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 BPF_PROG_TYPE_CGROUP_SOCKOPT, 174 }; 175 176 enum bpf_attach_type { 177 BPF_CGROUP_INET_INGRESS, 178 BPF_CGROUP_INET_EGRESS, 179 BPF_CGROUP_INET_SOCK_CREATE, 180 BPF_CGROUP_SOCK_OPS, 181 BPF_SK_SKB_STREAM_PARSER, 182 BPF_SK_SKB_STREAM_VERDICT, 183 BPF_CGROUP_DEVICE, 184 BPF_SK_MSG_VERDICT, 185 BPF_CGROUP_INET4_BIND, 186 BPF_CGROUP_INET6_BIND, 187 BPF_CGROUP_INET4_CONNECT, 188 BPF_CGROUP_INET6_CONNECT, 189 BPF_CGROUP_INET4_POST_BIND, 190 BPF_CGROUP_INET6_POST_BIND, 191 BPF_CGROUP_UDP4_SENDMSG, 192 BPF_CGROUP_UDP6_SENDMSG, 193 BPF_LIRC_MODE2, 194 BPF_FLOW_DISSECTOR, 195 BPF_CGROUP_SYSCTL, 196 BPF_CGROUP_UDP4_RECVMSG, 197 BPF_CGROUP_UDP6_RECVMSG, 198 BPF_CGROUP_GETSOCKOPT, 199 BPF_CGROUP_SETSOCKOPT, 200 __MAX_BPF_ATTACH_TYPE 201 }; 202 203 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 204 205 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 206 * 207 * NONE(default): No further bpf programs allowed in the subtree. 208 * 209 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 210 * the program in this cgroup yields to sub-cgroup program. 211 * 212 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 213 * that cgroup program gets run in addition to the program in this cgroup. 214 * 215 * Only one program is allowed to be attached to a cgroup with 216 * NONE or BPF_F_ALLOW_OVERRIDE flag. 217 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 218 * release old program and attach the new one. Attach flags has to match. 219 * 220 * Multiple programs are allowed to be attached to a cgroup with 221 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 222 * (those that were attached first, run first) 223 * The programs of sub-cgroup are executed first, then programs of 224 * this cgroup and then programs of parent cgroup. 225 * When children program makes decision (like picking TCP CA or sock bind) 226 * parent program has a chance to override it. 227 * 228 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 229 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 230 * Ex1: 231 * cgrp1 (MULTI progs A, B) -> 232 * cgrp2 (OVERRIDE prog C) -> 233 * cgrp3 (MULTI prog D) -> 234 * cgrp4 (OVERRIDE prog E) -> 235 * cgrp5 (NONE prog F) 236 * the event in cgrp5 triggers execution of F,D,A,B in that order. 237 * if prog F is detached, the execution is E,D,A,B 238 * if prog F and D are detached, the execution is E,A,B 239 * if prog F, E and D are detached, the execution is C,A,B 240 * 241 * All eligible programs are executed regardless of return code from 242 * earlier programs. 243 */ 244 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 245 #define BPF_F_ALLOW_MULTI (1U << 1) 246 247 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 248 * verifier will perform strict alignment checking as if the kernel 249 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 250 * and NET_IP_ALIGN defined to 2. 251 */ 252 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 253 254 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 255 * verifier will allow any alignment whatsoever. On platforms 256 * with strict alignment requirements for loads ands stores (such 257 * as sparc and mips) the verifier validates that all loads and 258 * stores provably follow this requirement. This flag turns that 259 * checking and enforcement off. 260 * 261 * It is mostly used for testing when we want to validate the 262 * context and memory access aspects of the verifier, but because 263 * of an unaligned access the alignment check would trigger before 264 * the one we are interested in. 265 */ 266 #define BPF_F_ANY_ALIGNMENT (1U << 1) 267 268 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 269 * Verifier does sub-register def/use analysis and identifies instructions whose 270 * def only matters for low 32-bit, high 32-bit is never referenced later 271 * through implicit zero extension. Therefore verifier notifies JIT back-ends 272 * that it is safe to ignore clearing high 32-bit for these instructions. This 273 * saves some back-ends a lot of code-gen. However such optimization is not 274 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 275 * hence hasn't used verifier's analysis result. But, we really want to have a 276 * way to be able to verify the correctness of the described optimization on 277 * x86_64 on which testsuites are frequently exercised. 278 * 279 * So, this flag is introduced. Once it is set, verifier will randomize high 280 * 32-bit for those instructions who has been identified as safe to ignore them. 281 * Then, if verifier is not doing correct analysis, such randomization will 282 * regress tests to expose bugs. 283 */ 284 #define BPF_F_TEST_RND_HI32 (1U << 2) 285 286 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 287 * two extensions: 288 * 289 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 290 * insn[0].imm: map fd map fd 291 * insn[1].imm: 0 offset into value 292 * insn[0].off: 0 0 293 * insn[1].off: 0 0 294 * ldimm64 rewrite: address of map address of map[0]+offset 295 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 296 */ 297 #define BPF_PSEUDO_MAP_FD 1 298 #define BPF_PSEUDO_MAP_VALUE 2 299 300 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 301 * offset to another bpf function 302 */ 303 #define BPF_PSEUDO_CALL 1 304 305 /* flags for BPF_MAP_UPDATE_ELEM command */ 306 #define BPF_ANY 0 /* create new element or update existing */ 307 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 308 #define BPF_EXIST 2 /* update existing element */ 309 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 310 311 /* flags for BPF_MAP_CREATE command */ 312 #define BPF_F_NO_PREALLOC (1U << 0) 313 /* Instead of having one common LRU list in the 314 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 315 * which can scale and perform better. 316 * Note, the LRU nodes (including free nodes) cannot be moved 317 * across different LRU lists. 318 */ 319 #define BPF_F_NO_COMMON_LRU (1U << 1) 320 /* Specify numa node during map creation */ 321 #define BPF_F_NUMA_NODE (1U << 2) 322 323 #define BPF_OBJ_NAME_LEN 16U 324 325 /* Flags for accessing BPF object from syscall side. */ 326 #define BPF_F_RDONLY (1U << 3) 327 #define BPF_F_WRONLY (1U << 4) 328 329 /* Flag for stack_map, store build_id+offset instead of pointer */ 330 #define BPF_F_STACK_BUILD_ID (1U << 5) 331 332 /* Zero-initialize hash function seed. This should only be used for testing. */ 333 #define BPF_F_ZERO_SEED (1U << 6) 334 335 /* Flags for accessing BPF object from program side. */ 336 #define BPF_F_RDONLY_PROG (1U << 7) 337 #define BPF_F_WRONLY_PROG (1U << 8) 338 339 /* flags for BPF_PROG_QUERY */ 340 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 341 342 enum bpf_stack_build_id_status { 343 /* user space need an empty entry to identify end of a trace */ 344 BPF_STACK_BUILD_ID_EMPTY = 0, 345 /* with valid build_id and offset */ 346 BPF_STACK_BUILD_ID_VALID = 1, 347 /* couldn't get build_id, fallback to ip */ 348 BPF_STACK_BUILD_ID_IP = 2, 349 }; 350 351 #define BPF_BUILD_ID_SIZE 20 352 struct bpf_stack_build_id { 353 __s32 status; 354 unsigned char build_id[BPF_BUILD_ID_SIZE]; 355 union { 356 __u64 offset; 357 __u64 ip; 358 }; 359 }; 360 361 union bpf_attr { 362 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 363 __u32 map_type; /* one of enum bpf_map_type */ 364 __u32 key_size; /* size of key in bytes */ 365 __u32 value_size; /* size of value in bytes */ 366 __u32 max_entries; /* max number of entries in a map */ 367 __u32 map_flags; /* BPF_MAP_CREATE related 368 * flags defined above. 369 */ 370 __u32 inner_map_fd; /* fd pointing to the inner map */ 371 __u32 numa_node; /* numa node (effective only if 372 * BPF_F_NUMA_NODE is set). 373 */ 374 char map_name[BPF_OBJ_NAME_LEN]; 375 __u32 map_ifindex; /* ifindex of netdev to create on */ 376 __u32 btf_fd; /* fd pointing to a BTF type data */ 377 __u32 btf_key_type_id; /* BTF type_id of the key */ 378 __u32 btf_value_type_id; /* BTF type_id of the value */ 379 }; 380 381 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 382 __u32 map_fd; 383 __aligned_u64 key; 384 union { 385 __aligned_u64 value; 386 __aligned_u64 next_key; 387 }; 388 __u64 flags; 389 }; 390 391 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 392 __u32 prog_type; /* one of enum bpf_prog_type */ 393 __u32 insn_cnt; 394 __aligned_u64 insns; 395 __aligned_u64 license; 396 __u32 log_level; /* verbosity level of verifier */ 397 __u32 log_size; /* size of user buffer */ 398 __aligned_u64 log_buf; /* user supplied buffer */ 399 __u32 kern_version; /* not used */ 400 __u32 prog_flags; 401 char prog_name[BPF_OBJ_NAME_LEN]; 402 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 403 /* For some prog types expected attach type must be known at 404 * load time to verify attach type specific parts of prog 405 * (context accesses, allowed helpers, etc). 406 */ 407 __u32 expected_attach_type; 408 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 409 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 410 __aligned_u64 func_info; /* func info */ 411 __u32 func_info_cnt; /* number of bpf_func_info records */ 412 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 413 __aligned_u64 line_info; /* line info */ 414 __u32 line_info_cnt; /* number of bpf_line_info records */ 415 }; 416 417 struct { /* anonymous struct used by BPF_OBJ_* commands */ 418 __aligned_u64 pathname; 419 __u32 bpf_fd; 420 __u32 file_flags; 421 }; 422 423 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 424 __u32 target_fd; /* container object to attach to */ 425 __u32 attach_bpf_fd; /* eBPF program to attach */ 426 __u32 attach_type; 427 __u32 attach_flags; 428 }; 429 430 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 431 __u32 prog_fd; 432 __u32 retval; 433 __u32 data_size_in; /* input: len of data_in */ 434 __u32 data_size_out; /* input/output: len of data_out 435 * returns ENOSPC if data_out 436 * is too small. 437 */ 438 __aligned_u64 data_in; 439 __aligned_u64 data_out; 440 __u32 repeat; 441 __u32 duration; 442 __u32 ctx_size_in; /* input: len of ctx_in */ 443 __u32 ctx_size_out; /* input/output: len of ctx_out 444 * returns ENOSPC if ctx_out 445 * is too small. 446 */ 447 __aligned_u64 ctx_in; 448 __aligned_u64 ctx_out; 449 } test; 450 451 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 452 union { 453 __u32 start_id; 454 __u32 prog_id; 455 __u32 map_id; 456 __u32 btf_id; 457 }; 458 __u32 next_id; 459 __u32 open_flags; 460 }; 461 462 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 463 __u32 bpf_fd; 464 __u32 info_len; 465 __aligned_u64 info; 466 } info; 467 468 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 469 __u32 target_fd; /* container object to query */ 470 __u32 attach_type; 471 __u32 query_flags; 472 __u32 attach_flags; 473 __aligned_u64 prog_ids; 474 __u32 prog_cnt; 475 } query; 476 477 struct { 478 __u64 name; 479 __u32 prog_fd; 480 } raw_tracepoint; 481 482 struct { /* anonymous struct for BPF_BTF_LOAD */ 483 __aligned_u64 btf; 484 __aligned_u64 btf_log_buf; 485 __u32 btf_size; 486 __u32 btf_log_size; 487 __u32 btf_log_level; 488 }; 489 490 struct { 491 __u32 pid; /* input: pid */ 492 __u32 fd; /* input: fd */ 493 __u32 flags; /* input: flags */ 494 __u32 buf_len; /* input/output: buf len */ 495 __aligned_u64 buf; /* input/output: 496 * tp_name for tracepoint 497 * symbol for kprobe 498 * filename for uprobe 499 */ 500 __u32 prog_id; /* output: prod_id */ 501 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 502 __u64 probe_offset; /* output: probe_offset */ 503 __u64 probe_addr; /* output: probe_addr */ 504 } task_fd_query; 505 } __attribute__((aligned(8))); 506 507 /* The description below is an attempt at providing documentation to eBPF 508 * developers about the multiple available eBPF helper functions. It can be 509 * parsed and used to produce a manual page. The workflow is the following, 510 * and requires the rst2man utility: 511 * 512 * $ ./scripts/bpf_helpers_doc.py \ 513 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 514 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 515 * $ man /tmp/bpf-helpers.7 516 * 517 * Note that in order to produce this external documentation, some RST 518 * formatting is used in the descriptions to get "bold" and "italics" in 519 * manual pages. Also note that the few trailing white spaces are 520 * intentional, removing them would break paragraphs for rst2man. 521 * 522 * Start of BPF helper function descriptions: 523 * 524 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 525 * Description 526 * Perform a lookup in *map* for an entry associated to *key*. 527 * Return 528 * Map value associated to *key*, or **NULL** if no entry was 529 * found. 530 * 531 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 532 * Description 533 * Add or update the value of the entry associated to *key* in 534 * *map* with *value*. *flags* is one of: 535 * 536 * **BPF_NOEXIST** 537 * The entry for *key* must not exist in the map. 538 * **BPF_EXIST** 539 * The entry for *key* must already exist in the map. 540 * **BPF_ANY** 541 * No condition on the existence of the entry for *key*. 542 * 543 * Flag value **BPF_NOEXIST** cannot be used for maps of types 544 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 545 * elements always exist), the helper would return an error. 546 * Return 547 * 0 on success, or a negative error in case of failure. 548 * 549 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 550 * Description 551 * Delete entry with *key* from *map*. 552 * Return 553 * 0 on success, or a negative error in case of failure. 554 * 555 * int bpf_probe_read(void *dst, u32 size, const void *src) 556 * Description 557 * For tracing programs, safely attempt to read *size* bytes from 558 * address *src* and store the data in *dst*. 559 * Return 560 * 0 on success, or a negative error in case of failure. 561 * 562 * u64 bpf_ktime_get_ns(void) 563 * Description 564 * Return the time elapsed since system boot, in nanoseconds. 565 * Return 566 * Current *ktime*. 567 * 568 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 569 * Description 570 * This helper is a "printk()-like" facility for debugging. It 571 * prints a message defined by format *fmt* (of size *fmt_size*) 572 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 573 * available. It can take up to three additional **u64** 574 * arguments (as an eBPF helpers, the total number of arguments is 575 * limited to five). 576 * 577 * Each time the helper is called, it appends a line to the trace. 578 * The format of the trace is customizable, and the exact output 579 * one will get depends on the options set in 580 * *\/sys/kernel/debug/tracing/trace_options* (see also the 581 * *README* file under the same directory). However, it usually 582 * defaults to something like: 583 * 584 * :: 585 * 586 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 587 * 588 * In the above: 589 * 590 * * ``telnet`` is the name of the current task. 591 * * ``470`` is the PID of the current task. 592 * * ``001`` is the CPU number on which the task is 593 * running. 594 * * In ``.N..``, each character refers to a set of 595 * options (whether irqs are enabled, scheduling 596 * options, whether hard/softirqs are running, level of 597 * preempt_disabled respectively). **N** means that 598 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 599 * are set. 600 * * ``419421.045894`` is a timestamp. 601 * * ``0x00000001`` is a fake value used by BPF for the 602 * instruction pointer register. 603 * * ``<formatted msg>`` is the message formatted with 604 * *fmt*. 605 * 606 * The conversion specifiers supported by *fmt* are similar, but 607 * more limited than for printk(). They are **%d**, **%i**, 608 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 609 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 610 * of field, padding with zeroes, etc.) is available, and the 611 * helper will return **-EINVAL** (but print nothing) if it 612 * encounters an unknown specifier. 613 * 614 * Also, note that **bpf_trace_printk**\ () is slow, and should 615 * only be used for debugging purposes. For this reason, a notice 616 * bloc (spanning several lines) is printed to kernel logs and 617 * states that the helper should not be used "for production use" 618 * the first time this helper is used (or more precisely, when 619 * **trace_printk**\ () buffers are allocated). For passing values 620 * to user space, perf events should be preferred. 621 * Return 622 * The number of bytes written to the buffer, or a negative error 623 * in case of failure. 624 * 625 * u32 bpf_get_prandom_u32(void) 626 * Description 627 * Get a pseudo-random number. 628 * 629 * From a security point of view, this helper uses its own 630 * pseudo-random internal state, and cannot be used to infer the 631 * seed of other random functions in the kernel. However, it is 632 * essential to note that the generator used by the helper is not 633 * cryptographically secure. 634 * Return 635 * A random 32-bit unsigned value. 636 * 637 * u32 bpf_get_smp_processor_id(void) 638 * Description 639 * Get the SMP (symmetric multiprocessing) processor id. Note that 640 * all programs run with preemption disabled, which means that the 641 * SMP processor id is stable during all the execution of the 642 * program. 643 * Return 644 * The SMP id of the processor running the program. 645 * 646 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 647 * Description 648 * Store *len* bytes from address *from* into the packet 649 * associated to *skb*, at *offset*. *flags* are a combination of 650 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 651 * checksum for the packet after storing the bytes) and 652 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 653 * **->swhash** and *skb*\ **->l4hash** to 0). 654 * 655 * A call to this helper is susceptible to change the underlying 656 * packet buffer. Therefore, at load time, all checks on pointers 657 * previously done by the verifier are invalidated and must be 658 * performed again, if the helper is used in combination with 659 * direct packet access. 660 * Return 661 * 0 on success, or a negative error in case of failure. 662 * 663 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 664 * Description 665 * Recompute the layer 3 (e.g. IP) checksum for the packet 666 * associated to *skb*. Computation is incremental, so the helper 667 * must know the former value of the header field that was 668 * modified (*from*), the new value of this field (*to*), and the 669 * number of bytes (2 or 4) for this field, stored in *size*. 670 * Alternatively, it is possible to store the difference between 671 * the previous and the new values of the header field in *to*, by 672 * setting *from* and *size* to 0. For both methods, *offset* 673 * indicates the location of the IP checksum within the packet. 674 * 675 * This helper works in combination with **bpf_csum_diff**\ (), 676 * which does not update the checksum in-place, but offers more 677 * flexibility and can handle sizes larger than 2 or 4 for the 678 * checksum to update. 679 * 680 * A call to this helper is susceptible to change the underlying 681 * packet buffer. Therefore, at load time, all checks on pointers 682 * previously done by the verifier are invalidated and must be 683 * performed again, if the helper is used in combination with 684 * direct packet access. 685 * Return 686 * 0 on success, or a negative error in case of failure. 687 * 688 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 689 * Description 690 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 691 * packet associated to *skb*. Computation is incremental, so the 692 * helper must know the former value of the header field that was 693 * modified (*from*), the new value of this field (*to*), and the 694 * number of bytes (2 or 4) for this field, stored on the lowest 695 * four bits of *flags*. Alternatively, it is possible to store 696 * the difference between the previous and the new values of the 697 * header field in *to*, by setting *from* and the four lowest 698 * bits of *flags* to 0. For both methods, *offset* indicates the 699 * location of the IP checksum within the packet. In addition to 700 * the size of the field, *flags* can be added (bitwise OR) actual 701 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 702 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 703 * for updates resulting in a null checksum the value is set to 704 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 705 * the checksum is to be computed against a pseudo-header. 706 * 707 * This helper works in combination with **bpf_csum_diff**\ (), 708 * which does not update the checksum in-place, but offers more 709 * flexibility and can handle sizes larger than 2 or 4 for the 710 * checksum to update. 711 * 712 * A call to this helper is susceptible to change the underlying 713 * packet buffer. Therefore, at load time, all checks on pointers 714 * previously done by the verifier are invalidated and must be 715 * performed again, if the helper is used in combination with 716 * direct packet access. 717 * Return 718 * 0 on success, or a negative error in case of failure. 719 * 720 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 721 * Description 722 * This special helper is used to trigger a "tail call", or in 723 * other words, to jump into another eBPF program. The same stack 724 * frame is used (but values on stack and in registers for the 725 * caller are not accessible to the callee). This mechanism allows 726 * for program chaining, either for raising the maximum number of 727 * available eBPF instructions, or to execute given programs in 728 * conditional blocks. For security reasons, there is an upper 729 * limit to the number of successive tail calls that can be 730 * performed. 731 * 732 * Upon call of this helper, the program attempts to jump into a 733 * program referenced at index *index* in *prog_array_map*, a 734 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 735 * *ctx*, a pointer to the context. 736 * 737 * If the call succeeds, the kernel immediately runs the first 738 * instruction of the new program. This is not a function call, 739 * and it never returns to the previous program. If the call 740 * fails, then the helper has no effect, and the caller continues 741 * to run its subsequent instructions. A call can fail if the 742 * destination program for the jump does not exist (i.e. *index* 743 * is superior to the number of entries in *prog_array_map*), or 744 * if the maximum number of tail calls has been reached for this 745 * chain of programs. This limit is defined in the kernel by the 746 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 747 * which is currently set to 32. 748 * Return 749 * 0 on success, or a negative error in case of failure. 750 * 751 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 752 * Description 753 * Clone and redirect the packet associated to *skb* to another 754 * net device of index *ifindex*. Both ingress and egress 755 * interfaces can be used for redirection. The **BPF_F_INGRESS** 756 * value in *flags* is used to make the distinction (ingress path 757 * is selected if the flag is present, egress path otherwise). 758 * This is the only flag supported for now. 759 * 760 * In comparison with **bpf_redirect**\ () helper, 761 * **bpf_clone_redirect**\ () has the associated cost of 762 * duplicating the packet buffer, but this can be executed out of 763 * the eBPF program. Conversely, **bpf_redirect**\ () is more 764 * efficient, but it is handled through an action code where the 765 * redirection happens only after the eBPF program has returned. 766 * 767 * A call to this helper is susceptible to change the underlying 768 * packet buffer. Therefore, at load time, all checks on pointers 769 * previously done by the verifier are invalidated and must be 770 * performed again, if the helper is used in combination with 771 * direct packet access. 772 * Return 773 * 0 on success, or a negative error in case of failure. 774 * 775 * u64 bpf_get_current_pid_tgid(void) 776 * Return 777 * A 64-bit integer containing the current tgid and pid, and 778 * created as such: 779 * *current_task*\ **->tgid << 32 \|** 780 * *current_task*\ **->pid**. 781 * 782 * u64 bpf_get_current_uid_gid(void) 783 * Return 784 * A 64-bit integer containing the current GID and UID, and 785 * created as such: *current_gid* **<< 32 \|** *current_uid*. 786 * 787 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 788 * Description 789 * Copy the **comm** attribute of the current task into *buf* of 790 * *size_of_buf*. The **comm** attribute contains the name of 791 * the executable (excluding the path) for the current task. The 792 * *size_of_buf* must be strictly positive. On success, the 793 * helper makes sure that the *buf* is NUL-terminated. On failure, 794 * it is filled with zeroes. 795 * Return 796 * 0 on success, or a negative error in case of failure. 797 * 798 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 799 * Description 800 * Retrieve the classid for the current task, i.e. for the net_cls 801 * cgroup to which *skb* belongs. 802 * 803 * This helper can be used on TC egress path, but not on ingress. 804 * 805 * The net_cls cgroup provides an interface to tag network packets 806 * based on a user-provided identifier for all traffic coming from 807 * the tasks belonging to the related cgroup. See also the related 808 * kernel documentation, available from the Linux sources in file 809 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 810 * 811 * The Linux kernel has two versions for cgroups: there are 812 * cgroups v1 and cgroups v2. Both are available to users, who can 813 * use a mixture of them, but note that the net_cls cgroup is for 814 * cgroup v1 only. This makes it incompatible with BPF programs 815 * run on cgroups, which is a cgroup-v2-only feature (a socket can 816 * only hold data for one version of cgroups at a time). 817 * 818 * This helper is only available is the kernel was compiled with 819 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 820 * "**y**" or to "**m**". 821 * Return 822 * The classid, or 0 for the default unconfigured classid. 823 * 824 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 825 * Description 826 * Push a *vlan_tci* (VLAN tag control information) of protocol 827 * *vlan_proto* to the packet associated to *skb*, then update 828 * the checksum. Note that if *vlan_proto* is different from 829 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 830 * be **ETH_P_8021Q**. 831 * 832 * A call to this helper is susceptible to change the underlying 833 * packet buffer. Therefore, at load time, all checks on pointers 834 * previously done by the verifier are invalidated and must be 835 * performed again, if the helper is used in combination with 836 * direct packet access. 837 * Return 838 * 0 on success, or a negative error in case of failure. 839 * 840 * int bpf_skb_vlan_pop(struct sk_buff *skb) 841 * Description 842 * Pop a VLAN header from the packet associated to *skb*. 843 * 844 * A call to this helper is susceptible to change the underlying 845 * packet buffer. Therefore, at load time, all checks on pointers 846 * previously done by the verifier are invalidated and must be 847 * performed again, if the helper is used in combination with 848 * direct packet access. 849 * Return 850 * 0 on success, or a negative error in case of failure. 851 * 852 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 853 * Description 854 * Get tunnel metadata. This helper takes a pointer *key* to an 855 * empty **struct bpf_tunnel_key** of **size**, that will be 856 * filled with tunnel metadata for the packet associated to *skb*. 857 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 858 * indicates that the tunnel is based on IPv6 protocol instead of 859 * IPv4. 860 * 861 * The **struct bpf_tunnel_key** is an object that generalizes the 862 * principal parameters used by various tunneling protocols into a 863 * single struct. This way, it can be used to easily make a 864 * decision based on the contents of the encapsulation header, 865 * "summarized" in this struct. In particular, it holds the IP 866 * address of the remote end (IPv4 or IPv6, depending on the case) 867 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 868 * this struct exposes the *key*\ **->tunnel_id**, which is 869 * generally mapped to a VNI (Virtual Network Identifier), making 870 * it programmable together with the **bpf_skb_set_tunnel_key**\ 871 * () helper. 872 * 873 * Let's imagine that the following code is part of a program 874 * attached to the TC ingress interface, on one end of a GRE 875 * tunnel, and is supposed to filter out all messages coming from 876 * remote ends with IPv4 address other than 10.0.0.1: 877 * 878 * :: 879 * 880 * int ret; 881 * struct bpf_tunnel_key key = {}; 882 * 883 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 884 * if (ret < 0) 885 * return TC_ACT_SHOT; // drop packet 886 * 887 * if (key.remote_ipv4 != 0x0a000001) 888 * return TC_ACT_SHOT; // drop packet 889 * 890 * return TC_ACT_OK; // accept packet 891 * 892 * This interface can also be used with all encapsulation devices 893 * that can operate in "collect metadata" mode: instead of having 894 * one network device per specific configuration, the "collect 895 * metadata" mode only requires a single device where the 896 * configuration can be extracted from this helper. 897 * 898 * This can be used together with various tunnels such as VXLan, 899 * Geneve, GRE or IP in IP (IPIP). 900 * Return 901 * 0 on success, or a negative error in case of failure. 902 * 903 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 904 * Description 905 * Populate tunnel metadata for packet associated to *skb.* The 906 * tunnel metadata is set to the contents of *key*, of *size*. The 907 * *flags* can be set to a combination of the following values: 908 * 909 * **BPF_F_TUNINFO_IPV6** 910 * Indicate that the tunnel is based on IPv6 protocol 911 * instead of IPv4. 912 * **BPF_F_ZERO_CSUM_TX** 913 * For IPv4 packets, add a flag to tunnel metadata 914 * indicating that checksum computation should be skipped 915 * and checksum set to zeroes. 916 * **BPF_F_DONT_FRAGMENT** 917 * Add a flag to tunnel metadata indicating that the 918 * packet should not be fragmented. 919 * **BPF_F_SEQ_NUMBER** 920 * Add a flag to tunnel metadata indicating that a 921 * sequence number should be added to tunnel header before 922 * sending the packet. This flag was added for GRE 923 * encapsulation, but might be used with other protocols 924 * as well in the future. 925 * 926 * Here is a typical usage on the transmit path: 927 * 928 * :: 929 * 930 * struct bpf_tunnel_key key; 931 * populate key ... 932 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 933 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 934 * 935 * See also the description of the **bpf_skb_get_tunnel_key**\ () 936 * helper for additional information. 937 * Return 938 * 0 on success, or a negative error in case of failure. 939 * 940 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 941 * Description 942 * Read the value of a perf event counter. This helper relies on a 943 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 944 * the perf event counter is selected when *map* is updated with 945 * perf event file descriptors. The *map* is an array whose size 946 * is the number of available CPUs, and each cell contains a value 947 * relative to one CPU. The value to retrieve is indicated by 948 * *flags*, that contains the index of the CPU to look up, masked 949 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 950 * **BPF_F_CURRENT_CPU** to indicate that the value for the 951 * current CPU should be retrieved. 952 * 953 * Note that before Linux 4.13, only hardware perf event can be 954 * retrieved. 955 * 956 * Also, be aware that the newer helper 957 * **bpf_perf_event_read_value**\ () is recommended over 958 * **bpf_perf_event_read**\ () in general. The latter has some ABI 959 * quirks where error and counter value are used as a return code 960 * (which is wrong to do since ranges may overlap). This issue is 961 * fixed with **bpf_perf_event_read_value**\ (), which at the same 962 * time provides more features over the **bpf_perf_event_read**\ 963 * () interface. Please refer to the description of 964 * **bpf_perf_event_read_value**\ () for details. 965 * Return 966 * The value of the perf event counter read from the map, or a 967 * negative error code in case of failure. 968 * 969 * int bpf_redirect(u32 ifindex, u64 flags) 970 * Description 971 * Redirect the packet to another net device of index *ifindex*. 972 * This helper is somewhat similar to **bpf_clone_redirect**\ 973 * (), except that the packet is not cloned, which provides 974 * increased performance. 975 * 976 * Except for XDP, both ingress and egress interfaces can be used 977 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 978 * to make the distinction (ingress path is selected if the flag 979 * is present, egress path otherwise). Currently, XDP only 980 * supports redirection to the egress interface, and accepts no 981 * flag at all. 982 * 983 * The same effect can be attained with the more generic 984 * **bpf_redirect_map**\ (), which requires specific maps to be 985 * used but offers better performance. 986 * Return 987 * For XDP, the helper returns **XDP_REDIRECT** on success or 988 * **XDP_ABORTED** on error. For other program types, the values 989 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 990 * error. 991 * 992 * u32 bpf_get_route_realm(struct sk_buff *skb) 993 * Description 994 * Retrieve the realm or the route, that is to say the 995 * **tclassid** field of the destination for the *skb*. The 996 * indentifier retrieved is a user-provided tag, similar to the 997 * one used with the net_cls cgroup (see description for 998 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 999 * held by a route (a destination entry), not by a task. 1000 * 1001 * Retrieving this identifier works with the clsact TC egress hook 1002 * (see also **tc-bpf(8)**), or alternatively on conventional 1003 * classful egress qdiscs, but not on TC ingress path. In case of 1004 * clsact TC egress hook, this has the advantage that, internally, 1005 * the destination entry has not been dropped yet in the transmit 1006 * path. Therefore, the destination entry does not need to be 1007 * artificially held via **netif_keep_dst**\ () for a classful 1008 * qdisc until the *skb* is freed. 1009 * 1010 * This helper is available only if the kernel was compiled with 1011 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1012 * Return 1013 * The realm of the route for the packet associated to *skb*, or 0 1014 * if none was found. 1015 * 1016 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1017 * Description 1018 * Write raw *data* blob into a special BPF perf event held by 1019 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1020 * event must have the following attributes: **PERF_SAMPLE_RAW** 1021 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1022 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1023 * 1024 * The *flags* are used to indicate the index in *map* for which 1025 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1026 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1027 * to indicate that the index of the current CPU core should be 1028 * used. 1029 * 1030 * The value to write, of *size*, is passed through eBPF stack and 1031 * pointed by *data*. 1032 * 1033 * The context of the program *ctx* needs also be passed to the 1034 * helper. 1035 * 1036 * On user space, a program willing to read the values needs to 1037 * call **perf_event_open**\ () on the perf event (either for 1038 * one or for all CPUs) and to store the file descriptor into the 1039 * *map*. This must be done before the eBPF program can send data 1040 * into it. An example is available in file 1041 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1042 * tree (the eBPF program counterpart is in 1043 * *samples/bpf/trace_output_kern.c*). 1044 * 1045 * **bpf_perf_event_output**\ () achieves better performance 1046 * than **bpf_trace_printk**\ () for sharing data with user 1047 * space, and is much better suitable for streaming data from eBPF 1048 * programs. 1049 * 1050 * Note that this helper is not restricted to tracing use cases 1051 * and can be used with programs attached to TC or XDP as well, 1052 * where it allows for passing data to user space listeners. Data 1053 * can be: 1054 * 1055 * * Only custom structs, 1056 * * Only the packet payload, or 1057 * * A combination of both. 1058 * Return 1059 * 0 on success, or a negative error in case of failure. 1060 * 1061 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 1062 * Description 1063 * This helper was provided as an easy way to load data from a 1064 * packet. It can be used to load *len* bytes from *offset* from 1065 * the packet associated to *skb*, into the buffer pointed by 1066 * *to*. 1067 * 1068 * Since Linux 4.7, usage of this helper has mostly been replaced 1069 * by "direct packet access", enabling packet data to be 1070 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1071 * pointing respectively to the first byte of packet data and to 1072 * the byte after the last byte of packet data. However, it 1073 * remains useful if one wishes to read large quantities of data 1074 * at once from a packet into the eBPF stack. 1075 * Return 1076 * 0 on success, or a negative error in case of failure. 1077 * 1078 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 1079 * Description 1080 * Walk a user or a kernel stack and return its id. To achieve 1081 * this, the helper needs *ctx*, which is a pointer to the context 1082 * on which the tracing program is executed, and a pointer to a 1083 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1084 * 1085 * The last argument, *flags*, holds the number of stack frames to 1086 * skip (from 0 to 255), masked with 1087 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1088 * a combination of the following flags: 1089 * 1090 * **BPF_F_USER_STACK** 1091 * Collect a user space stack instead of a kernel stack. 1092 * **BPF_F_FAST_STACK_CMP** 1093 * Compare stacks by hash only. 1094 * **BPF_F_REUSE_STACKID** 1095 * If two different stacks hash into the same *stackid*, 1096 * discard the old one. 1097 * 1098 * The stack id retrieved is a 32 bit long integer handle which 1099 * can be further combined with other data (including other stack 1100 * ids) and used as a key into maps. This can be useful for 1101 * generating a variety of graphs (such as flame graphs or off-cpu 1102 * graphs). 1103 * 1104 * For walking a stack, this helper is an improvement over 1105 * **bpf_probe_read**\ (), which can be used with unrolled loops 1106 * but is not efficient and consumes a lot of eBPF instructions. 1107 * Instead, **bpf_get_stackid**\ () can collect up to 1108 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1109 * this limit can be controlled with the **sysctl** program, and 1110 * that it should be manually increased in order to profile long 1111 * user stacks (such as stacks for Java programs). To do so, use: 1112 * 1113 * :: 1114 * 1115 * # sysctl kernel.perf_event_max_stack=<new value> 1116 * Return 1117 * The positive or null stack id on success, or a negative error 1118 * in case of failure. 1119 * 1120 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1121 * Description 1122 * Compute a checksum difference, from the raw buffer pointed by 1123 * *from*, of length *from_size* (that must be a multiple of 4), 1124 * towards the raw buffer pointed by *to*, of size *to_size* 1125 * (same remark). An optional *seed* can be added to the value 1126 * (this can be cascaded, the seed may come from a previous call 1127 * to the helper). 1128 * 1129 * This is flexible enough to be used in several ways: 1130 * 1131 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1132 * checksum, it can be used when pushing new data. 1133 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1134 * checksum, it can be used when removing data from a packet. 1135 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1136 * can be used to compute a diff. Note that *from_size* and 1137 * *to_size* do not need to be equal. 1138 * 1139 * This helper can be used in combination with 1140 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1141 * which one can feed in the difference computed with 1142 * **bpf_csum_diff**\ (). 1143 * Return 1144 * The checksum result, or a negative error code in case of 1145 * failure. 1146 * 1147 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1148 * Description 1149 * Retrieve tunnel options metadata for the packet associated to 1150 * *skb*, and store the raw tunnel option data to the buffer *opt* 1151 * of *size*. 1152 * 1153 * This helper can be used with encapsulation devices that can 1154 * operate in "collect metadata" mode (please refer to the related 1155 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1156 * more details). A particular example where this can be used is 1157 * in combination with the Geneve encapsulation protocol, where it 1158 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1159 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1160 * the eBPF program. This allows for full customization of these 1161 * headers. 1162 * Return 1163 * The size of the option data retrieved. 1164 * 1165 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1166 * Description 1167 * Set tunnel options metadata for the packet associated to *skb* 1168 * to the option data contained in the raw buffer *opt* of *size*. 1169 * 1170 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1171 * helper for additional information. 1172 * Return 1173 * 0 on success, or a negative error in case of failure. 1174 * 1175 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1176 * Description 1177 * Change the protocol of the *skb* to *proto*. Currently 1178 * supported are transition from IPv4 to IPv6, and from IPv6 to 1179 * IPv4. The helper takes care of the groundwork for the 1180 * transition, including resizing the socket buffer. The eBPF 1181 * program is expected to fill the new headers, if any, via 1182 * **skb_store_bytes**\ () and to recompute the checksums with 1183 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1184 * (). The main case for this helper is to perform NAT64 1185 * operations out of an eBPF program. 1186 * 1187 * Internally, the GSO type is marked as dodgy so that headers are 1188 * checked and segments are recalculated by the GSO/GRO engine. 1189 * The size for GSO target is adapted as well. 1190 * 1191 * All values for *flags* are reserved for future usage, and must 1192 * be left at zero. 1193 * 1194 * A call to this helper is susceptible to change the underlying 1195 * packet buffer. Therefore, at load time, all checks on pointers 1196 * previously done by the verifier are invalidated and must be 1197 * performed again, if the helper is used in combination with 1198 * direct packet access. 1199 * Return 1200 * 0 on success, or a negative error in case of failure. 1201 * 1202 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1203 * Description 1204 * Change the packet type for the packet associated to *skb*. This 1205 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1206 * the eBPF program does not have a write access to *skb*\ 1207 * **->pkt_type** beside this helper. Using a helper here allows 1208 * for graceful handling of errors. 1209 * 1210 * The major use case is to change incoming *skb*s to 1211 * **PACKET_HOST** in a programmatic way instead of having to 1212 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1213 * example. 1214 * 1215 * Note that *type* only allows certain values. At this time, they 1216 * are: 1217 * 1218 * **PACKET_HOST** 1219 * Packet is for us. 1220 * **PACKET_BROADCAST** 1221 * Send packet to all. 1222 * **PACKET_MULTICAST** 1223 * Send packet to group. 1224 * **PACKET_OTHERHOST** 1225 * Send packet to someone else. 1226 * Return 1227 * 0 on success, or a negative error in case of failure. 1228 * 1229 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1230 * Description 1231 * Check whether *skb* is a descendant of the cgroup2 held by 1232 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1233 * Return 1234 * The return value depends on the result of the test, and can be: 1235 * 1236 * * 0, if the *skb* failed the cgroup2 descendant test. 1237 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1238 * * A negative error code, if an error occurred. 1239 * 1240 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1241 * Description 1242 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1243 * not set, in particular if the hash was cleared due to mangling, 1244 * recompute this hash. Later accesses to the hash can be done 1245 * directly with *skb*\ **->hash**. 1246 * 1247 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1248 * prototype with **bpf_skb_change_proto**\ (), or calling 1249 * **bpf_skb_store_bytes**\ () with the 1250 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1251 * the hash and to trigger a new computation for the next call to 1252 * **bpf_get_hash_recalc**\ (). 1253 * Return 1254 * The 32-bit hash. 1255 * 1256 * u64 bpf_get_current_task(void) 1257 * Return 1258 * A pointer to the current task struct. 1259 * 1260 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1261 * Description 1262 * Attempt in a safe way to write *len* bytes from the buffer 1263 * *src* to *dst* in memory. It only works for threads that are in 1264 * user context, and *dst* must be a valid user space address. 1265 * 1266 * This helper should not be used to implement any kind of 1267 * security mechanism because of TOC-TOU attacks, but rather to 1268 * debug, divert, and manipulate execution of semi-cooperative 1269 * processes. 1270 * 1271 * Keep in mind that this feature is meant for experiments, and it 1272 * has a risk of crashing the system and running programs. 1273 * Therefore, when an eBPF program using this helper is attached, 1274 * a warning including PID and process name is printed to kernel 1275 * logs. 1276 * Return 1277 * 0 on success, or a negative error in case of failure. 1278 * 1279 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1280 * Description 1281 * Check whether the probe is being run is the context of a given 1282 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1283 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1284 * Return 1285 * The return value depends on the result of the test, and can be: 1286 * 1287 * * 0, if the *skb* task belongs to the cgroup2. 1288 * * 1, if the *skb* task does not belong to the cgroup2. 1289 * * A negative error code, if an error occurred. 1290 * 1291 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1292 * Description 1293 * Resize (trim or grow) the packet associated to *skb* to the 1294 * new *len*. The *flags* are reserved for future usage, and must 1295 * be left at zero. 1296 * 1297 * The basic idea is that the helper performs the needed work to 1298 * change the size of the packet, then the eBPF program rewrites 1299 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1300 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1301 * and others. This helper is a slow path utility intended for 1302 * replies with control messages. And because it is targeted for 1303 * slow path, the helper itself can afford to be slow: it 1304 * implicitly linearizes, unclones and drops offloads from the 1305 * *skb*. 1306 * 1307 * A call to this helper is susceptible to change the underlying 1308 * packet buffer. Therefore, at load time, all checks on pointers 1309 * previously done by the verifier are invalidated and must be 1310 * performed again, if the helper is used in combination with 1311 * direct packet access. 1312 * Return 1313 * 0 on success, or a negative error in case of failure. 1314 * 1315 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1316 * Description 1317 * Pull in non-linear data in case the *skb* is non-linear and not 1318 * all of *len* are part of the linear section. Make *len* bytes 1319 * from *skb* readable and writable. If a zero value is passed for 1320 * *len*, then the whole length of the *skb* is pulled. 1321 * 1322 * This helper is only needed for reading and writing with direct 1323 * packet access. 1324 * 1325 * For direct packet access, testing that offsets to access 1326 * are within packet boundaries (test on *skb*\ **->data_end**) is 1327 * susceptible to fail if offsets are invalid, or if the requested 1328 * data is in non-linear parts of the *skb*. On failure the 1329 * program can just bail out, or in the case of a non-linear 1330 * buffer, use a helper to make the data available. The 1331 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1332 * the data. Another one consists in using **bpf_skb_pull_data** 1333 * to pull in once the non-linear parts, then retesting and 1334 * eventually access the data. 1335 * 1336 * At the same time, this also makes sure the *skb* is uncloned, 1337 * which is a necessary condition for direct write. As this needs 1338 * to be an invariant for the write part only, the verifier 1339 * detects writes and adds a prologue that is calling 1340 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1341 * the very beginning in case it is indeed cloned. 1342 * 1343 * A call to this helper is susceptible to change the underlying 1344 * packet buffer. Therefore, at load time, all checks on pointers 1345 * previously done by the verifier are invalidated and must be 1346 * performed again, if the helper is used in combination with 1347 * direct packet access. 1348 * Return 1349 * 0 on success, or a negative error in case of failure. 1350 * 1351 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1352 * Description 1353 * Add the checksum *csum* into *skb*\ **->csum** in case the 1354 * driver has supplied a checksum for the entire packet into that 1355 * field. Return an error otherwise. This helper is intended to be 1356 * used in combination with **bpf_csum_diff**\ (), in particular 1357 * when the checksum needs to be updated after data has been 1358 * written into the packet through direct packet access. 1359 * Return 1360 * The checksum on success, or a negative error code in case of 1361 * failure. 1362 * 1363 * void bpf_set_hash_invalid(struct sk_buff *skb) 1364 * Description 1365 * Invalidate the current *skb*\ **->hash**. It can be used after 1366 * mangling on headers through direct packet access, in order to 1367 * indicate that the hash is outdated and to trigger a 1368 * recalculation the next time the kernel tries to access this 1369 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1370 * 1371 * int bpf_get_numa_node_id(void) 1372 * Description 1373 * Return the id of the current NUMA node. The primary use case 1374 * for this helper is the selection of sockets for the local NUMA 1375 * node, when the program is attached to sockets using the 1376 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1377 * but the helper is also available to other eBPF program types, 1378 * similarly to **bpf_get_smp_processor_id**\ (). 1379 * Return 1380 * The id of current NUMA node. 1381 * 1382 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1383 * Description 1384 * Grows headroom of packet associated to *skb* and adjusts the 1385 * offset of the MAC header accordingly, adding *len* bytes of 1386 * space. It automatically extends and reallocates memory as 1387 * required. 1388 * 1389 * This helper can be used on a layer 3 *skb* to push a MAC header 1390 * for redirection into a layer 2 device. 1391 * 1392 * All values for *flags* are reserved for future usage, and must 1393 * be left at zero. 1394 * 1395 * A call to this helper is susceptible to change the underlying 1396 * packet buffer. Therefore, at load time, all checks on pointers 1397 * previously done by the verifier are invalidated and must be 1398 * performed again, if the helper is used in combination with 1399 * direct packet access. 1400 * Return 1401 * 0 on success, or a negative error in case of failure. 1402 * 1403 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1404 * Description 1405 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1406 * it is possible to use a negative value for *delta*. This helper 1407 * can be used to prepare the packet for pushing or popping 1408 * headers. 1409 * 1410 * A call to this helper is susceptible to change the underlying 1411 * packet buffer. Therefore, at load time, all checks on pointers 1412 * previously done by the verifier are invalidated and must be 1413 * performed again, if the helper is used in combination with 1414 * direct packet access. 1415 * Return 1416 * 0 on success, or a negative error in case of failure. 1417 * 1418 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1419 * Description 1420 * Copy a NUL terminated string from an unsafe address 1421 * *unsafe_ptr* to *dst*. The *size* should include the 1422 * terminating NUL byte. In case the string length is smaller than 1423 * *size*, the target is not padded with further NUL bytes. If the 1424 * string length is larger than *size*, just *size*-1 bytes are 1425 * copied and the last byte is set to NUL. 1426 * 1427 * On success, the length of the copied string is returned. This 1428 * makes this helper useful in tracing programs for reading 1429 * strings, and more importantly to get its length at runtime. See 1430 * the following snippet: 1431 * 1432 * :: 1433 * 1434 * SEC("kprobe/sys_open") 1435 * void bpf_sys_open(struct pt_regs *ctx) 1436 * { 1437 * char buf[PATHLEN]; // PATHLEN is defined to 256 1438 * int res = bpf_probe_read_str(buf, sizeof(buf), 1439 * ctx->di); 1440 * 1441 * // Consume buf, for example push it to 1442 * // userspace via bpf_perf_event_output(); we 1443 * // can use res (the string length) as event 1444 * // size, after checking its boundaries. 1445 * } 1446 * 1447 * In comparison, using **bpf_probe_read()** helper here instead 1448 * to read the string would require to estimate the length at 1449 * compile time, and would often result in copying more memory 1450 * than necessary. 1451 * 1452 * Another useful use case is when parsing individual process 1453 * arguments or individual environment variables navigating 1454 * *current*\ **->mm->arg_start** and *current*\ 1455 * **->mm->env_start**: using this helper and the return value, 1456 * one can quickly iterate at the right offset of the memory area. 1457 * Return 1458 * On success, the strictly positive length of the string, 1459 * including the trailing NUL character. On error, a negative 1460 * value. 1461 * 1462 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1463 * Description 1464 * If the **struct sk_buff** pointed by *skb* has a known socket, 1465 * retrieve the cookie (generated by the kernel) of this socket. 1466 * If no cookie has been set yet, generate a new cookie. Once 1467 * generated, the socket cookie remains stable for the life of the 1468 * socket. This helper can be useful for monitoring per socket 1469 * networking traffic statistics as it provides a unique socket 1470 * identifier per namespace. 1471 * Return 1472 * A 8-byte long non-decreasing number on success, or 0 if the 1473 * socket field is missing inside *skb*. 1474 * 1475 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1476 * Description 1477 * Equivalent to bpf_get_socket_cookie() helper that accepts 1478 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1479 * Return 1480 * A 8-byte long non-decreasing number. 1481 * 1482 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1483 * Description 1484 * Equivalent to bpf_get_socket_cookie() helper that accepts 1485 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1486 * Return 1487 * A 8-byte long non-decreasing number. 1488 * 1489 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1490 * Return 1491 * The owner UID of the socket associated to *skb*. If the socket 1492 * is **NULL**, or if it is not a full socket (i.e. if it is a 1493 * time-wait or a request socket instead), **overflowuid** value 1494 * is returned (note that **overflowuid** might also be the actual 1495 * UID value for the socket). 1496 * 1497 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1498 * Description 1499 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1500 * to value *hash*. 1501 * Return 1502 * 0 1503 * 1504 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1505 * Description 1506 * Emulate a call to **setsockopt()** on the socket associated to 1507 * *bpf_socket*, which must be a full socket. The *level* at 1508 * which the option resides and the name *optname* of the option 1509 * must be specified, see **setsockopt(2)** for more information. 1510 * The option value of length *optlen* is pointed by *optval*. 1511 * 1512 * This helper actually implements a subset of **setsockopt()**. 1513 * It supports the following *level*\ s: 1514 * 1515 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1516 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1517 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1518 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1519 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1520 * **TCP_BPF_SNDCWND_CLAMP**. 1521 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1522 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1523 * Return 1524 * 0 on success, or a negative error in case of failure. 1525 * 1526 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1527 * Description 1528 * Grow or shrink the room for data in the packet associated to 1529 * *skb* by *len_diff*, and according to the selected *mode*. 1530 * 1531 * There are two supported modes at this time: 1532 * 1533 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1534 * (room space is added or removed below the layer 2 header). 1535 * 1536 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1537 * (room space is added or removed below the layer 3 header). 1538 * 1539 * The following flags are supported at this time: 1540 * 1541 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1542 * Adjusting mss in this way is not allowed for datagrams. 1543 * 1544 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1545 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1546 * Any new space is reserved to hold a tunnel header. 1547 * Configure skb offsets and other fields accordingly. 1548 * 1549 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1550 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1551 * Use with ENCAP_L3 flags to further specify the tunnel type. 1552 * 1553 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1554 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1555 * type; *len* is the length of the inner MAC header. 1556 * 1557 * A call to this helper is susceptible to change the underlying 1558 * packet buffer. Therefore, at load time, all checks on pointers 1559 * previously done by the verifier are invalidated and must be 1560 * performed again, if the helper is used in combination with 1561 * direct packet access. 1562 * Return 1563 * 0 on success, or a negative error in case of failure. 1564 * 1565 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1566 * Description 1567 * Redirect the packet to the endpoint referenced by *map* at 1568 * index *key*. Depending on its type, this *map* can contain 1569 * references to net devices (for forwarding packets through other 1570 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1571 * but this is only implemented for native XDP (with driver 1572 * support) as of this writing). 1573 * 1574 * All values for *flags* are reserved for future usage, and must 1575 * be left at zero. 1576 * 1577 * When used to redirect packets to net devices, this helper 1578 * provides a high performance increase over **bpf_redirect**\ (). 1579 * This is due to various implementation details of the underlying 1580 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1581 * () tries to send packet as a "bulk" to the device. 1582 * Return 1583 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1584 * 1585 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1586 * Description 1587 * Redirect the packet to the socket referenced by *map* (of type 1588 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1589 * egress interfaces can be used for redirection. The 1590 * **BPF_F_INGRESS** value in *flags* is used to make the 1591 * distinction (ingress path is selected if the flag is present, 1592 * egress path otherwise). This is the only flag supported for now. 1593 * Return 1594 * **SK_PASS** on success, or **SK_DROP** on error. 1595 * 1596 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1597 * Description 1598 * Add an entry to, or update a *map* referencing sockets. The 1599 * *skops* is used as a new value for the entry associated to 1600 * *key*. *flags* is one of: 1601 * 1602 * **BPF_NOEXIST** 1603 * The entry for *key* must not exist in the map. 1604 * **BPF_EXIST** 1605 * The entry for *key* must already exist in the map. 1606 * **BPF_ANY** 1607 * No condition on the existence of the entry for *key*. 1608 * 1609 * If the *map* has eBPF programs (parser and verdict), those will 1610 * be inherited by the socket being added. If the socket is 1611 * already attached to eBPF programs, this results in an error. 1612 * Return 1613 * 0 on success, or a negative error in case of failure. 1614 * 1615 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1616 * Description 1617 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1618 * *delta* (which can be positive or negative). Note that this 1619 * operation modifies the address stored in *xdp_md*\ **->data**, 1620 * so the latter must be loaded only after the helper has been 1621 * called. 1622 * 1623 * The use of *xdp_md*\ **->data_meta** is optional and programs 1624 * are not required to use it. The rationale is that when the 1625 * packet is processed with XDP (e.g. as DoS filter), it is 1626 * possible to push further meta data along with it before passing 1627 * to the stack, and to give the guarantee that an ingress eBPF 1628 * program attached as a TC classifier on the same device can pick 1629 * this up for further post-processing. Since TC works with socket 1630 * buffers, it remains possible to set from XDP the **mark** or 1631 * **priority** pointers, or other pointers for the socket buffer. 1632 * Having this scratch space generic and programmable allows for 1633 * more flexibility as the user is free to store whatever meta 1634 * data they need. 1635 * 1636 * A call to this helper is susceptible to change the underlying 1637 * packet buffer. Therefore, at load time, all checks on pointers 1638 * previously done by the verifier are invalidated and must be 1639 * performed again, if the helper is used in combination with 1640 * direct packet access. 1641 * Return 1642 * 0 on success, or a negative error in case of failure. 1643 * 1644 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1645 * Description 1646 * Read the value of a perf event counter, and store it into *buf* 1647 * of size *buf_size*. This helper relies on a *map* of type 1648 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1649 * counter is selected when *map* is updated with perf event file 1650 * descriptors. The *map* is an array whose size is the number of 1651 * available CPUs, and each cell contains a value relative to one 1652 * CPU. The value to retrieve is indicated by *flags*, that 1653 * contains the index of the CPU to look up, masked with 1654 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1655 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1656 * current CPU should be retrieved. 1657 * 1658 * This helper behaves in a way close to 1659 * **bpf_perf_event_read**\ () helper, save that instead of 1660 * just returning the value observed, it fills the *buf* 1661 * structure. This allows for additional data to be retrieved: in 1662 * particular, the enabled and running times (in *buf*\ 1663 * **->enabled** and *buf*\ **->running**, respectively) are 1664 * copied. In general, **bpf_perf_event_read_value**\ () is 1665 * recommended over **bpf_perf_event_read**\ (), which has some 1666 * ABI issues and provides fewer functionalities. 1667 * 1668 * These values are interesting, because hardware PMU (Performance 1669 * Monitoring Unit) counters are limited resources. When there are 1670 * more PMU based perf events opened than available counters, 1671 * kernel will multiplex these events so each event gets certain 1672 * percentage (but not all) of the PMU time. In case that 1673 * multiplexing happens, the number of samples or counter value 1674 * will not reflect the case compared to when no multiplexing 1675 * occurs. This makes comparison between different runs difficult. 1676 * Typically, the counter value should be normalized before 1677 * comparing to other experiments. The usual normalization is done 1678 * as follows. 1679 * 1680 * :: 1681 * 1682 * normalized_counter = counter * t_enabled / t_running 1683 * 1684 * Where t_enabled is the time enabled for event and t_running is 1685 * the time running for event since last normalization. The 1686 * enabled and running times are accumulated since the perf event 1687 * open. To achieve scaling factor between two invocations of an 1688 * eBPF program, users can can use CPU id as the key (which is 1689 * typical for perf array usage model) to remember the previous 1690 * value and do the calculation inside the eBPF program. 1691 * Return 1692 * 0 on success, or a negative error in case of failure. 1693 * 1694 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1695 * Description 1696 * For en eBPF program attached to a perf event, retrieve the 1697 * value of the event counter associated to *ctx* and store it in 1698 * the structure pointed by *buf* and of size *buf_size*. Enabled 1699 * and running times are also stored in the structure (see 1700 * description of helper **bpf_perf_event_read_value**\ () for 1701 * more details). 1702 * Return 1703 * 0 on success, or a negative error in case of failure. 1704 * 1705 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1706 * Description 1707 * Emulate a call to **getsockopt()** on the socket associated to 1708 * *bpf_socket*, which must be a full socket. The *level* at 1709 * which the option resides and the name *optname* of the option 1710 * must be specified, see **getsockopt(2)** for more information. 1711 * The retrieved value is stored in the structure pointed by 1712 * *opval* and of length *optlen*. 1713 * 1714 * This helper actually implements a subset of **getsockopt()**. 1715 * It supports the following *level*\ s: 1716 * 1717 * * **IPPROTO_TCP**, which supports *optname* 1718 * **TCP_CONGESTION**. 1719 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1720 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1721 * Return 1722 * 0 on success, or a negative error in case of failure. 1723 * 1724 * int bpf_override_return(struct pt_reg *regs, u64 rc) 1725 * Description 1726 * Used for error injection, this helper uses kprobes to override 1727 * the return value of the probed function, and to set it to *rc*. 1728 * The first argument is the context *regs* on which the kprobe 1729 * works. 1730 * 1731 * This helper works by setting setting the PC (program counter) 1732 * to an override function which is run in place of the original 1733 * probed function. This means the probed function is not run at 1734 * all. The replacement function just returns with the required 1735 * value. 1736 * 1737 * This helper has security implications, and thus is subject to 1738 * restrictions. It is only available if the kernel was compiled 1739 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1740 * option, and in this case it only works on functions tagged with 1741 * **ALLOW_ERROR_INJECTION** in the kernel code. 1742 * 1743 * Also, the helper is only available for the architectures having 1744 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1745 * x86 architecture is the only one to support this feature. 1746 * Return 1747 * 0 1748 * 1749 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1750 * Description 1751 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1752 * for the full TCP socket associated to *bpf_sock_ops* to 1753 * *argval*. 1754 * 1755 * The primary use of this field is to determine if there should 1756 * be calls to eBPF programs of type 1757 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1758 * code. A program of the same type can change its value, per 1759 * connection and as necessary, when the connection is 1760 * established. This field is directly accessible for reading, but 1761 * this helper must be used for updates in order to return an 1762 * error if an eBPF program tries to set a callback that is not 1763 * supported in the current kernel. 1764 * 1765 * *argval* is a flag array which can combine these flags: 1766 * 1767 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1768 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1769 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1770 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1771 * 1772 * Therefore, this function can be used to clear a callback flag by 1773 * setting the appropriate bit to zero. e.g. to disable the RTO 1774 * callback: 1775 * 1776 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1777 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1778 * 1779 * Here are some examples of where one could call such eBPF 1780 * program: 1781 * 1782 * * When RTO fires. 1783 * * When a packet is retransmitted. 1784 * * When the connection terminates. 1785 * * When a packet is sent. 1786 * * When a packet is received. 1787 * Return 1788 * Code **-EINVAL** if the socket is not a full TCP socket; 1789 * otherwise, a positive number containing the bits that could not 1790 * be set is returned (which comes down to 0 if all bits were set 1791 * as required). 1792 * 1793 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1794 * Description 1795 * This helper is used in programs implementing policies at the 1796 * socket level. If the message *msg* is allowed to pass (i.e. if 1797 * the verdict eBPF program returns **SK_PASS**), redirect it to 1798 * the socket referenced by *map* (of type 1799 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1800 * egress interfaces can be used for redirection. The 1801 * **BPF_F_INGRESS** value in *flags* is used to make the 1802 * distinction (ingress path is selected if the flag is present, 1803 * egress path otherwise). This is the only flag supported for now. 1804 * Return 1805 * **SK_PASS** on success, or **SK_DROP** on error. 1806 * 1807 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1808 * Description 1809 * For socket policies, apply the verdict of the eBPF program to 1810 * the next *bytes* (number of bytes) of message *msg*. 1811 * 1812 * For example, this helper can be used in the following cases: 1813 * 1814 * * A single **sendmsg**\ () or **sendfile**\ () system call 1815 * contains multiple logical messages that the eBPF program is 1816 * supposed to read and for which it should apply a verdict. 1817 * * An eBPF program only cares to read the first *bytes* of a 1818 * *msg*. If the message has a large payload, then setting up 1819 * and calling the eBPF program repeatedly for all bytes, even 1820 * though the verdict is already known, would create unnecessary 1821 * overhead. 1822 * 1823 * When called from within an eBPF program, the helper sets a 1824 * counter internal to the BPF infrastructure, that is used to 1825 * apply the last verdict to the next *bytes*. If *bytes* is 1826 * smaller than the current data being processed from a 1827 * **sendmsg**\ () or **sendfile**\ () system call, the first 1828 * *bytes* will be sent and the eBPF program will be re-run with 1829 * the pointer for start of data pointing to byte number *bytes* 1830 * **+ 1**. If *bytes* is larger than the current data being 1831 * processed, then the eBPF verdict will be applied to multiple 1832 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1833 * consumed. 1834 * 1835 * Note that if a socket closes with the internal counter holding 1836 * a non-zero value, this is not a problem because data is not 1837 * being buffered for *bytes* and is sent as it is received. 1838 * Return 1839 * 0 1840 * 1841 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1842 * Description 1843 * For socket policies, prevent the execution of the verdict eBPF 1844 * program for message *msg* until *bytes* (byte number) have been 1845 * accumulated. 1846 * 1847 * This can be used when one needs a specific number of bytes 1848 * before a verdict can be assigned, even if the data spans 1849 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1850 * case would be a user calling **sendmsg**\ () repeatedly with 1851 * 1-byte long message segments. Obviously, this is bad for 1852 * performance, but it is still valid. If the eBPF program needs 1853 * *bytes* bytes to validate a header, this helper can be used to 1854 * prevent the eBPF program to be called again until *bytes* have 1855 * been accumulated. 1856 * Return 1857 * 0 1858 * 1859 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1860 * Description 1861 * For socket policies, pull in non-linear data from user space 1862 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1863 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1864 * respectively. 1865 * 1866 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1867 * *msg* it can only parse data that the (**data**, **data_end**) 1868 * pointers have already consumed. For **sendmsg**\ () hooks this 1869 * is likely the first scatterlist element. But for calls relying 1870 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1871 * be the range (**0**, **0**) because the data is shared with 1872 * user space and by default the objective is to avoid allowing 1873 * user space to modify data while (or after) eBPF verdict is 1874 * being decided. This helper can be used to pull in data and to 1875 * set the start and end pointer to given values. Data will be 1876 * copied if necessary (i.e. if data was not linear and if start 1877 * and end pointers do not point to the same chunk). 1878 * 1879 * A call to this helper is susceptible to change the underlying 1880 * packet buffer. Therefore, at load time, all checks on pointers 1881 * previously done by the verifier are invalidated and must be 1882 * performed again, if the helper is used in combination with 1883 * direct packet access. 1884 * 1885 * All values for *flags* are reserved for future usage, and must 1886 * be left at zero. 1887 * Return 1888 * 0 on success, or a negative error in case of failure. 1889 * 1890 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1891 * Description 1892 * Bind the socket associated to *ctx* to the address pointed by 1893 * *addr*, of length *addr_len*. This allows for making outgoing 1894 * connection from the desired IP address, which can be useful for 1895 * example when all processes inside a cgroup should use one 1896 * single IP address on a host that has multiple IP configured. 1897 * 1898 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1899 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1900 * **AF_INET6**). Looking for a free port to bind to can be 1901 * expensive, therefore binding to port is not permitted by the 1902 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1903 * must be set to zero. 1904 * Return 1905 * 0 on success, or a negative error in case of failure. 1906 * 1907 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1908 * Description 1909 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1910 * only possible to shrink the packet as of this writing, 1911 * therefore *delta* must be a negative integer. 1912 * 1913 * A call to this helper is susceptible to change the underlying 1914 * packet buffer. Therefore, at load time, all checks on pointers 1915 * previously done by the verifier are invalidated and must be 1916 * performed again, if the helper is used in combination with 1917 * direct packet access. 1918 * Return 1919 * 0 on success, or a negative error in case of failure. 1920 * 1921 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1922 * Description 1923 * Retrieve the XFRM state (IP transform framework, see also 1924 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1925 * 1926 * The retrieved value is stored in the **struct bpf_xfrm_state** 1927 * pointed by *xfrm_state* and of length *size*. 1928 * 1929 * All values for *flags* are reserved for future usage, and must 1930 * be left at zero. 1931 * 1932 * This helper is available only if the kernel was compiled with 1933 * **CONFIG_XFRM** configuration option. 1934 * Return 1935 * 0 on success, or a negative error in case of failure. 1936 * 1937 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1938 * Description 1939 * Return a user or a kernel stack in bpf program provided buffer. 1940 * To achieve this, the helper needs *ctx*, which is a pointer 1941 * to the context on which the tracing program is executed. 1942 * To store the stacktrace, the bpf program provides *buf* with 1943 * a nonnegative *size*. 1944 * 1945 * The last argument, *flags*, holds the number of stack frames to 1946 * skip (from 0 to 255), masked with 1947 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1948 * the following flags: 1949 * 1950 * **BPF_F_USER_STACK** 1951 * Collect a user space stack instead of a kernel stack. 1952 * **BPF_F_USER_BUILD_ID** 1953 * Collect buildid+offset instead of ips for user stack, 1954 * only valid if **BPF_F_USER_STACK** is also specified. 1955 * 1956 * **bpf_get_stack**\ () can collect up to 1957 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1958 * to sufficient large buffer size. Note that 1959 * this limit can be controlled with the **sysctl** program, and 1960 * that it should be manually increased in order to profile long 1961 * user stacks (such as stacks for Java programs). To do so, use: 1962 * 1963 * :: 1964 * 1965 * # sysctl kernel.perf_event_max_stack=<new value> 1966 * Return 1967 * A non-negative value equal to or less than *size* on success, 1968 * or a negative error in case of failure. 1969 * 1970 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1971 * Description 1972 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1973 * it provides an easy way to load *len* bytes from *offset* 1974 * from the packet associated to *skb*, into the buffer pointed 1975 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1976 * a fifth argument *start_header* exists in order to select a 1977 * base offset to start from. *start_header* can be one of: 1978 * 1979 * **BPF_HDR_START_MAC** 1980 * Base offset to load data from is *skb*'s mac header. 1981 * **BPF_HDR_START_NET** 1982 * Base offset to load data from is *skb*'s network header. 1983 * 1984 * In general, "direct packet access" is the preferred method to 1985 * access packet data, however, this helper is in particular useful 1986 * in socket filters where *skb*\ **->data** does not always point 1987 * to the start of the mac header and where "direct packet access" 1988 * is not available. 1989 * Return 1990 * 0 on success, or a negative error in case of failure. 1991 * 1992 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 1993 * Description 1994 * Do FIB lookup in kernel tables using parameters in *params*. 1995 * If lookup is successful and result shows packet is to be 1996 * forwarded, the neighbor tables are searched for the nexthop. 1997 * If successful (ie., FIB lookup shows forwarding and nexthop 1998 * is resolved), the nexthop address is returned in ipv4_dst 1999 * or ipv6_dst based on family, smac is set to mac address of 2000 * egress device, dmac is set to nexthop mac address, rt_metric 2001 * is set to metric from route (IPv4/IPv6 only), and ifindex 2002 * is set to the device index of the nexthop from the FIB lookup. 2003 * 2004 * *plen* argument is the size of the passed in struct. 2005 * *flags* argument can be a combination of one or more of the 2006 * following values: 2007 * 2008 * **BPF_FIB_LOOKUP_DIRECT** 2009 * Do a direct table lookup vs full lookup using FIB 2010 * rules. 2011 * **BPF_FIB_LOOKUP_OUTPUT** 2012 * Perform lookup from an egress perspective (default is 2013 * ingress). 2014 * 2015 * *ctx* is either **struct xdp_md** for XDP programs or 2016 * **struct sk_buff** tc cls_act programs. 2017 * Return 2018 * * < 0 if any input argument is invalid 2019 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2020 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2021 * packet is not forwarded or needs assist from full stack 2022 * 2023 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 2024 * Description 2025 * Add an entry to, or update a sockhash *map* referencing sockets. 2026 * The *skops* is used as a new value for the entry associated to 2027 * *key*. *flags* is one of: 2028 * 2029 * **BPF_NOEXIST** 2030 * The entry for *key* must not exist in the map. 2031 * **BPF_EXIST** 2032 * The entry for *key* must already exist in the map. 2033 * **BPF_ANY** 2034 * No condition on the existence of the entry for *key*. 2035 * 2036 * If the *map* has eBPF programs (parser and verdict), those will 2037 * be inherited by the socket being added. If the socket is 2038 * already attached to eBPF programs, this results in an error. 2039 * Return 2040 * 0 on success, or a negative error in case of failure. 2041 * 2042 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2043 * Description 2044 * This helper is used in programs implementing policies at the 2045 * socket level. If the message *msg* is allowed to pass (i.e. if 2046 * the verdict eBPF program returns **SK_PASS**), redirect it to 2047 * the socket referenced by *map* (of type 2048 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2049 * egress interfaces can be used for redirection. The 2050 * **BPF_F_INGRESS** value in *flags* is used to make the 2051 * distinction (ingress path is selected if the flag is present, 2052 * egress path otherwise). This is the only flag supported for now. 2053 * Return 2054 * **SK_PASS** on success, or **SK_DROP** on error. 2055 * 2056 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2057 * Description 2058 * This helper is used in programs implementing policies at the 2059 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2060 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2061 * to the socket referenced by *map* (of type 2062 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2063 * egress interfaces can be used for redirection. The 2064 * **BPF_F_INGRESS** value in *flags* is used to make the 2065 * distinction (ingress path is selected if the flag is present, 2066 * egress otherwise). This is the only flag supported for now. 2067 * Return 2068 * **SK_PASS** on success, or **SK_DROP** on error. 2069 * 2070 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2071 * Description 2072 * Encapsulate the packet associated to *skb* within a Layer 3 2073 * protocol header. This header is provided in the buffer at 2074 * address *hdr*, with *len* its size in bytes. *type* indicates 2075 * the protocol of the header and can be one of: 2076 * 2077 * **BPF_LWT_ENCAP_SEG6** 2078 * IPv6 encapsulation with Segment Routing Header 2079 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2080 * the IPv6 header is computed by the kernel. 2081 * **BPF_LWT_ENCAP_SEG6_INLINE** 2082 * Only works if *skb* contains an IPv6 packet. Insert a 2083 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2084 * the IPv6 header. 2085 * **BPF_LWT_ENCAP_IP** 2086 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2087 * must be IPv4 or IPv6, followed by zero or more 2088 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2089 * total bytes in all prepended headers. Please note that 2090 * if **skb_is_gso**\ (*skb*) is true, no more than two 2091 * headers can be prepended, and the inner header, if 2092 * present, should be either GRE or UDP/GUE. 2093 * 2094 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2095 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2096 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2097 * **BPF_PROG_TYPE_LWT_XMIT**. 2098 * 2099 * A call to this helper is susceptible to change the underlying 2100 * packet buffer. Therefore, at load time, all checks on pointers 2101 * previously done by the verifier are invalidated and must be 2102 * performed again, if the helper is used in combination with 2103 * direct packet access. 2104 * Return 2105 * 0 on success, or a negative error in case of failure. 2106 * 2107 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2108 * Description 2109 * Store *len* bytes from address *from* into the packet 2110 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2111 * inside the outermost IPv6 Segment Routing Header can be 2112 * modified through this helper. 2113 * 2114 * A call to this helper is susceptible to change the underlying 2115 * packet buffer. Therefore, at load time, all checks on pointers 2116 * previously done by the verifier are invalidated and must be 2117 * performed again, if the helper is used in combination with 2118 * direct packet access. 2119 * Return 2120 * 0 on success, or a negative error in case of failure. 2121 * 2122 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2123 * Description 2124 * Adjust the size allocated to TLVs in the outermost IPv6 2125 * Segment Routing Header contained in the packet associated to 2126 * *skb*, at position *offset* by *delta* bytes. Only offsets 2127 * after the segments are accepted. *delta* can be as well 2128 * positive (growing) as negative (shrinking). 2129 * 2130 * A call to this helper is susceptible to change the underlying 2131 * packet buffer. Therefore, at load time, all checks on pointers 2132 * previously done by the verifier are invalidated and must be 2133 * performed again, if the helper is used in combination with 2134 * direct packet access. 2135 * Return 2136 * 0 on success, or a negative error in case of failure. 2137 * 2138 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2139 * Description 2140 * Apply an IPv6 Segment Routing action of type *action* to the 2141 * packet associated to *skb*. Each action takes a parameter 2142 * contained at address *param*, and of length *param_len* bytes. 2143 * *action* can be one of: 2144 * 2145 * **SEG6_LOCAL_ACTION_END_X** 2146 * End.X action: Endpoint with Layer-3 cross-connect. 2147 * Type of *param*: **struct in6_addr**. 2148 * **SEG6_LOCAL_ACTION_END_T** 2149 * End.T action: Endpoint with specific IPv6 table lookup. 2150 * Type of *param*: **int**. 2151 * **SEG6_LOCAL_ACTION_END_B6** 2152 * End.B6 action: Endpoint bound to an SRv6 policy. 2153 * Type of *param*: **struct ipv6_sr_hdr**. 2154 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2155 * End.B6.Encap action: Endpoint bound to an SRv6 2156 * encapsulation policy. 2157 * Type of *param*: **struct ipv6_sr_hdr**. 2158 * 2159 * A call to this helper is susceptible to change the underlying 2160 * packet buffer. Therefore, at load time, all checks on pointers 2161 * previously done by the verifier are invalidated and must be 2162 * performed again, if the helper is used in combination with 2163 * direct packet access. 2164 * Return 2165 * 0 on success, or a negative error in case of failure. 2166 * 2167 * int bpf_rc_repeat(void *ctx) 2168 * Description 2169 * This helper is used in programs implementing IR decoding, to 2170 * report a successfully decoded repeat key message. This delays 2171 * the generation of a key up event for previously generated 2172 * key down event. 2173 * 2174 * Some IR protocols like NEC have a special IR message for 2175 * repeating last button, for when a button is held down. 2176 * 2177 * The *ctx* should point to the lirc sample as passed into 2178 * the program. 2179 * 2180 * This helper is only available is the kernel was compiled with 2181 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2182 * "**y**". 2183 * Return 2184 * 0 2185 * 2186 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2187 * Description 2188 * This helper is used in programs implementing IR decoding, to 2189 * report a successfully decoded key press with *scancode*, 2190 * *toggle* value in the given *protocol*. The scancode will be 2191 * translated to a keycode using the rc keymap, and reported as 2192 * an input key down event. After a period a key up event is 2193 * generated. This period can be extended by calling either 2194 * **bpf_rc_keydown**\ () again with the same values, or calling 2195 * **bpf_rc_repeat**\ (). 2196 * 2197 * Some protocols include a toggle bit, in case the button was 2198 * released and pressed again between consecutive scancodes. 2199 * 2200 * The *ctx* should point to the lirc sample as passed into 2201 * the program. 2202 * 2203 * The *protocol* is the decoded protocol number (see 2204 * **enum rc_proto** for some predefined values). 2205 * 2206 * This helper is only available is the kernel was compiled with 2207 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2208 * "**y**". 2209 * Return 2210 * 0 2211 * 2212 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2213 * Description 2214 * Return the cgroup v2 id of the socket associated with the *skb*. 2215 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2216 * helper for cgroup v1 by providing a tag resp. identifier that 2217 * can be matched on or used for map lookups e.g. to implement 2218 * policy. The cgroup v2 id of a given path in the hierarchy is 2219 * exposed in user space through the f_handle API in order to get 2220 * to the same 64-bit id. 2221 * 2222 * This helper can be used on TC egress path, but not on ingress, 2223 * and is available only if the kernel was compiled with the 2224 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2225 * Return 2226 * The id is returned or 0 in case the id could not be retrieved. 2227 * 2228 * u64 bpf_get_current_cgroup_id(void) 2229 * Return 2230 * A 64-bit integer containing the current cgroup id based 2231 * on the cgroup within which the current task is running. 2232 * 2233 * void *bpf_get_local_storage(void *map, u64 flags) 2234 * Description 2235 * Get the pointer to the local storage area. 2236 * The type and the size of the local storage is defined 2237 * by the *map* argument. 2238 * The *flags* meaning is specific for each map type, 2239 * and has to be 0 for cgroup local storage. 2240 * 2241 * Depending on the BPF program type, a local storage area 2242 * can be shared between multiple instances of the BPF program, 2243 * running simultaneously. 2244 * 2245 * A user should care about the synchronization by himself. 2246 * For example, by using the **BPF_STX_XADD** instruction to alter 2247 * the shared data. 2248 * Return 2249 * A pointer to the local storage area. 2250 * 2251 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2252 * Description 2253 * Select a **SO_REUSEPORT** socket from a 2254 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2255 * It checks the selected socket is matching the incoming 2256 * request in the socket buffer. 2257 * Return 2258 * 0 on success, or a negative error in case of failure. 2259 * 2260 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2261 * Description 2262 * Return id of cgroup v2 that is ancestor of cgroup associated 2263 * with the *skb* at the *ancestor_level*. The root cgroup is at 2264 * *ancestor_level* zero and each step down the hierarchy 2265 * increments the level. If *ancestor_level* == level of cgroup 2266 * associated with *skb*, then return value will be same as that 2267 * of **bpf_skb_cgroup_id**\ (). 2268 * 2269 * The helper is useful to implement policies based on cgroups 2270 * that are upper in hierarchy than immediate cgroup associated 2271 * with *skb*. 2272 * 2273 * The format of returned id and helper limitations are same as in 2274 * **bpf_skb_cgroup_id**\ (). 2275 * Return 2276 * The id is returned or 0 in case the id could not be retrieved. 2277 * 2278 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2279 * Description 2280 * Look for TCP socket matching *tuple*, optionally in a child 2281 * network namespace *netns*. The return value must be checked, 2282 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2283 * 2284 * The *ctx* should point to the context of the program, such as 2285 * the skb or socket (depending on the hook in use). This is used 2286 * to determine the base network namespace for the lookup. 2287 * 2288 * *tuple_size* must be one of: 2289 * 2290 * **sizeof**\ (*tuple*\ **->ipv4**) 2291 * Look for an IPv4 socket. 2292 * **sizeof**\ (*tuple*\ **->ipv6**) 2293 * Look for an IPv6 socket. 2294 * 2295 * If the *netns* is a negative signed 32-bit integer, then the 2296 * socket lookup table in the netns associated with the *ctx* will 2297 * will be used. For the TC hooks, this is the netns of the device 2298 * in the skb. For socket hooks, this is the netns of the socket. 2299 * If *netns* is any other signed 32-bit value greater than or 2300 * equal to zero then it specifies the ID of the netns relative to 2301 * the netns associated with the *ctx*. *netns* values beyond the 2302 * range of 32-bit integers are reserved for future use. 2303 * 2304 * All values for *flags* are reserved for future usage, and must 2305 * be left at zero. 2306 * 2307 * This helper is available only if the kernel was compiled with 2308 * **CONFIG_NET** configuration option. 2309 * Return 2310 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2311 * For sockets with reuseport option, the **struct bpf_sock** 2312 * result is from *reuse*\ **->socks**\ [] using the hash of the 2313 * tuple. 2314 * 2315 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2316 * Description 2317 * Look for UDP socket matching *tuple*, optionally in a child 2318 * network namespace *netns*. The return value must be checked, 2319 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2320 * 2321 * The *ctx* should point to the context of the program, such as 2322 * the skb or socket (depending on the hook in use). This is used 2323 * to determine the base network namespace for the lookup. 2324 * 2325 * *tuple_size* must be one of: 2326 * 2327 * **sizeof**\ (*tuple*\ **->ipv4**) 2328 * Look for an IPv4 socket. 2329 * **sizeof**\ (*tuple*\ **->ipv6**) 2330 * Look for an IPv6 socket. 2331 * 2332 * If the *netns* is a negative signed 32-bit integer, then the 2333 * socket lookup table in the netns associated with the *ctx* will 2334 * will be used. For the TC hooks, this is the netns of the device 2335 * in the skb. For socket hooks, this is the netns of the socket. 2336 * If *netns* is any other signed 32-bit value greater than or 2337 * equal to zero then it specifies the ID of the netns relative to 2338 * the netns associated with the *ctx*. *netns* values beyond the 2339 * range of 32-bit integers are reserved for future use. 2340 * 2341 * All values for *flags* are reserved for future usage, and must 2342 * be left at zero. 2343 * 2344 * This helper is available only if the kernel was compiled with 2345 * **CONFIG_NET** configuration option. 2346 * Return 2347 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2348 * For sockets with reuseport option, the **struct bpf_sock** 2349 * result is from *reuse*\ **->socks**\ [] using the hash of the 2350 * tuple. 2351 * 2352 * int bpf_sk_release(struct bpf_sock *sock) 2353 * Description 2354 * Release the reference held by *sock*. *sock* must be a 2355 * non-**NULL** pointer that was returned from 2356 * **bpf_sk_lookup_xxx**\ (). 2357 * Return 2358 * 0 on success, or a negative error in case of failure. 2359 * 2360 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2361 * Description 2362 * Push an element *value* in *map*. *flags* is one of: 2363 * 2364 * **BPF_EXIST** 2365 * If the queue/stack is full, the oldest element is 2366 * removed to make room for this. 2367 * Return 2368 * 0 on success, or a negative error in case of failure. 2369 * 2370 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2371 * Description 2372 * Pop an element from *map*. 2373 * Return 2374 * 0 on success, or a negative error in case of failure. 2375 * 2376 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2377 * Description 2378 * Get an element from *map* without removing it. 2379 * Return 2380 * 0 on success, or a negative error in case of failure. 2381 * 2382 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 2383 * Description 2384 * For socket policies, insert *len* bytes into *msg* at offset 2385 * *start*. 2386 * 2387 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2388 * *msg* it may want to insert metadata or options into the *msg*. 2389 * This can later be read and used by any of the lower layer BPF 2390 * hooks. 2391 * 2392 * This helper may fail if under memory pressure (a malloc 2393 * fails) in these cases BPF programs will get an appropriate 2394 * error and BPF programs will need to handle them. 2395 * Return 2396 * 0 on success, or a negative error in case of failure. 2397 * 2398 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags) 2399 * Description 2400 * Will remove *pop* bytes from a *msg* starting at byte *start*. 2401 * This may result in **ENOMEM** errors under certain situations if 2402 * an allocation and copy are required due to a full ring buffer. 2403 * However, the helper will try to avoid doing the allocation 2404 * if possible. Other errors can occur if input parameters are 2405 * invalid either due to *start* byte not being valid part of *msg* 2406 * payload and/or *pop* value being to large. 2407 * Return 2408 * 0 on success, or a negative error in case of failure. 2409 * 2410 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2411 * Description 2412 * This helper is used in programs implementing IR decoding, to 2413 * report a successfully decoded pointer movement. 2414 * 2415 * The *ctx* should point to the lirc sample as passed into 2416 * the program. 2417 * 2418 * This helper is only available is the kernel was compiled with 2419 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2420 * "**y**". 2421 * Return 2422 * 0 2423 * 2424 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2425 * Description 2426 * Acquire a spinlock represented by the pointer *lock*, which is 2427 * stored as part of a value of a map. Taking the lock allows to 2428 * safely update the rest of the fields in that value. The 2429 * spinlock can (and must) later be released with a call to 2430 * **bpf_spin_unlock**\ (\ *lock*\ ). 2431 * 2432 * Spinlocks in BPF programs come with a number of restrictions 2433 * and constraints: 2434 * 2435 * * **bpf_spin_lock** objects are only allowed inside maps of 2436 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2437 * list could be extended in the future). 2438 * * BTF description of the map is mandatory. 2439 * * The BPF program can take ONE lock at a time, since taking two 2440 * or more could cause dead locks. 2441 * * Only one **struct bpf_spin_lock** is allowed per map element. 2442 * * When the lock is taken, calls (either BPF to BPF or helpers) 2443 * are not allowed. 2444 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2445 * allowed inside a spinlock-ed region. 2446 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2447 * the lock, on all execution paths, before it returns. 2448 * * The BPF program can access **struct bpf_spin_lock** only via 2449 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2450 * helpers. Loading or storing data into the **struct 2451 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2452 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2453 * of the map value must be a struct and have **struct 2454 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2455 * Nested lock inside another struct is not allowed. 2456 * * The **struct bpf_spin_lock** *lock* field in a map value must 2457 * be aligned on a multiple of 4 bytes in that value. 2458 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2459 * the **bpf_spin_lock** field to user space. 2460 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2461 * a BPF program, do not update the **bpf_spin_lock** field. 2462 * * **bpf_spin_lock** cannot be on the stack or inside a 2463 * networking packet (it can only be inside of a map values). 2464 * * **bpf_spin_lock** is available to root only. 2465 * * Tracing programs and socket filter programs cannot use 2466 * **bpf_spin_lock**\ () due to insufficient preemption checks 2467 * (but this may change in the future). 2468 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2469 * Return 2470 * 0 2471 * 2472 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2473 * Description 2474 * Release the *lock* previously locked by a call to 2475 * **bpf_spin_lock**\ (\ *lock*\ ). 2476 * Return 2477 * 0 2478 * 2479 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2480 * Description 2481 * This helper gets a **struct bpf_sock** pointer such 2482 * that all the fields in this **bpf_sock** can be accessed. 2483 * Return 2484 * A **struct bpf_sock** pointer on success, or **NULL** in 2485 * case of failure. 2486 * 2487 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2488 * Description 2489 * This helper gets a **struct bpf_tcp_sock** pointer from a 2490 * **struct bpf_sock** pointer. 2491 * Return 2492 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2493 * case of failure. 2494 * 2495 * int bpf_skb_ecn_set_ce(struct sk_buf *skb) 2496 * Description 2497 * Set ECN (Explicit Congestion Notification) field of IP header 2498 * to **CE** (Congestion Encountered) if current value is **ECT** 2499 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2500 * and IPv4. 2501 * Return 2502 * 1 if the **CE** flag is set (either by the current helper call 2503 * or because it was already present), 0 if it is not set. 2504 * 2505 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2506 * Description 2507 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2508 * **bpf_sk_release**\ () is unnecessary and not allowed. 2509 * Return 2510 * A **struct bpf_sock** pointer on success, or **NULL** in 2511 * case of failure. 2512 * 2513 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2514 * Description 2515 * Look for TCP socket matching *tuple*, optionally in a child 2516 * network namespace *netns*. The return value must be checked, 2517 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2518 * 2519 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2520 * that it also returns timewait or request sockets. Use 2521 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2522 * full structure. 2523 * 2524 * This helper is available only if the kernel was compiled with 2525 * **CONFIG_NET** configuration option. 2526 * Return 2527 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2528 * For sockets with reuseport option, the **struct bpf_sock** 2529 * result is from *reuse*\ **->socks**\ [] using the hash of the 2530 * tuple. 2531 * 2532 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2533 * Description 2534 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2535 * the listening socket in *sk*. 2536 * 2537 * *iph* points to the start of the IPv4 or IPv6 header, while 2538 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2539 * **sizeof**\ (**struct ip6hdr**). 2540 * 2541 * *th* points to the start of the TCP header, while *th_len* 2542 * contains **sizeof**\ (**struct tcphdr**). 2543 * 2544 * Return 2545 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2546 * error otherwise. 2547 * 2548 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2549 * Description 2550 * Get name of sysctl in /proc/sys/ and copy it into provided by 2551 * program buffer *buf* of size *buf_len*. 2552 * 2553 * The buffer is always NUL terminated, unless it's zero-sized. 2554 * 2555 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2556 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2557 * only (e.g. "tcp_mem"). 2558 * Return 2559 * Number of character copied (not including the trailing NUL). 2560 * 2561 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2562 * truncated name in this case). 2563 * 2564 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2565 * Description 2566 * Get current value of sysctl as it is presented in /proc/sys 2567 * (incl. newline, etc), and copy it as a string into provided 2568 * by program buffer *buf* of size *buf_len*. 2569 * 2570 * The whole value is copied, no matter what file position user 2571 * space issued e.g. sys_read at. 2572 * 2573 * The buffer is always NUL terminated, unless it's zero-sized. 2574 * Return 2575 * Number of character copied (not including the trailing NUL). 2576 * 2577 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2578 * truncated name in this case). 2579 * 2580 * **-EINVAL** if current value was unavailable, e.g. because 2581 * sysctl is uninitialized and read returns -EIO for it. 2582 * 2583 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2584 * Description 2585 * Get new value being written by user space to sysctl (before 2586 * the actual write happens) and copy it as a string into 2587 * provided by program buffer *buf* of size *buf_len*. 2588 * 2589 * User space may write new value at file position > 0. 2590 * 2591 * The buffer is always NUL terminated, unless it's zero-sized. 2592 * Return 2593 * Number of character copied (not including the trailing NUL). 2594 * 2595 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2596 * truncated name in this case). 2597 * 2598 * **-EINVAL** if sysctl is being read. 2599 * 2600 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2601 * Description 2602 * Override new value being written by user space to sysctl with 2603 * value provided by program in buffer *buf* of size *buf_len*. 2604 * 2605 * *buf* should contain a string in same form as provided by user 2606 * space on sysctl write. 2607 * 2608 * User space may write new value at file position > 0. To override 2609 * the whole sysctl value file position should be set to zero. 2610 * Return 2611 * 0 on success. 2612 * 2613 * **-E2BIG** if the *buf_len* is too big. 2614 * 2615 * **-EINVAL** if sysctl is being read. 2616 * 2617 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2618 * Description 2619 * Convert the initial part of the string from buffer *buf* of 2620 * size *buf_len* to a long integer according to the given base 2621 * and save the result in *res*. 2622 * 2623 * The string may begin with an arbitrary amount of white space 2624 * (as determined by **isspace**\ (3)) followed by a single 2625 * optional '**-**' sign. 2626 * 2627 * Five least significant bits of *flags* encode base, other bits 2628 * are currently unused. 2629 * 2630 * Base must be either 8, 10, 16 or 0 to detect it automatically 2631 * similar to user space **strtol**\ (3). 2632 * Return 2633 * Number of characters consumed on success. Must be positive but 2634 * no more than *buf_len*. 2635 * 2636 * **-EINVAL** if no valid digits were found or unsupported base 2637 * was provided. 2638 * 2639 * **-ERANGE** if resulting value was out of range. 2640 * 2641 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2642 * Description 2643 * Convert the initial part of the string from buffer *buf* of 2644 * size *buf_len* to an unsigned long integer according to the 2645 * given base and save the result in *res*. 2646 * 2647 * The string may begin with an arbitrary amount of white space 2648 * (as determined by **isspace**\ (3)). 2649 * 2650 * Five least significant bits of *flags* encode base, other bits 2651 * are currently unused. 2652 * 2653 * Base must be either 8, 10, 16 or 0 to detect it automatically 2654 * similar to user space **strtoul**\ (3). 2655 * Return 2656 * Number of characters consumed on success. Must be positive but 2657 * no more than *buf_len*. 2658 * 2659 * **-EINVAL** if no valid digits were found or unsupported base 2660 * was provided. 2661 * 2662 * **-ERANGE** if resulting value was out of range. 2663 * 2664 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2665 * Description 2666 * Get a bpf-local-storage from a *sk*. 2667 * 2668 * Logically, it could be thought of getting the value from 2669 * a *map* with *sk* as the **key**. From this 2670 * perspective, the usage is not much different from 2671 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2672 * helper enforces the key must be a full socket and the map must 2673 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2674 * 2675 * Underneath, the value is stored locally at *sk* instead of 2676 * the *map*. The *map* is used as the bpf-local-storage 2677 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2678 * searched against all bpf-local-storages residing at *sk*. 2679 * 2680 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2681 * used such that a new bpf-local-storage will be 2682 * created if one does not exist. *value* can be used 2683 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2684 * the initial value of a bpf-local-storage. If *value* is 2685 * **NULL**, the new bpf-local-storage will be zero initialized. 2686 * Return 2687 * A bpf-local-storage pointer is returned on success. 2688 * 2689 * **NULL** if not found or there was an error in adding 2690 * a new bpf-local-storage. 2691 * 2692 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2693 * Description 2694 * Delete a bpf-local-storage from a *sk*. 2695 * Return 2696 * 0 on success. 2697 * 2698 * **-ENOENT** if the bpf-local-storage cannot be found. 2699 * 2700 * int bpf_send_signal(u32 sig) 2701 * Description 2702 * Send signal *sig* to the current task. 2703 * Return 2704 * 0 on success or successfully queued. 2705 * 2706 * **-EBUSY** if work queue under nmi is full. 2707 * 2708 * **-EINVAL** if *sig* is invalid. 2709 * 2710 * **-EPERM** if no permission to send the *sig*. 2711 * 2712 * **-EAGAIN** if bpf program can try again. 2713 */ 2714 #define __BPF_FUNC_MAPPER(FN) \ 2715 FN(unspec), \ 2716 FN(map_lookup_elem), \ 2717 FN(map_update_elem), \ 2718 FN(map_delete_elem), \ 2719 FN(probe_read), \ 2720 FN(ktime_get_ns), \ 2721 FN(trace_printk), \ 2722 FN(get_prandom_u32), \ 2723 FN(get_smp_processor_id), \ 2724 FN(skb_store_bytes), \ 2725 FN(l3_csum_replace), \ 2726 FN(l4_csum_replace), \ 2727 FN(tail_call), \ 2728 FN(clone_redirect), \ 2729 FN(get_current_pid_tgid), \ 2730 FN(get_current_uid_gid), \ 2731 FN(get_current_comm), \ 2732 FN(get_cgroup_classid), \ 2733 FN(skb_vlan_push), \ 2734 FN(skb_vlan_pop), \ 2735 FN(skb_get_tunnel_key), \ 2736 FN(skb_set_tunnel_key), \ 2737 FN(perf_event_read), \ 2738 FN(redirect), \ 2739 FN(get_route_realm), \ 2740 FN(perf_event_output), \ 2741 FN(skb_load_bytes), \ 2742 FN(get_stackid), \ 2743 FN(csum_diff), \ 2744 FN(skb_get_tunnel_opt), \ 2745 FN(skb_set_tunnel_opt), \ 2746 FN(skb_change_proto), \ 2747 FN(skb_change_type), \ 2748 FN(skb_under_cgroup), \ 2749 FN(get_hash_recalc), \ 2750 FN(get_current_task), \ 2751 FN(probe_write_user), \ 2752 FN(current_task_under_cgroup), \ 2753 FN(skb_change_tail), \ 2754 FN(skb_pull_data), \ 2755 FN(csum_update), \ 2756 FN(set_hash_invalid), \ 2757 FN(get_numa_node_id), \ 2758 FN(skb_change_head), \ 2759 FN(xdp_adjust_head), \ 2760 FN(probe_read_str), \ 2761 FN(get_socket_cookie), \ 2762 FN(get_socket_uid), \ 2763 FN(set_hash), \ 2764 FN(setsockopt), \ 2765 FN(skb_adjust_room), \ 2766 FN(redirect_map), \ 2767 FN(sk_redirect_map), \ 2768 FN(sock_map_update), \ 2769 FN(xdp_adjust_meta), \ 2770 FN(perf_event_read_value), \ 2771 FN(perf_prog_read_value), \ 2772 FN(getsockopt), \ 2773 FN(override_return), \ 2774 FN(sock_ops_cb_flags_set), \ 2775 FN(msg_redirect_map), \ 2776 FN(msg_apply_bytes), \ 2777 FN(msg_cork_bytes), \ 2778 FN(msg_pull_data), \ 2779 FN(bind), \ 2780 FN(xdp_adjust_tail), \ 2781 FN(skb_get_xfrm_state), \ 2782 FN(get_stack), \ 2783 FN(skb_load_bytes_relative), \ 2784 FN(fib_lookup), \ 2785 FN(sock_hash_update), \ 2786 FN(msg_redirect_hash), \ 2787 FN(sk_redirect_hash), \ 2788 FN(lwt_push_encap), \ 2789 FN(lwt_seg6_store_bytes), \ 2790 FN(lwt_seg6_adjust_srh), \ 2791 FN(lwt_seg6_action), \ 2792 FN(rc_repeat), \ 2793 FN(rc_keydown), \ 2794 FN(skb_cgroup_id), \ 2795 FN(get_current_cgroup_id), \ 2796 FN(get_local_storage), \ 2797 FN(sk_select_reuseport), \ 2798 FN(skb_ancestor_cgroup_id), \ 2799 FN(sk_lookup_tcp), \ 2800 FN(sk_lookup_udp), \ 2801 FN(sk_release), \ 2802 FN(map_push_elem), \ 2803 FN(map_pop_elem), \ 2804 FN(map_peek_elem), \ 2805 FN(msg_push_data), \ 2806 FN(msg_pop_data), \ 2807 FN(rc_pointer_rel), \ 2808 FN(spin_lock), \ 2809 FN(spin_unlock), \ 2810 FN(sk_fullsock), \ 2811 FN(tcp_sock), \ 2812 FN(skb_ecn_set_ce), \ 2813 FN(get_listener_sock), \ 2814 FN(skc_lookup_tcp), \ 2815 FN(tcp_check_syncookie), \ 2816 FN(sysctl_get_name), \ 2817 FN(sysctl_get_current_value), \ 2818 FN(sysctl_get_new_value), \ 2819 FN(sysctl_set_new_value), \ 2820 FN(strtol), \ 2821 FN(strtoul), \ 2822 FN(sk_storage_get), \ 2823 FN(sk_storage_delete), \ 2824 FN(send_signal), 2825 2826 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2827 * function eBPF program intends to call 2828 */ 2829 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2830 enum bpf_func_id { 2831 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2832 __BPF_FUNC_MAX_ID, 2833 }; 2834 #undef __BPF_ENUM_FN 2835 2836 /* All flags used by eBPF helper functions, placed here. */ 2837 2838 /* BPF_FUNC_skb_store_bytes flags. */ 2839 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2840 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2841 2842 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2843 * First 4 bits are for passing the header field size. 2844 */ 2845 #define BPF_F_HDR_FIELD_MASK 0xfULL 2846 2847 /* BPF_FUNC_l4_csum_replace flags. */ 2848 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2849 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2850 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2851 2852 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2853 #define BPF_F_INGRESS (1ULL << 0) 2854 2855 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2856 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2857 2858 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2859 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2860 #define BPF_F_USER_STACK (1ULL << 8) 2861 /* flags used by BPF_FUNC_get_stackid only. */ 2862 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2863 #define BPF_F_REUSE_STACKID (1ULL << 10) 2864 /* flags used by BPF_FUNC_get_stack only. */ 2865 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2866 2867 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2868 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2869 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2870 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2871 2872 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2873 * BPF_FUNC_perf_event_read_value flags. 2874 */ 2875 #define BPF_F_INDEX_MASK 0xffffffffULL 2876 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2877 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2878 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2879 2880 /* Current network namespace */ 2881 #define BPF_F_CURRENT_NETNS (-1L) 2882 2883 /* BPF_FUNC_skb_adjust_room flags. */ 2884 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 2885 2886 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 2887 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 2888 2889 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 2890 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 2891 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 2892 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 2893 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 2894 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 2895 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 2896 2897 /* BPF_FUNC_sysctl_get_name flags. */ 2898 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0) 2899 2900 /* BPF_FUNC_sk_storage_get flags */ 2901 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0) 2902 2903 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 2904 enum bpf_adj_room_mode { 2905 BPF_ADJ_ROOM_NET, 2906 BPF_ADJ_ROOM_MAC, 2907 }; 2908 2909 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 2910 enum bpf_hdr_start_off { 2911 BPF_HDR_START_MAC, 2912 BPF_HDR_START_NET, 2913 }; 2914 2915 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2916 enum bpf_lwt_encap_mode { 2917 BPF_LWT_ENCAP_SEG6, 2918 BPF_LWT_ENCAP_SEG6_INLINE, 2919 BPF_LWT_ENCAP_IP, 2920 }; 2921 2922 #define __bpf_md_ptr(type, name) \ 2923 union { \ 2924 type name; \ 2925 __u64 :64; \ 2926 } __attribute__((aligned(8))) 2927 2928 /* user accessible mirror of in-kernel sk_buff. 2929 * new fields can only be added to the end of this structure 2930 */ 2931 struct __sk_buff { 2932 __u32 len; 2933 __u32 pkt_type; 2934 __u32 mark; 2935 __u32 queue_mapping; 2936 __u32 protocol; 2937 __u32 vlan_present; 2938 __u32 vlan_tci; 2939 __u32 vlan_proto; 2940 __u32 priority; 2941 __u32 ingress_ifindex; 2942 __u32 ifindex; 2943 __u32 tc_index; 2944 __u32 cb[5]; 2945 __u32 hash; 2946 __u32 tc_classid; 2947 __u32 data; 2948 __u32 data_end; 2949 __u32 napi_id; 2950 2951 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 2952 __u32 family; 2953 __u32 remote_ip4; /* Stored in network byte order */ 2954 __u32 local_ip4; /* Stored in network byte order */ 2955 __u32 remote_ip6[4]; /* Stored in network byte order */ 2956 __u32 local_ip6[4]; /* Stored in network byte order */ 2957 __u32 remote_port; /* Stored in network byte order */ 2958 __u32 local_port; /* stored in host byte order */ 2959 /* ... here. */ 2960 2961 __u32 data_meta; 2962 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 2963 __u64 tstamp; 2964 __u32 wire_len; 2965 __u32 gso_segs; 2966 __bpf_md_ptr(struct bpf_sock *, sk); 2967 }; 2968 2969 struct bpf_tunnel_key { 2970 __u32 tunnel_id; 2971 union { 2972 __u32 remote_ipv4; 2973 __u32 remote_ipv6[4]; 2974 }; 2975 __u8 tunnel_tos; 2976 __u8 tunnel_ttl; 2977 __u16 tunnel_ext; /* Padding, future use. */ 2978 __u32 tunnel_label; 2979 }; 2980 2981 /* user accessible mirror of in-kernel xfrm_state. 2982 * new fields can only be added to the end of this structure 2983 */ 2984 struct bpf_xfrm_state { 2985 __u32 reqid; 2986 __u32 spi; /* Stored in network byte order */ 2987 __u16 family; 2988 __u16 ext; /* Padding, future use. */ 2989 union { 2990 __u32 remote_ipv4; /* Stored in network byte order */ 2991 __u32 remote_ipv6[4]; /* Stored in network byte order */ 2992 }; 2993 }; 2994 2995 /* Generic BPF return codes which all BPF program types may support. 2996 * The values are binary compatible with their TC_ACT_* counter-part to 2997 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 2998 * programs. 2999 * 3000 * XDP is handled seprately, see XDP_*. 3001 */ 3002 enum bpf_ret_code { 3003 BPF_OK = 0, 3004 /* 1 reserved */ 3005 BPF_DROP = 2, 3006 /* 3-6 reserved */ 3007 BPF_REDIRECT = 7, 3008 /* >127 are reserved for prog type specific return codes. 3009 * 3010 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3011 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3012 * changed and should be routed based on its new L3 header. 3013 * (This is an L3 redirect, as opposed to L2 redirect 3014 * represented by BPF_REDIRECT above). 3015 */ 3016 BPF_LWT_REROUTE = 128, 3017 }; 3018 3019 struct bpf_sock { 3020 __u32 bound_dev_if; 3021 __u32 family; 3022 __u32 type; 3023 __u32 protocol; 3024 __u32 mark; 3025 __u32 priority; 3026 /* IP address also allows 1 and 2 bytes access */ 3027 __u32 src_ip4; 3028 __u32 src_ip6[4]; 3029 __u32 src_port; /* host byte order */ 3030 __u32 dst_port; /* network byte order */ 3031 __u32 dst_ip4; 3032 __u32 dst_ip6[4]; 3033 __u32 state; 3034 }; 3035 3036 struct bpf_tcp_sock { 3037 __u32 snd_cwnd; /* Sending congestion window */ 3038 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3039 __u32 rtt_min; 3040 __u32 snd_ssthresh; /* Slow start size threshold */ 3041 __u32 rcv_nxt; /* What we want to receive next */ 3042 __u32 snd_nxt; /* Next sequence we send */ 3043 __u32 snd_una; /* First byte we want an ack for */ 3044 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3045 __u32 ecn_flags; /* ECN status bits. */ 3046 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3047 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3048 __u32 packets_out; /* Packets which are "in flight" */ 3049 __u32 retrans_out; /* Retransmitted packets out */ 3050 __u32 total_retrans; /* Total retransmits for entire connection */ 3051 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3052 * total number of segments in. 3053 */ 3054 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3055 * total number of data segments in. 3056 */ 3057 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3058 * The total number of segments sent. 3059 */ 3060 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3061 * total number of data segments sent. 3062 */ 3063 __u32 lost_out; /* Lost packets */ 3064 __u32 sacked_out; /* SACK'd packets */ 3065 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3066 * sum(delta(rcv_nxt)), or how many bytes 3067 * were acked. 3068 */ 3069 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3070 * sum(delta(snd_una)), or how many bytes 3071 * were acked. 3072 */ 3073 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3074 * total number of DSACK blocks received 3075 */ 3076 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3077 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3078 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3079 }; 3080 3081 struct bpf_sock_tuple { 3082 union { 3083 struct { 3084 __be32 saddr; 3085 __be32 daddr; 3086 __be16 sport; 3087 __be16 dport; 3088 } ipv4; 3089 struct { 3090 __be32 saddr[4]; 3091 __be32 daddr[4]; 3092 __be16 sport; 3093 __be16 dport; 3094 } ipv6; 3095 }; 3096 }; 3097 3098 struct bpf_xdp_sock { 3099 __u32 queue_id; 3100 }; 3101 3102 #define XDP_PACKET_HEADROOM 256 3103 3104 /* User return codes for XDP prog type. 3105 * A valid XDP program must return one of these defined values. All other 3106 * return codes are reserved for future use. Unknown return codes will 3107 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3108 */ 3109 enum xdp_action { 3110 XDP_ABORTED = 0, 3111 XDP_DROP, 3112 XDP_PASS, 3113 XDP_TX, 3114 XDP_REDIRECT, 3115 }; 3116 3117 /* user accessible metadata for XDP packet hook 3118 * new fields must be added to the end of this structure 3119 */ 3120 struct xdp_md { 3121 __u32 data; 3122 __u32 data_end; 3123 __u32 data_meta; 3124 /* Below access go through struct xdp_rxq_info */ 3125 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3126 __u32 rx_queue_index; /* rxq->queue_index */ 3127 }; 3128 3129 enum sk_action { 3130 SK_DROP = 0, 3131 SK_PASS, 3132 }; 3133 3134 /* user accessible metadata for SK_MSG packet hook, new fields must 3135 * be added to the end of this structure 3136 */ 3137 struct sk_msg_md { 3138 __bpf_md_ptr(void *, data); 3139 __bpf_md_ptr(void *, data_end); 3140 3141 __u32 family; 3142 __u32 remote_ip4; /* Stored in network byte order */ 3143 __u32 local_ip4; /* Stored in network byte order */ 3144 __u32 remote_ip6[4]; /* Stored in network byte order */ 3145 __u32 local_ip6[4]; /* Stored in network byte order */ 3146 __u32 remote_port; /* Stored in network byte order */ 3147 __u32 local_port; /* stored in host byte order */ 3148 __u32 size; /* Total size of sk_msg */ 3149 }; 3150 3151 struct sk_reuseport_md { 3152 /* 3153 * Start of directly accessible data. It begins from 3154 * the tcp/udp header. 3155 */ 3156 __bpf_md_ptr(void *, data); 3157 /* End of directly accessible data */ 3158 __bpf_md_ptr(void *, data_end); 3159 /* 3160 * Total length of packet (starting from the tcp/udp header). 3161 * Note that the directly accessible bytes (data_end - data) 3162 * could be less than this "len". Those bytes could be 3163 * indirectly read by a helper "bpf_skb_load_bytes()". 3164 */ 3165 __u32 len; 3166 /* 3167 * Eth protocol in the mac header (network byte order). e.g. 3168 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3169 */ 3170 __u32 eth_protocol; 3171 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3172 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3173 __u32 hash; /* A hash of the packet 4 tuples */ 3174 }; 3175 3176 #define BPF_TAG_SIZE 8 3177 3178 struct bpf_prog_info { 3179 __u32 type; 3180 __u32 id; 3181 __u8 tag[BPF_TAG_SIZE]; 3182 __u32 jited_prog_len; 3183 __u32 xlated_prog_len; 3184 __aligned_u64 jited_prog_insns; 3185 __aligned_u64 xlated_prog_insns; 3186 __u64 load_time; /* ns since boottime */ 3187 __u32 created_by_uid; 3188 __u32 nr_map_ids; 3189 __aligned_u64 map_ids; 3190 char name[BPF_OBJ_NAME_LEN]; 3191 __u32 ifindex; 3192 __u32 gpl_compatible:1; 3193 __u32 :31; /* alignment pad */ 3194 __u64 netns_dev; 3195 __u64 netns_ino; 3196 __u32 nr_jited_ksyms; 3197 __u32 nr_jited_func_lens; 3198 __aligned_u64 jited_ksyms; 3199 __aligned_u64 jited_func_lens; 3200 __u32 btf_id; 3201 __u32 func_info_rec_size; 3202 __aligned_u64 func_info; 3203 __u32 nr_func_info; 3204 __u32 nr_line_info; 3205 __aligned_u64 line_info; 3206 __aligned_u64 jited_line_info; 3207 __u32 nr_jited_line_info; 3208 __u32 line_info_rec_size; 3209 __u32 jited_line_info_rec_size; 3210 __u32 nr_prog_tags; 3211 __aligned_u64 prog_tags; 3212 __u64 run_time_ns; 3213 __u64 run_cnt; 3214 } __attribute__((aligned(8))); 3215 3216 struct bpf_map_info { 3217 __u32 type; 3218 __u32 id; 3219 __u32 key_size; 3220 __u32 value_size; 3221 __u32 max_entries; 3222 __u32 map_flags; 3223 char name[BPF_OBJ_NAME_LEN]; 3224 __u32 ifindex; 3225 __u32 :32; 3226 __u64 netns_dev; 3227 __u64 netns_ino; 3228 __u32 btf_id; 3229 __u32 btf_key_type_id; 3230 __u32 btf_value_type_id; 3231 } __attribute__((aligned(8))); 3232 3233 struct bpf_btf_info { 3234 __aligned_u64 btf; 3235 __u32 btf_size; 3236 __u32 id; 3237 } __attribute__((aligned(8))); 3238 3239 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3240 * by user and intended to be used by socket (e.g. to bind to, depends on 3241 * attach attach type). 3242 */ 3243 struct bpf_sock_addr { 3244 __u32 user_family; /* Allows 4-byte read, but no write. */ 3245 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3246 * Stored in network byte order. 3247 */ 3248 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3249 * Stored in network byte order. 3250 */ 3251 __u32 user_port; /* Allows 4-byte read and write. 3252 * Stored in network byte order 3253 */ 3254 __u32 family; /* Allows 4-byte read, but no write */ 3255 __u32 type; /* Allows 4-byte read, but no write */ 3256 __u32 protocol; /* Allows 4-byte read, but no write */ 3257 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3258 * Stored in network byte order. 3259 */ 3260 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3261 * Stored in network byte order. 3262 */ 3263 __bpf_md_ptr(struct bpf_sock *, sk); 3264 }; 3265 3266 /* User bpf_sock_ops struct to access socket values and specify request ops 3267 * and their replies. 3268 * Some of this fields are in network (bigendian) byte order and may need 3269 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3270 * New fields can only be added at the end of this structure 3271 */ 3272 struct bpf_sock_ops { 3273 __u32 op; 3274 union { 3275 __u32 args[4]; /* Optionally passed to bpf program */ 3276 __u32 reply; /* Returned by bpf program */ 3277 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3278 }; 3279 __u32 family; 3280 __u32 remote_ip4; /* Stored in network byte order */ 3281 __u32 local_ip4; /* Stored in network byte order */ 3282 __u32 remote_ip6[4]; /* Stored in network byte order */ 3283 __u32 local_ip6[4]; /* Stored in network byte order */ 3284 __u32 remote_port; /* Stored in network byte order */ 3285 __u32 local_port; /* stored in host byte order */ 3286 __u32 is_fullsock; /* Some TCP fields are only valid if 3287 * there is a full socket. If not, the 3288 * fields read as zero. 3289 */ 3290 __u32 snd_cwnd; 3291 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3292 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3293 __u32 state; 3294 __u32 rtt_min; 3295 __u32 snd_ssthresh; 3296 __u32 rcv_nxt; 3297 __u32 snd_nxt; 3298 __u32 snd_una; 3299 __u32 mss_cache; 3300 __u32 ecn_flags; 3301 __u32 rate_delivered; 3302 __u32 rate_interval_us; 3303 __u32 packets_out; 3304 __u32 retrans_out; 3305 __u32 total_retrans; 3306 __u32 segs_in; 3307 __u32 data_segs_in; 3308 __u32 segs_out; 3309 __u32 data_segs_out; 3310 __u32 lost_out; 3311 __u32 sacked_out; 3312 __u32 sk_txhash; 3313 __u64 bytes_received; 3314 __u64 bytes_acked; 3315 __bpf_md_ptr(struct bpf_sock *, sk); 3316 }; 3317 3318 /* Definitions for bpf_sock_ops_cb_flags */ 3319 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3320 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3321 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3322 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3) 3323 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently 3324 * supported cb flags 3325 */ 3326 3327 /* List of known BPF sock_ops operators. 3328 * New entries can only be added at the end 3329 */ 3330 enum { 3331 BPF_SOCK_OPS_VOID, 3332 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3333 * -1 if default value should be used 3334 */ 3335 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3336 * window (in packets) or -1 if default 3337 * value should be used 3338 */ 3339 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3340 * active connection is initialized 3341 */ 3342 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3343 * active connection is 3344 * established 3345 */ 3346 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3347 * passive connection is 3348 * established 3349 */ 3350 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3351 * needs ECN 3352 */ 3353 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3354 * based on the path and may be 3355 * dependent on the congestion control 3356 * algorithm. In general it indicates 3357 * a congestion threshold. RTTs above 3358 * this indicate congestion 3359 */ 3360 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3361 * Arg1: value of icsk_retransmits 3362 * Arg2: value of icsk_rto 3363 * Arg3: whether RTO has expired 3364 */ 3365 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3366 * Arg1: sequence number of 1st byte 3367 * Arg2: # segments 3368 * Arg3: return value of 3369 * tcp_transmit_skb (0 => success) 3370 */ 3371 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3372 * Arg1: old_state 3373 * Arg2: new_state 3374 */ 3375 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3376 * socket transition to LISTEN state. 3377 */ 3378 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 3379 */ 3380 }; 3381 3382 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3383 * changes between the TCP and BPF versions. Ideally this should never happen. 3384 * If it does, we need to add code to convert them before calling 3385 * the BPF sock_ops function. 3386 */ 3387 enum { 3388 BPF_TCP_ESTABLISHED = 1, 3389 BPF_TCP_SYN_SENT, 3390 BPF_TCP_SYN_RECV, 3391 BPF_TCP_FIN_WAIT1, 3392 BPF_TCP_FIN_WAIT2, 3393 BPF_TCP_TIME_WAIT, 3394 BPF_TCP_CLOSE, 3395 BPF_TCP_CLOSE_WAIT, 3396 BPF_TCP_LAST_ACK, 3397 BPF_TCP_LISTEN, 3398 BPF_TCP_CLOSING, /* Now a valid state */ 3399 BPF_TCP_NEW_SYN_RECV, 3400 3401 BPF_TCP_MAX_STATES /* Leave at the end! */ 3402 }; 3403 3404 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3405 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3406 3407 struct bpf_perf_event_value { 3408 __u64 counter; 3409 __u64 enabled; 3410 __u64 running; 3411 }; 3412 3413 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3414 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3415 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3416 3417 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3418 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3419 3420 struct bpf_cgroup_dev_ctx { 3421 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3422 __u32 access_type; 3423 __u32 major; 3424 __u32 minor; 3425 }; 3426 3427 struct bpf_raw_tracepoint_args { 3428 __u64 args[0]; 3429 }; 3430 3431 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3432 * OUTPUT: Do lookup from egress perspective; default is ingress 3433 */ 3434 #define BPF_FIB_LOOKUP_DIRECT (1U << 0) 3435 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1) 3436 3437 enum { 3438 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3439 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3440 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3441 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3442 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3443 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3444 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3445 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3446 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3447 }; 3448 3449 struct bpf_fib_lookup { 3450 /* input: network family for lookup (AF_INET, AF_INET6) 3451 * output: network family of egress nexthop 3452 */ 3453 __u8 family; 3454 3455 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3456 __u8 l4_protocol; 3457 __be16 sport; 3458 __be16 dport; 3459 3460 /* total length of packet from network header - used for MTU check */ 3461 __u16 tot_len; 3462 3463 /* input: L3 device index for lookup 3464 * output: device index from FIB lookup 3465 */ 3466 __u32 ifindex; 3467 3468 union { 3469 /* inputs to lookup */ 3470 __u8 tos; /* AF_INET */ 3471 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3472 3473 /* output: metric of fib result (IPv4/IPv6 only) */ 3474 __u32 rt_metric; 3475 }; 3476 3477 union { 3478 __be32 ipv4_src; 3479 __u32 ipv6_src[4]; /* in6_addr; network order */ 3480 }; 3481 3482 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3483 * network header. output: bpf_fib_lookup sets to gateway address 3484 * if FIB lookup returns gateway route 3485 */ 3486 union { 3487 __be32 ipv4_dst; 3488 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3489 }; 3490 3491 /* output */ 3492 __be16 h_vlan_proto; 3493 __be16 h_vlan_TCI; 3494 __u8 smac[6]; /* ETH_ALEN */ 3495 __u8 dmac[6]; /* ETH_ALEN */ 3496 }; 3497 3498 enum bpf_task_fd_type { 3499 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3500 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3501 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3502 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3503 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3504 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3505 }; 3506 3507 struct bpf_flow_keys { 3508 __u16 nhoff; 3509 __u16 thoff; 3510 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3511 __u8 is_frag; 3512 __u8 is_first_frag; 3513 __u8 is_encap; 3514 __u8 ip_proto; 3515 __be16 n_proto; 3516 __be16 sport; 3517 __be16 dport; 3518 union { 3519 struct { 3520 __be32 ipv4_src; 3521 __be32 ipv4_dst; 3522 }; 3523 struct { 3524 __u32 ipv6_src[4]; /* in6_addr; network order */ 3525 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3526 }; 3527 }; 3528 }; 3529 3530 struct bpf_func_info { 3531 __u32 insn_off; 3532 __u32 type_id; 3533 }; 3534 3535 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3536 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3537 3538 struct bpf_line_info { 3539 __u32 insn_off; 3540 __u32 file_name_off; 3541 __u32 line_off; 3542 __u32 line_col; 3543 }; 3544 3545 struct bpf_spin_lock { 3546 __u32 val; 3547 }; 3548 3549 struct bpf_sysctl { 3550 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3551 * Allows 1,2,4-byte read, but no write. 3552 */ 3553 __u32 file_pos; /* Sysctl file position to read from, write to. 3554 * Allows 1,2,4-byte read an 4-byte write. 3555 */ 3556 }; 3557 3558 struct bpf_sockopt { 3559 __bpf_md_ptr(struct bpf_sock *, sk); 3560 __bpf_md_ptr(void *, optval); 3561 __bpf_md_ptr(void *, optval_end); 3562 3563 __s32 level; 3564 __s32 optname; 3565 __s32 optlen; 3566 __s32 retval; 3567 }; 3568 3569 #endif /* _UAPI__LINUX_BPF_H__ */ 3570