitrary amount of white space 2642 * (as determined by **isspace**\ (3)). 2643 * 2644 * Five least significant bits of *flags* encode base, other bits 2645 * are currently unused. 2646 * 2647 * Base must be either 8, 10, 16 or 0 to detect it automatically 2648 * similar to user space **strtoul**\ (3). 2649 * Return 2650 * Number of characters consumed on success. Must be positive but 2651 * no more than *buf_len*. 2652 * 2653 * **-EINVAL** if no valid digits were found or unsupported base 2654 * was provided. 2655 * 2656 * **-ERANGE** if resulting value was out of range. 2657 * 2658 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2659 * Description 2660 * Get a bpf-local-storage from a *sk*. 2661 * 2662 * Logically, it could be thought of getting the value from 2663 * a *map* with *sk* as the **key**. From this 2664 * perspective, the usage is not much different from 2665 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2666 * helper enforces the key must be a full socket and the map must 2667 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2668 * 2669 * Underneath, the value is stored locally at *sk* instead of 2670 * the *map*. The *map* is used as the bpf-local-storage 2671 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2672 * searched against all bpf-local-storages residing at *sk*. 2673 * 2674 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2675 * used such that a new bpf-local-storage will be 2676 * created if one does not exist. *value* can be used 2677 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2678 * the initial value of a bpf-local-storage. If *value* is 2679 * **NULL**, the new bpf-local-storage will be zero initialized. 2680 * Return 2681 * A bpf-local-storage pointer is returned on success. 2682 * 2683 * **NULL** if not found or there was an error in adding 2684 * a new bpf-local-storage. 2685 * 2686 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2687 * Description 2688 * Delete a bpf-local-storage from a *sk*. 2689 * Return 2690 * 0 on success. 2691 * 2692 * **-ENOENT** if the bpf-local-storage cannot be found. 2693 * 2694 * int bpf_send_signal(u32 sig) 2695 * Description 2696 * Send signal *sig* to the current task. 2697 * Return 2698 * 0 on success or successfully queued. 2699 * 2700 * **-EBUSY** if work queue under nmi is full. 2701 * 2702 * **-EINVAL** if *sig* is invalid. 2703 * 2704 * **-EPERM** if no permission to send the *sig*. 2705 * 2706 * **-EAGAIN** if bpf program can try again. 2707 */ 2708 #define __BPF_FUNC_MAPPER(FN) \ 2709 FN(unspec), \ 2710 FN(map_lookup_elem), \ 2711 FN(map_update_elem), \ 2712 FN(map_delete_elem), \ 2713 FN(probe_read), \ 2714 FN(ktime_get_ns), \ 2715 FN(trace_printk), \ 2716 FN(get_prandom_u32), \ 2717 FN(get_smp_processor_id), \ 2718 FN(skb_store_bytes), \ 2719 FN(l3_csum_replace), \ 2720 FN(l4_csum_replace), \ 2721 FN(tail_call), \ 2722 FN(clone_redirect), \ 2723 FN(get_current_pid_tgid), \ 2724 FN(get_current_uid_gid), \ 2725 FN(get_current_comm), \ 2726 FN(get_cgroup_classid), \ 2727 FN(skb_vlan_push), \ 2728 FN(skb_vlan_pop), \ 2729 FN(skb_get_tunnel_key), \ 2730 FN(skb_set_tunnel_key), \ 2731 FN(perf_event_read), \ 2732 FN(redirect), \ 2733 FN(get_route_realm), \ 2734 FN(perf_event_output), \ 2735 FN(skb_load_bytes), \ 2736 FN(get_stackid), \ 2737 FN(csum_diff), \ 2738 FN(skb_get_tunnel_opt), \ 2739 FN(skb_set_tunnel_opt), \ 2740 FN(skb_change_proto), \ 2741 FN(skb_change_type), \ 2742 FN(skb_under_cgroup), \ 2743 FN(get_hash_recalc), \ 2744 FN(get_current_task), \ 2745 FN(probe_write_user), \ 2746 FN(current_task_under_cgroup), \ 2747 FN(skb_change_tail), \ 2748 FN(skb_pull_data), \ 2749 FN(csum_update), \ 2750 FN(set_hash_invalid), \ 2751 FN(get_numa_node_id), \ 2752 FN(skb_change_head), \ 2753 FN(xdp_adjust_head), \ 2754 FN(probe_read_str), \ 2755 FN(get_socket_cookie), \ 2756 FN(get_socket_uid), \ 2757 FN(set_hash), \ 2758 FN(setsockopt), \ 2759 FN(skb_adjust_room), \ 2760 FN(redirect_map), \ 2761 FN(sk_redirect_map), \ 2762 FN(sock_map_update), \ 2763 FN(xdp_adjust_meta), \ 2764 FN(perf_event_read_value), \ 2765 FN(perf_prog_read_value), \ 2766 FN(getsockopt), \ 2767 FN(override_return), \ 2768 FN(sock_ops_cb_flags_set), \ 2769 FN(msg_redirect_map), \ 2770 FN(msg_apply_bytes), \ 2771 FN(msg_cork_bytes), \ 2772 FN(msg_pull_data), \ 2773 FN(bind), \ 2774 FN(xdp_adjust_tail), \ 2775 FN(skb_get_xfrm_state), \ 2776 FN(get_stack), \ 2777 FN(skb_load_bytes_relative), \ 2778 FN(fib_lookup), \ 2779 FN(sock_hash_update), \ 2780 FN(msg_redirect_hash), \ 2781 FN(sk_redirect_hash), \ 2782 FN(lwt_push_encap), \ 2783 FN(lwt_seg6_store_bytes), \ 2784 FN(lwt_seg6_adjust_srh), \ 2785 FN(lwt_seg6_action), \ 2786 FN(rc_repeat), \ 2787 FN(rc_keydown), \ 2788 FN(skb_cgroup_id), \ 2789 FN(get_current_cgroup_id), \ 2790 FN(get_local_storage), \ 2791 FN(sk_select_reuseport), \ 2792 FN(skb_ancestor_cgroup_id), \ 2793 FN(sk_lookup_tcp), \ 2794 FN(sk_lookup_udp), \ 2795 FN(sk_release), \ 2796 FN(map_push_elem), \ 2797 FN(map_pop_elem), \ 2798 FN(map_peek_elem), \ 2799 FN(msg_push_data), \ 2800 FN(msg_pop_data), \ 2801 FN(rc_pointer_rel), \ 2802 FN(spin_lock), \ 2803 FN(spin_unlock), \ 2804 FN(sk_fullsock), \ 2805 FN(tcp_sock), \ 2806 FN(skb_ecn_set_ce), \ 2807 FN(get_listener_sock), \ 2808 FN(skc_lookup_tcp), \ 2809 FN(tcp_check_syncookie), \ 2810 FN(sysctl_get_name), \ 2811 FN(sysctl_get_current_value), \ 2812 FN(sysctl_get_new_value), \ 2813 FN(sysctl_set_new_value), \ 2814 FN(strtol), \ 2815 FN(strtoul), \ 2816 FN(sk_storage_get), \ 2817 FN(sk_storage_delete), \ 2818 FN(send_signal), 2819 2820 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2821 * function eBPF program intends to call 2822 */ 2823 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2824 enum bpf_func_id { 2825 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2826 __BPF_FUNC_MAX_ID, 2827 }; 2828 #undef __BPF_ENUM_FN 2829 2830 /* All flags used by eBPF helper functions, placed here. */ 2831 2832 /* BPF_FUNC_skb_store_bytes flags. */ 2833 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2834 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2835 2836 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2837 * First 4 bits are for passing the header field size. 2838 */ 2839 #define BPF_F_HDR_FIELD_MASK 0xfULL 2840 2841 /* BPF_FUNC_l4_csum_replace flags. */ 2842 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2843 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2844 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2845 2846 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2847 #define BPF_F_INGRESS (1ULL << 0) 2848 2849 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2850 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2851 2852 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2853 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2854 #define BPF_F_USER_STACK (1ULL << 8) 2855 /* flags used by BPF_FUNC_get_stackid only. */ 2856 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2857 #define BPF_F_REUSE_STACKID (1ULL << 10) 2858 /* flags used by BPF_FUNC_get_stack only. */ 2859 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2860 2861 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2862 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2863 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2864 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2865 2866 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2867 * BPF_FUNC_perf_event_read_value flags. 2868 */ 2869 #define BPF_F_INDEX_MASK 0xffffffffULL 2870 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2871 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2872 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2873 2874 /* Current network namespace */ 2875 #define BPF_F_CURRENT_NETNS (-1L) 2876 2877 /* BPF_FUNC_skb_adjust_room flags. */ 2878 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 2879 2880 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 2881 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 2882 2883 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 2884 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 2885 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 2886 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 2887 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 2888 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 2889 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 2890 2891 /* BPF_FUNC_sysctl_get_name flags. */ 2892 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0) 2893 2894 /* BPF_FUNC_sk_storage_get flags */ 2895 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0) 2896 2897 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 2898 enum bpf_adj_room_mode { 2899 BPF_ADJ_ROOM_NET, 2900 BPF_ADJ_ROOM_MAC, 2901 }; 2902 2903 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 2904 enum bpf_hdr_start_off { 2905 BPF_HDR_START_MAC, 2906 BPF_HDR_START_NET, 2907 }; 2908 2909 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2910 enum bpf_lwt_encap_mode { 2911 BPF_LWT_ENCAP_SEG6, 2912 BPF_LWT_ENCAP_SEG6_INLINE, 2913 BPF_LWT_ENCAP_IP, 2914 }; 2915 2916 #define __bpf_md_ptr(type, name) \ 2917 union { \ 2918 type name; \ 2919 __u64 :64; \ 2920 } __attribute__((aligned(8))) 2921 2922 /* user accessible mirror of in-kernel sk_buff. 2923 * new fields can only be added to the end of this structure 2924 */ 2925 struct __sk_buff { 2926 __u32 len; 2927 __u32 pkt_type; 2928 __u32 mark; 2929 __u32 queue_mapping; 2930 __u32 protocol; 2931 __u32 vlan_present; 2932 __u32 vlan_tci; 2933 __u32 vlan_proto; 2934 __u32 priority; 2935 __u32 ingress_ifindex; 2936 __u32 ifindex; 2937 __u32 tc_index; 2938 __u32 cb[5]; 2939 __u32 hash; 2940 __u32 tc_classid; 2941 __u32 data; 2942 __u32 data_end; 2943 __u32 napi_id; 2944 2945 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 2946 __u32 family; 2947 __u32 remote_ip4; /* Stored in network byte order */ 2948 __u32 local_ip4; /* Stored in network byte order */ 2949 __u32 remote_ip6[4]; /* Stored in network byte order */ 2950 __u32 local_ip6[4]; /* Stored in network byte order */ 2951 __u32 remote_port; /* Stored in network byte order */ 2952 __u32 local_port; /* stored in host byte order */ 2953 /* ... here. */ 2954 2955 __u32 data_meta; 2956 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 2957 __u64 tstamp; 2958 __u32 wire_len; 2959 __u32 gso_segs; 2960 __bpf_md_ptr(struct bpf_sock *, sk); 2961 }; 2962 2963 struct bpf_tunnel_key { 2964 __u32 tunnel_id; 2965 union { 2966 __u32 remote_ipv4; 2967 __u32 remote_ipv6[4]; 2968 }; 2969 __u8 tunnel_tos; 2970 __u8 tunnel_ttl; 2971 __u16 tunnel_ext; /* Padding, future use. */ 2972 __u32 tunnel_label; 2973 }; 2974 2975 /* user accessible mirror of in-kernel xfrm_state. 2976 * new fields can only be added to the end of this structure 2977 */ 2978 struct bpf_xfrm_state { 2979 __u32 reqid; 2980 __u32 spi; /* Stored in network byte order */ 2981 __u16 family; 2982 __u16 ext; /* Padding, future use. */ 2983 union { 2984 __u32 remote_ipv4; /* Stored in network byte order */ 2985 __u32 remote_ipv6[4]; /* Stored in network byte order */ 2986 }; 2987 }; 2988 2989 /* Generic BPF return codes which all BPF program types may support. 2990 * The values are binary compatible with their TC_ACT_* counter-part to 2991 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 2992 * programs. 2993 * 2994 * XDP is handled seprately, see XDP_*. 2995 */ 2996 enum bpf_ret_code { 2997 BPF_OK = 0, 2998 /* 1 reserved */ 2999 BPF_DROP = 2, 3000 /* 3-6 reserved */ 3001 BPF_REDIRECT = 7, 3002 /* >127 are reserved for prog type specific return codes. 3003 * 3004 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3005 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3006 * changed and should be routed based on its new L3 header. 3007 * (This is an L3 redirect, as opposed to L2 redirect 3008 * represented by BPF_REDIRECT above). 3009 */ 3010 BPF_LWT_REROUTE = 128, 3011 }; 3012 3013 struct bpf_sock { 3014 __u32 bound_dev_if; 3015 __u32 family; 3016 __u32 type; 3017 __u32 protocol; 3018 __u32 mark; 3019 __u32 priority; 3020 /* IP address also allows 1 and 2 bytes access */ 3021 __u32 src_ip4; 3022 __u32 src_ip6[4]; 3023 __u32 src_port; /* host byte order */ 3024 __u32 dst_port; /* network byte order */ 3025 __u32 dst_ip4; 3026 __u32 dst_ip6[4]; 3027 __u32 state; 3028 }; 3029 3030 struct bpf_tcp_sock { 3031 __u32 snd_cwnd; /* Sending congestion window */ 3032 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3033 __u32 rtt_min; 3034 __u32 snd_ssthresh; /* Slow start size threshold */ 3035 __u32 rcv_nxt; /* What we want to receive next */ 3036 __u32 snd_nxt; /* Next sequence we send */ 3037 __u32 snd_una; /* First byte we want an ack for */ 3038 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3039 __u32 ecn_flags; /* ECN status bits. */ 3040 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3041 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3042 __u32 packets_out; /* Packets which are "in flight" */ 3043 __u32 retrans_out; /* Retransmitted packets out */ 3044 __u32 total_retrans; /* Total retransmits for entire connection */ 3045 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3046 * total number of segments in. 3047 */ 3048 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3049 * total number of data segments in. 3050 */ 3051 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3052 * The total number of segments sent. 3053 */ 3054 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3055 * total number of data segments sent. 3056 */ 3057 __u32 lost_out; /* Lost packets */ 3058 __u32 sacked_out; /* SACK'd packets */ 3059 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3060 * sum(delta(rcv_nxt)), or how many bytes 3061 * were acked. 3062 */ 3063 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3064 * sum(delta(snd_una)), or how many bytes 3065 * were acked. 3066 */ 3067 }; 3068 3069 struct bpf_sock_tuple { 3070 union { 3071 struct { 3072 __be32 saddr; 3073 __be32 daddr; 3074 __be16 sport; 3075 __be16 dport; 3076 } ipv4; 3077 struct { 3078 __be32 saddr[4]; 3079 __be32 daddr[4]; 3080 __be16 sport; 3081 __be16 dport; 3082 } ipv6; 3083 }; 3084 }; 3085 3086 #define XDP_PACKET_HEADROOM 256 3087 3088 /* User return codes for XDP prog type. 3089 * A valid XDP program must return one of these defined values. All other 3090 * return codes are reserved for future use. Unknown return codes will 3091 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3092 */ 3093 enum xdp_action { 3094 XDP_ABORTED = 0, 3095 XDP_DROP, 3096 XDP_PASS, 3097 XDP_TX, 3098 XDP_REDIRECT, 3099 }; 3100 3101 /* user accessible metadata for XDP packet hook 3102 * new fields must be added to the end of this structure 3103 */ 3104 struct xdp_md { 3105 __u32 data; 3106 __u32 data_end; 3107 __u32 data_meta; 3108 /* Below access go through struct xdp_rxq_info */ 3109 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3110 __u32 rx_queue_index; /* rxq->queue_index */ 3111 }; 3112 3113 enum sk_action { 3114 SK_DROP = 0, 3115 SK_PASS, 3116 }; 3117 3118 /* user accessible metadata for SK_MSG packet hook, new fields must 3119 * be added to the end of this structure 3120 */ 3121 struct sk_msg_md { 3122 __bpf_md_ptr(void *, data); 3123 __bpf_md_ptr(void *, data_end); 3124 3125 __u32 family; 3126 __u32 remote_ip4; /* Stored in network byte order */ 3127 __u32 local_ip4; /* Stored in network byte order */ 3128 __u32 remote_ip6[4]; /* Stored in network byte order */ 3129 __u32 local_ip6[4]; /* Stored in network byte order */ 3130 __u32 remote_port; /* Stored in network byte order */ 3131 __u32 local_port; /* stored in host byte order */ 3132 __u32 size; /* Total size of sk_msg */ 3133 }; 3134 3135 struct sk_reuseport_md { 3136 /* 3137 * Start of directly accessible data. It begins from 3138 * the tcp/udp header. 3139 */ 3140 __bpf_md_ptr(void *, data); 3141 /* End of directly accessible data */ 3142 __bpf_md_ptr(void *, data_end); 3143 /* 3144 * Total length of packet (starting from the tcp/udp header). 3145 * Note that the directly accessible bytes (data_end - data) 3146 * could be less than this "len". Those bytes could be 3147 * indirectly read by a helper "bpf_skb_load_bytes()". 3148 */ 3149 __u32 len; 3150 /* 3151 * Eth protocol in the mac header (network byte order). e.g. 3152 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3153 */ 3154 __u32 eth_protocol; 3155 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3156 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3157 __u32 hash; /* A hash of the packet 4 tuples */ 3158 }; 3159 3160 #define BPF_TAG_SIZE 8 3161 3162 struct bpf_prog_info { 3163 __u32 type; 3164 __u32 id; 3165 __u8 tag[BPF_TAG_SIZE]; 3166 __u32 jited_prog_len; 3167 __u32 xlated_prog_len; 3168 __aligned_u64 jited_prog_insns; 3169 __aligned_u64 xlated_prog_insns; 3170 __u64 load_time; /* ns since boottime */ 3171 __u32 created_by_uid; 3172 __u32 nr_map_ids; 3173 __aligned_u64 map_ids; 3174 char name[BPF_OBJ_NAME_LEN]; 3175 __u32 ifindex; 3176 __u32 gpl_compatible:1; 3177 __u64 netns_dev; 3178 __u64 netns_ino; 3179 __u32 nr_jited_ksyms; 3180 __u32 nr_jited_func_lens; 3181 __aligned_u64 jited_ksyms; 3182 __aligned_u64 jited_func_lens; 3183 __u32 btf_id; 3184 __u32 func_info_rec_size; 3185 __aligned_u64 func_info; 3186 __u32 nr_func_info; 3187 __u32 nr_line_info; 3188 __aligned_u64 line_info; 3189 __aligned_u64 jited_line_info; 3190 __u32 nr_jited_line_info; 3191 __u32 line_info_rec_size; 3192 __u32 jited_line_info_rec_size; 3193 __u32 nr_prog_tags; 3194 __aligned_u64 prog_tags; 3195 __u64 run_time_ns; 3196 __u64 run_cnt; 3197 } __attribute__((aligned(8))); 3198 3199 struct bpf_map_info { 3200 __u32 type; 3201 __u32 id; 3202 __u32 key_size; 3203 __u32 value_size; 3204 __u32 max_entries; 3205 __u32 map_flags; 3206 char name[BPF_OBJ_NAME_LEN]; 3207 __u32 ifindex; 3208 __u32 :32; 3209 __u64 netns_dev; 3210 __u64 netns_ino; 3211 __u32 btf_id; 3212 __u32 btf_key_type_id; 3213 __u32 btf_value_type_id; 3214 } __attribute__((aligned(8))); 3215 3216 struct bpf_btf_info { 3217 __aligned_u64 btf; 3218 __u32 btf_size; 3219 __u32 id; 3220 } __attribute__((aligned(8))); 3221 3222 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3223 * by user and intended to be used by socket (e.g. to bind to, depends on 3224 * attach attach type). 3225 */ 3226 struct bpf_sock_addr { 3227 __u32 user_family; /* Allows 4-byte read, but no write. */ 3228 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3229 * Stored in network byte order. 3230 */ 3231 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 3232 * Stored in network byte order. 3233 */ 3234 __u32 user_port; /* Allows 4-byte read and write. 3235 * Stored in network byte order 3236 */ 3237 __u32 family; /* Allows 4-byte read, but no write */ 3238 __u32 type; /* Allows 4-byte read, but no write */ 3239 __u32 protocol; /* Allows 4-byte read, but no write */ 3240 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write. 3241 * Stored in network byte order. 3242 */ 3243 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 3244 * Stored in network byte order. 3245 */ 3246 }; 3247 3248 /* User bpf_sock_ops struct to access socket values and specify request ops 3249 * and their replies. 3250 * Some of this fields are in network (bigendian) byte order and may need 3251 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3252 * New fields can only be added at the end of this structure 3253 */ 3254 struct bpf_sock_ops { 3255 __u32 op; 3256 union { 3257 __u32 args[4]; /* Optionally passed to bpf program */ 3258 __u32 reply; /* Returned by bpf program */ 3259 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3260 }; 3261 __u32 family; 3262 __u32 remote_ip4; /* Stored in network byte order */ 3263 __u32 local_ip4; /* Stored in network byte order */ 3264 __u32 remote_ip6[4]; /* Stored in network byte order */ 3265 __u32 local_ip6[4]; /* Stored in network byte order */ 3266 __u32 remote_port; /* Stored in network byte order */ 3267 __u32 local_port; /* stored in host byte order */ 3268 __u32 is_fullsock; /* Some TCP fields are only valid if 3269 * there is a full socket. If not, the 3270 * fields read as zero. 3271 */ 3272 __u32 snd_cwnd; 3273 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3274 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3275 __u32 state; 3276 __u32 rtt_min; 3277 __u32 snd_ssthresh; 3278 __u32 rcv_nxt; 3279 __u32 snd_nxt; 3280 __u32 snd_una; 3281 __u32 mss_cache; 3282 __u32 ecn_flags; 3283 __u32 rate_delivered; 3284 __u32 rate_interval_us; 3285 __u32 packets_out; 3286 __u32 retrans_out; 3287 __u32 total_retrans; 3288 __u32 segs_in; 3289 __u32 data_segs_in; 3290 __u32 segs_out; 3291 __u32 data_segs_out; 3292 __u32 lost_out; 3293 __u32 sacked_out; 3294 __u32 sk_txhash; 3295 __u64 bytes_received; 3296 __u64 bytes_acked; 3297 }; 3298 3299 /* Definitions for bpf_sock_ops_cb_flags */ 3300 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3301 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3302 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3303 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 3304 * supported cb flags 3305 */ 3306 3307 /* List of known BPF sock_ops operators. 3308 * New entries can only be added at the end 3309 */ 3310 enum { 3311 BPF_SOCK_OPS_VOID, 3312 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3313 * -1 if default value should be used 3314 */ 3315 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3316 * window (in packets) or -1 if default 3317 * value should be used 3318 */ 3319 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3320 * active connection is initialized 3321 */ 3322 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3323 * active connection is 3324 * established 3325 */ 3326 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3327 * passive connection is 3328 * established 3329 */ 3330 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3331 * needs ECN 3332 */ 3333 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3334 * based on the path and may be 3335 * dependent on the congestion control 3336 * algorithm. In general it indicates 3337 * a congestion threshold. RTTs above 3338 * this indicate congestion 3339 */ 3340 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3341 * Arg1: value of icsk_retransmits 3342 * Arg2: value of icsk_rto 3343 * Arg3: whether RTO has expired 3344 */ 3345 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3346 * Arg1: sequence number of 1st byte 3347 * Arg2: # segments 3348 * Arg3: return value of 3349 * tcp_transmit_skb (0 => success) 3350 */ 3351 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3352 * Arg1: old_state 3353 * Arg2: new_state 3354 */ 3355 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3356 * socket transition to LISTEN state. 3357 */ 3358 }; 3359 3360 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3361 * changes between the TCP and BPF versions. Ideally this should never happen. 3362 * If it does, we need to add code to convert them before calling 3363 * the BPF sock_ops function. 3364 */ 3365 enum { 3366 BPF_TCP_ESTABLISHED = 1, 3367 BPF_TCP_SYN_SENT, 3368 BPF_TCP_SYN_RECV, 3369 BPF_TCP_FIN_WAIT1, 3370 BPF_TCP_FIN_WAIT2, 3371 BPF_TCP_TIME_WAIT, 3372 BPF_TCP_CLOSE, 3373 BPF_TCP_CLOSE_WAIT, 3374 BPF_TCP_LAST_ACK, 3375 BPF_TCP_LISTEN, 3376 BPF_TCP_CLOSING, /* Now a valid state */ 3377 BPF_TCP_NEW_SYN_RECV, 3378 3379 BPF_TCP_MAX_STATES /* Leave at the end! */ 3380 }; 3381 3382 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3383 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3384 3385 struct bpf_perf_event_value { 3386 __u64 counter; 3387 __u64 enabled; 3388 __u64 running; 3389 }; 3390 3391 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3392 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3393 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3394 3395 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3396 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3397 3398 struct bpf_cgroup_dev_ctx { 3399 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3400 __u32 access_type; 3401 __u32 major; 3402 __u32 minor; 3403 }; 3404 3405 struct bpf_raw_tracepoint_args { 3406 __u64 args[0]; 3407 }; 3408 3409 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3410 * OUTPUT: Do lookup from egress perspective; default is ingress 3411 */ 3412 #define BPF_FIB_LOOKUP_DIRECT BIT(0) 3413 #define BPF_FIB_LOOKUP_OUTPUT BIT(1) 3414 3415 enum { 3416 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3417 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3418 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3419 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3420 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3421 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3422 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3423 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3424 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3425 }; 3426 3427 struct bpf_fib_lookup { 3428 /* input: network family for lookup (AF_INET, AF_INET6) 3429 * output: network family of egress nexthop 3430 */ 3431 __u8 family; 3432 3433 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3434 __u8 l4_protocol; 3435 __be16 sport; 3436 __be16 dport; 3437 3438 /* total length of packet from network header - used for MTU check */ 3439 __u16 tot_len; 3440 3441 /* input: L3 device index for lookup 3442 * output: device index from FIB lookup 3443 */ 3444 __u32 ifindex; 3445 3446 union { 3447 /* inputs to lookup */ 3448 __u8 tos; /* AF_INET */ 3449 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3450 3451 /* output: metric of fib result (IPv4/IPv6 only) */ 3452 __u32 rt_metric; 3453 }; 3454 3455 union { 3456 __be32 ipv4_src; 3457 __u32 ipv6_src[4]; /* in6_addr; network order */ 3458 }; 3459 3460 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3461 * network header. output: bpf_fib_lookup sets to gateway address 3462 * if FIB lookup returns gateway route 3463 */ 3464 union { 3465 __be32 ipv4_dst; 3466 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3467 }; 3468 3469 /* output */ 3470 __be16 h_vlan_proto; 3471 __be16 h_vlan_TCI; 3472 __u8 smac[6]; /* ETH_ALEN */ 3473 __u8 dmac[6]; /* ETH_ALEN */ 3474 }; 3475 3476 enum bpf_task_fd_type { 3477 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3478 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3479 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3480 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3481 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3482 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3483 }; 3484 3485 struct bpf_flow_keys { 3486 __u16 nhoff; 3487 __u16 thoff; 3488 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3489 __u8 is_frag; 3490 __u8 is_first_frag; 3491 __u8 is_encap; 3492 __u8 ip_proto; 3493 __be16 n_proto; 3494 __be16 sport; 3495 __be16 dport; 3496 union { 3497 struct { 3498 __be32 ipv4_src; 3499 __be32 ipv4_dst; 3500 }; 3501 struct { 3502 __u32 ipv6_src[4]; /* in6_addr; network order */ 3503 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3504 }; 3505 }; 3506 }; 3507 3508 struct bpf_func_info { 3509 __u32 insn_off; 3510 __u32 type_id; 3511 }; 3512 3513 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3514 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3515 3516 struct bpf_line_info { 3517 __u32 insn_off; 3518 __u32 file_name_off; 3519 __u32 line_off; 3520 __u32 line_col; 3521 }; 3522 3523 struct bpf_spin_lock { 3524 __u32 val; 3525 }; 3526 3527 struct bpf_sysctl { 3528 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3529 * Allows 1,2,4-byte read, but no write. 3530 */ 3531 __u32 file_pos; /* Sysctl file position to read from, write to. 3532 * Allows 1,2,4-byte read an 4-byte write. 3533 */ 3534 }; 3535 3536 #endif /* _UAPI__LINUX_BPF_H__ */ 3537