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 BPF_BTF_GET_NEXT_ID, 110 }; 111 112 enum bpf_map_type { 113 BPF_MAP_TYPE_UNSPEC, 114 BPF_MAP_TYPE_HASH, 115 BPF_MAP_TYPE_ARRAY, 116 BPF_MAP_TYPE_PROG_ARRAY, 117 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 118 BPF_MAP_TYPE_PERCPU_HASH, 119 BPF_MAP_TYPE_PERCPU_ARRAY, 120 BPF_MAP_TYPE_STACK_TRACE, 121 BPF_MAP_TYPE_CGROUP_ARRAY, 122 BPF_MAP_TYPE_LRU_HASH, 123 BPF_MAP_TYPE_LRU_PERCPU_HASH, 124 BPF_MAP_TYPE_LPM_TRIE, 125 BPF_MAP_TYPE_ARRAY_OF_MAPS, 126 BPF_MAP_TYPE_HASH_OF_MAPS, 127 BPF_MAP_TYPE_DEVMAP, 128 BPF_MAP_TYPE_SOCKMAP, 129 BPF_MAP_TYPE_CPUMAP, 130 BPF_MAP_TYPE_XSKMAP, 131 BPF_MAP_TYPE_SOCKHASH, 132 BPF_MAP_TYPE_CGROUP_STORAGE, 133 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 134 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 135 BPF_MAP_TYPE_QUEUE, 136 BPF_MAP_TYPE_STACK, 137 BPF_MAP_TYPE_SK_STORAGE, 138 BPF_MAP_TYPE_DEVMAP_HASH, 139 }; 140 141 /* Note that tracing related programs such as 142 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 143 * are not subject to a stable API since kernel internal data 144 * structures can change from release to release and may 145 * therefore break existing tracing BPF programs. Tracing BPF 146 * programs correspond to /a/ specific kernel which is to be 147 * analyzed, and not /a/ specific kernel /and/ all future ones. 148 */ 149 enum bpf_prog_type { 150 BPF_PROG_TYPE_UNSPEC, 151 BPF_PROG_TYPE_SOCKET_FILTER, 152 BPF_PROG_TYPE_KPROBE, 153 BPF_PROG_TYPE_SCHED_CLS, 154 BPF_PROG_TYPE_SCHED_ACT, 155 BPF_PROG_TYPE_TRACEPOINT, 156 BPF_PROG_TYPE_XDP, 157 BPF_PROG_TYPE_PERF_EVENT, 158 BPF_PROG_TYPE_CGROUP_SKB, 159 BPF_PROG_TYPE_CGROUP_SOCK, 160 BPF_PROG_TYPE_LWT_IN, 161 BPF_PROG_TYPE_LWT_OUT, 162 BPF_PROG_TYPE_LWT_XMIT, 163 BPF_PROG_TYPE_SOCK_OPS, 164 BPF_PROG_TYPE_SK_SKB, 165 BPF_PROG_TYPE_CGROUP_DEVICE, 166 BPF_PROG_TYPE_SK_MSG, 167 BPF_PROG_TYPE_RAW_TRACEPOINT, 168 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 169 BPF_PROG_TYPE_LWT_SEG6LOCAL, 170 BPF_PROG_TYPE_LIRC_MODE2, 171 BPF_PROG_TYPE_SK_REUSEPORT, 172 BPF_PROG_TYPE_FLOW_DISSECTOR, 173 BPF_PROG_TYPE_CGROUP_SYSCTL, 174 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 175 BPF_PROG_TYPE_CGROUP_SOCKOPT, 176 }; 177 178 enum bpf_attach_type { 179 BPF_CGROUP_INET_INGRESS, 180 BPF_CGROUP_INET_EGRESS, 181 BPF_CGROUP_INET_SOCK_CREATE, 182 BPF_CGROUP_SOCK_OPS, 183 BPF_SK_SKB_STREAM_PARSER, 184 BPF_SK_SKB_STREAM_VERDICT, 185 BPF_CGROUP_DEVICE, 186 BPF_SK_MSG_VERDICT, 187 BPF_CGROUP_INET4_BIND, 188 BPF_CGROUP_INET6_BIND, 189 BPF_CGROUP_INET4_CONNECT, 190 BPF_CGROUP_INET6_CONNECT, 191 BPF_CGROUP_INET4_POST_BIND, 192 BPF_CGROUP_INET6_POST_BIND, 193 BPF_CGROUP_UDP4_SENDMSG, 194 BPF_CGROUP_UDP6_SENDMSG, 195 BPF_LIRC_MODE2, 196 BPF_FLOW_DISSECTOR, 197 BPF_CGROUP_SYSCTL, 198 BPF_CGROUP_UDP4_RECVMSG, 199 BPF_CGROUP_UDP6_RECVMSG, 200 BPF_CGROUP_GETSOCKOPT, 201 BPF_CGROUP_SETSOCKOPT, 202 __MAX_BPF_ATTACH_TYPE 203 }; 204 205 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 206 207 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 208 * 209 * NONE(default): No further bpf programs allowed in the subtree. 210 * 211 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 212 * the program in this cgroup yields to sub-cgroup program. 213 * 214 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 215 * that cgroup program gets run in addition to the program in this cgroup. 216 * 217 * Only one program is allowed to be attached to a cgroup with 218 * NONE or BPF_F_ALLOW_OVERRIDE flag. 219 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 220 * release old program and attach the new one. Attach flags has to match. 221 * 222 * Multiple programs are allowed to be attached to a cgroup with 223 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 224 * (those that were attached first, run first) 225 * The programs of sub-cgroup are executed first, then programs of 226 * this cgroup and then programs of parent cgroup. 227 * When children program makes decision (like picking TCP CA or sock bind) 228 * parent program has a chance to override it. 229 * 230 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 231 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 232 * Ex1: 233 * cgrp1 (MULTI progs A, B) -> 234 * cgrp2 (OVERRIDE prog C) -> 235 * cgrp3 (MULTI prog D) -> 236 * cgrp4 (OVERRIDE prog E) -> 237 * cgrp5 (NONE prog F) 238 * the event in cgrp5 triggers execution of F,D,A,B in that order. 239 * if prog F is detached, the execution is E,D,A,B 240 * if prog F and D are detached, the execution is E,A,B 241 * if prog F, E and D are detached, the execution is C,A,B 242 * 243 * All eligible programs are executed regardless of return code from 244 * earlier programs. 245 */ 246 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 247 #define BPF_F_ALLOW_MULTI (1U << 1) 248 249 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 250 * verifier will perform strict alignment checking as if the kernel 251 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 252 * and NET_IP_ALIGN defined to 2. 253 */ 254 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 255 256 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 257 * verifier will allow any alignment whatsoever. On platforms 258 * with strict alignment requirements for loads ands stores (such 259 * as sparc and mips) the verifier validates that all loads and 260 * stores provably follow this requirement. This flag turns that 261 * checking and enforcement off. 262 * 263 * It is mostly used for testing when we want to validate the 264 * context and memory access aspects of the verifier, but because 265 * of an unaligned access the alignment check would trigger before 266 * the one we are interested in. 267 */ 268 #define BPF_F_ANY_ALIGNMENT (1U << 1) 269 270 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 271 * Verifier does sub-register def/use analysis and identifies instructions whose 272 * def only matters for low 32-bit, high 32-bit is never referenced later 273 * through implicit zero extension. Therefore verifier notifies JIT back-ends 274 * that it is safe to ignore clearing high 32-bit for these instructions. This 275 * saves some back-ends a lot of code-gen. However such optimization is not 276 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 277 * hence hasn't used verifier's analysis result. But, we really want to have a 278 * way to be able to verify the correctness of the described optimization on 279 * x86_64 on which testsuites are frequently exercised. 280 * 281 * So, this flag is introduced. Once it is set, verifier will randomize high 282 * 32-bit for those instructions who has been identified as safe to ignore them. 283 * Then, if verifier is not doing correct analysis, such randomization will 284 * regress tests to expose bugs. 285 */ 286 #define BPF_F_TEST_RND_HI32 (1U << 2) 287 288 /* The verifier internal test flag. Behavior is undefined */ 289 #define BPF_F_TEST_STATE_FREQ (1U << 3) 290 291 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 292 * two extensions: 293 * 294 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 295 * insn[0].imm: map fd map fd 296 * insn[1].imm: 0 offset into value 297 * insn[0].off: 0 0 298 * insn[1].off: 0 0 299 * ldimm64 rewrite: address of map address of map[0]+offset 300 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 301 */ 302 #define BPF_PSEUDO_MAP_FD 1 303 #define BPF_PSEUDO_MAP_VALUE 2 304 305 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 306 * offset to another bpf function 307 */ 308 #define BPF_PSEUDO_CALL 1 309 310 /* flags for BPF_MAP_UPDATE_ELEM command */ 311 #define BPF_ANY 0 /* create new element or update existing */ 312 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 313 #define BPF_EXIST 2 /* update existing element */ 314 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 315 316 /* flags for BPF_MAP_CREATE command */ 317 #define BPF_F_NO_PREALLOC (1U << 0) 318 /* Instead of having one common LRU list in the 319 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 320 * which can scale and perform better. 321 * Note, the LRU nodes (including free nodes) cannot be moved 322 * across different LRU lists. 323 */ 324 #define BPF_F_NO_COMMON_LRU (1U << 1) 325 /* Specify numa node during map creation */ 326 #define BPF_F_NUMA_NODE (1U << 2) 327 328 #define BPF_OBJ_NAME_LEN 16U 329 330 /* Flags for accessing BPF object from syscall side. */ 331 #define BPF_F_RDONLY (1U << 3) 332 #define BPF_F_WRONLY (1U << 4) 333 334 /* Flag for stack_map, store build_id+offset instead of pointer */ 335 #define BPF_F_STACK_BUILD_ID (1U << 5) 336 337 /* Zero-initialize hash function seed. This should only be used for testing. */ 338 #define BPF_F_ZERO_SEED (1U << 6) 339 340 /* Flags for accessing BPF object from program side. */ 341 #define BPF_F_RDONLY_PROG (1U << 7) 342 #define BPF_F_WRONLY_PROG (1U << 8) 343 344 /* Clone map from listener for newly accepted socket */ 345 #define BPF_F_CLONE (1U << 9) 346 347 /* flags for BPF_PROG_QUERY */ 348 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 349 350 enum bpf_stack_build_id_status { 351 /* user space need an empty entry to identify end of a trace */ 352 BPF_STACK_BUILD_ID_EMPTY = 0, 353 /* with valid build_id and offset */ 354 BPF_STACK_BUILD_ID_VALID = 1, 355 /* couldn't get build_id, fallback to ip */ 356 BPF_STACK_BUILD_ID_IP = 2, 357 }; 358 359 #define BPF_BUILD_ID_SIZE 20 360 struct bpf_stack_build_id { 361 __s32 status; 362 unsigned char build_id[BPF_BUILD_ID_SIZE]; 363 union { 364 __u64 offset; 365 __u64 ip; 366 }; 367 }; 368 369 union bpf_attr { 370 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 371 __u32 map_type; /* one of enum bpf_map_type */ 372 __u32 key_size; /* size of key in bytes */ 373 __u32 value_size; /* size of value in bytes */ 374 __u32 max_entries; /* max number of entries in a map */ 375 __u32 map_flags; /* BPF_MAP_CREATE related 376 * flags defined above. 377 */ 378 __u32 inner_map_fd; /* fd pointing to the inner map */ 379 __u32 numa_node; /* numa node (effective only if 380 * BPF_F_NUMA_NODE is set). 381 */ 382 char map_name[BPF_OBJ_NAME_LEN]; 383 __u32 map_ifindex; /* ifindex of netdev to create on */ 384 __u32 btf_fd; /* fd pointing to a BTF type data */ 385 __u32 btf_key_type_id; /* BTF type_id of the key */ 386 __u32 btf_value_type_id; /* BTF type_id of the value */ 387 }; 388 389 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 390 __u32 map_fd; 391 __aligned_u64 key; 392 union { 393 __aligned_u64 value; 394 __aligned_u64 next_key; 395 }; 396 __u64 flags; 397 }; 398 399 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 400 __u32 prog_type; /* one of enum bpf_prog_type */ 401 __u32 insn_cnt; 402 __aligned_u64 insns; 403 __aligned_u64 license; 404 __u32 log_level; /* verbosity level of verifier */ 405 __u32 log_size; /* size of user buffer */ 406 __aligned_u64 log_buf; /* user supplied buffer */ 407 __u32 kern_version; /* not used */ 408 __u32 prog_flags; 409 char prog_name[BPF_OBJ_NAME_LEN]; 410 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 411 /* For some prog types expected attach type must be known at 412 * load time to verify attach type specific parts of prog 413 * (context accesses, allowed helpers, etc). 414 */ 415 __u32 expected_attach_type; 416 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 417 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 418 __aligned_u64 func_info; /* func info */ 419 __u32 func_info_cnt; /* number of bpf_func_info records */ 420 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 421 __aligned_u64 line_info; /* line info */ 422 __u32 line_info_cnt; /* number of bpf_line_info records */ 423 }; 424 425 struct { /* anonymous struct used by BPF_OBJ_* commands */ 426 __aligned_u64 pathname; 427 __u32 bpf_fd; 428 __u32 file_flags; 429 }; 430 431 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 432 __u32 target_fd; /* container object to attach to */ 433 __u32 attach_bpf_fd; /* eBPF program to attach */ 434 __u32 attach_type; 435 __u32 attach_flags; 436 }; 437 438 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 439 __u32 prog_fd; 440 __u32 retval; 441 __u32 data_size_in; /* input: len of data_in */ 442 __u32 data_size_out; /* input/output: len of data_out 443 * returns ENOSPC if data_out 444 * is too small. 445 */ 446 __aligned_u64 data_in; 447 __aligned_u64 data_out; 448 __u32 repeat; 449 __u32 duration; 450 __u32 ctx_size_in; /* input: len of ctx_in */ 451 __u32 ctx_size_out; /* input/output: len of ctx_out 452 * returns ENOSPC if ctx_out 453 * is too small. 454 */ 455 __aligned_u64 ctx_in; 456 __aligned_u64 ctx_out; 457 } test; 458 459 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 460 union { 461 __u32 start_id; 462 __u32 prog_id; 463 __u32 map_id; 464 __u32 btf_id; 465 }; 466 __u32 next_id; 467 __u32 open_flags; 468 }; 469 470 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 471 __u32 bpf_fd; 472 __u32 info_len; 473 __aligned_u64 info; 474 } info; 475 476 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 477 __u32 target_fd; /* container object to query */ 478 __u32 attach_type; 479 __u32 query_flags; 480 __u32 attach_flags; 481 __aligned_u64 prog_ids; 482 __u32 prog_cnt; 483 } query; 484 485 struct { 486 __u64 name; 487 __u32 prog_fd; 488 } raw_tracepoint; 489 490 struct { /* anonymous struct for BPF_BTF_LOAD */ 491 __aligned_u64 btf; 492 __aligned_u64 btf_log_buf; 493 __u32 btf_size; 494 __u32 btf_log_size; 495 __u32 btf_log_level; 496 }; 497 498 struct { 499 __u32 pid; /* input: pid */ 500 __u32 fd; /* input: fd */ 501 __u32 flags; /* input: flags */ 502 __u32 buf_len; /* input/output: buf len */ 503 __aligned_u64 buf; /* input/output: 504 * tp_name for tracepoint 505 * symbol for kprobe 506 * filename for uprobe 507 */ 508 __u32 prog_id; /* output: prod_id */ 509 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 510 __u64 probe_offset; /* output: probe_offset */ 511 __u64 probe_addr; /* output: probe_addr */ 512 } task_fd_query; 513 } __attribute__((aligned(8))); 514 515 /* The description below is an attempt at providing documentation to eBPF 516 * developers about the multiple available eBPF helper functions. It can be 517 * parsed and used to produce a manual page. The workflow is the following, 518 * and requires the rst2man utility: 519 * 520 * $ ./scripts/bpf_helpers_doc.py \ 521 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 522 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 523 * $ man /tmp/bpf-helpers.7 524 * 525 * Note that in order to produce this external documentation, some RST 526 * formatting is used in the descriptions to get "bold" and "italics" in 527 * manual pages. Also note that the few trailing white spaces are 528 * intentional, removing them would break paragraphs for rst2man. 529 * 530 * Start of BPF helper function descriptions: 531 * 532 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 533 * Description 534 * Perform a lookup in *map* for an entry associated to *key*. 535 * Return 536 * Map value associated to *key*, or **NULL** if no entry was 537 * found. 538 * 539 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 540 * Description 541 * Add or update the value of the entry associated to *key* in 542 * *map* with *value*. *flags* is one of: 543 * 544 * **BPF_NOEXIST** 545 * The entry for *key* must not exist in the map. 546 * **BPF_EXIST** 547 * The entry for *key* must already exist in the map. 548 * **BPF_ANY** 549 * No condition on the existence of the entry for *key*. 550 * 551 * Flag value **BPF_NOEXIST** cannot be used for maps of types 552 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 553 * elements always exist), the helper would return an error. 554 * Return 555 * 0 on success, or a negative error in case of failure. 556 * 557 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 558 * Description 559 * Delete entry with *key* from *map*. 560 * Return 561 * 0 on success, or a negative error in case of failure. 562 * 563 * int bpf_probe_read(void *dst, u32 size, const void *src) 564 * Description 565 * For tracing programs, safely attempt to read *size* bytes from 566 * address *src* and store the data in *dst*. 567 * Return 568 * 0 on success, or a negative error in case of failure. 569 * 570 * u64 bpf_ktime_get_ns(void) 571 * Description 572 * Return the time elapsed since system boot, in nanoseconds. 573 * Does not include time the system was suspended. 574 * See: clock_gettime(CLOCK_MONOTONIC) 575 * Return 576 * Current *ktime*. 577 * 578 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 579 * Description 580 * This helper is a "printk()-like" facility for debugging. It 581 * prints a message defined by format *fmt* (of size *fmt_size*) 582 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 583 * available. It can take up to three additional **u64** 584 * arguments (as an eBPF helpers, the total number of arguments is 585 * limited to five). 586 * 587 * Each time the helper is called, it appends a line to the trace. 588 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 589 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 590 * The format of the trace is customizable, and the exact output 591 * one will get depends on the options set in 592 * *\/sys/kernel/debug/tracing/trace_options* (see also the 593 * *README* file under the same directory). However, it usually 594 * defaults to something like: 595 * 596 * :: 597 * 598 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 599 * 600 * In the above: 601 * 602 * * ``telnet`` is the name of the current task. 603 * * ``470`` is the PID of the current task. 604 * * ``001`` is the CPU number on which the task is 605 * running. 606 * * In ``.N..``, each character refers to a set of 607 * options (whether irqs are enabled, scheduling 608 * options, whether hard/softirqs are running, level of 609 * preempt_disabled respectively). **N** means that 610 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 611 * are set. 612 * * ``419421.045894`` is a timestamp. 613 * * ``0x00000001`` is a fake value used by BPF for the 614 * instruction pointer register. 615 * * ``<formatted msg>`` is the message formatted with 616 * *fmt*. 617 * 618 * The conversion specifiers supported by *fmt* are similar, but 619 * more limited than for printk(). They are **%d**, **%i**, 620 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 621 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 622 * of field, padding with zeroes, etc.) is available, and the 623 * helper will return **-EINVAL** (but print nothing) if it 624 * encounters an unknown specifier. 625 * 626 * Also, note that **bpf_trace_printk**\ () is slow, and should 627 * only be used for debugging purposes. For this reason, a notice 628 * bloc (spanning several lines) is printed to kernel logs and 629 * states that the helper should not be used "for production use" 630 * the first time this helper is used (or more precisely, when 631 * **trace_printk**\ () buffers are allocated). For passing values 632 * to user space, perf events should be preferred. 633 * Return 634 * The number of bytes written to the buffer, or a negative error 635 * in case of failure. 636 * 637 * u32 bpf_get_prandom_u32(void) 638 * Description 639 * Get a pseudo-random number. 640 * 641 * From a security point of view, this helper uses its own 642 * pseudo-random internal state, and cannot be used to infer the 643 * seed of other random functions in the kernel. However, it is 644 * essential to note that the generator used by the helper is not 645 * cryptographically secure. 646 * Return 647 * A random 32-bit unsigned value. 648 * 649 * u32 bpf_get_smp_processor_id(void) 650 * Description 651 * Get the SMP (symmetric multiprocessing) processor id. Note that 652 * all programs run with preemption disabled, which means that the 653 * SMP processor id is stable during all the execution of the 654 * program. 655 * Return 656 * The SMP id of the processor running the program. 657 * 658 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 659 * Description 660 * Store *len* bytes from address *from* into the packet 661 * associated to *skb*, at *offset*. *flags* are a combination of 662 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 663 * checksum for the packet after storing the bytes) and 664 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 665 * **->swhash** and *skb*\ **->l4hash** to 0). 666 * 667 * A call to this helper is susceptible to change the underlying 668 * packet buffer. Therefore, at load time, all checks on pointers 669 * previously done by the verifier are invalidated and must be 670 * performed again, if the helper is used in combination with 671 * direct packet access. 672 * Return 673 * 0 on success, or a negative error in case of failure. 674 * 675 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 676 * Description 677 * Recompute the layer 3 (e.g. IP) checksum for the packet 678 * associated to *skb*. Computation is incremental, so the helper 679 * must know the former value of the header field that was 680 * modified (*from*), the new value of this field (*to*), and the 681 * number of bytes (2 or 4) for this field, stored in *size*. 682 * Alternatively, it is possible to store the difference between 683 * the previous and the new values of the header field in *to*, by 684 * setting *from* and *size* to 0. For both methods, *offset* 685 * indicates the location of the IP checksum within the packet. 686 * 687 * This helper works in combination with **bpf_csum_diff**\ (), 688 * which does not update the checksum in-place, but offers more 689 * flexibility and can handle sizes larger than 2 or 4 for the 690 * checksum to update. 691 * 692 * A call to this helper is susceptible to change the underlying 693 * packet buffer. Therefore, at load time, all checks on pointers 694 * previously done by the verifier are invalidated and must be 695 * performed again, if the helper is used in combination with 696 * direct packet access. 697 * Return 698 * 0 on success, or a negative error in case of failure. 699 * 700 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 701 * Description 702 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 703 * packet associated to *skb*. Computation is incremental, so the 704 * helper must know the former value of the header field that was 705 * modified (*from*), the new value of this field (*to*), and the 706 * number of bytes (2 or 4) for this field, stored on the lowest 707 * four bits of *flags*. Alternatively, it is possible to store 708 * the difference between the previous and the new values of the 709 * header field in *to*, by setting *from* and the four lowest 710 * bits of *flags* to 0. For both methods, *offset* indicates the 711 * location of the IP checksum within the packet. In addition to 712 * the size of the field, *flags* can be added (bitwise OR) actual 713 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 714 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 715 * for updates resulting in a null checksum the value is set to 716 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 717 * the checksum is to be computed against a pseudo-header. 718 * 719 * This helper works in combination with **bpf_csum_diff**\ (), 720 * which does not update the checksum in-place, but offers more 721 * flexibility and can handle sizes larger than 2 or 4 for the 722 * checksum to update. 723 * 724 * A call to this helper is susceptible to change the underlying 725 * packet buffer. Therefore, at load time, all checks on pointers 726 * previously done by the verifier are invalidated and must be 727 * performed again, if the helper is used in combination with 728 * direct packet access. 729 * Return 730 * 0 on success, or a negative error in case of failure. 731 * 732 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 733 * Description 734 * This special helper is used to trigger a "tail call", or in 735 * other words, to jump into another eBPF program. The same stack 736 * frame is used (but values on stack and in registers for the 737 * caller are not accessible to the callee). This mechanism allows 738 * for program chaining, either for raising the maximum number of 739 * available eBPF instructions, or to execute given programs in 740 * conditional blocks. For security reasons, there is an upper 741 * limit to the number of successive tail calls that can be 742 * performed. 743 * 744 * Upon call of this helper, the program attempts to jump into a 745 * program referenced at index *index* in *prog_array_map*, a 746 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 747 * *ctx*, a pointer to the context. 748 * 749 * If the call succeeds, the kernel immediately runs the first 750 * instruction of the new program. This is not a function call, 751 * and it never returns to the previous program. If the call 752 * fails, then the helper has no effect, and the caller continues 753 * to run its subsequent instructions. A call can fail if the 754 * destination program for the jump does not exist (i.e. *index* 755 * is superior to the number of entries in *prog_array_map*), or 756 * if the maximum number of tail calls has been reached for this 757 * chain of programs. This limit is defined in the kernel by the 758 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 759 * which is currently set to 32. 760 * Return 761 * 0 on success, or a negative error in case of failure. 762 * 763 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 764 * Description 765 * Clone and redirect the packet associated to *skb* to another 766 * net device of index *ifindex*. Both ingress and egress 767 * interfaces can be used for redirection. The **BPF_F_INGRESS** 768 * value in *flags* is used to make the distinction (ingress path 769 * is selected if the flag is present, egress path otherwise). 770 * This is the only flag supported for now. 771 * 772 * In comparison with **bpf_redirect**\ () helper, 773 * **bpf_clone_redirect**\ () has the associated cost of 774 * duplicating the packet buffer, but this can be executed out of 775 * the eBPF program. Conversely, **bpf_redirect**\ () is more 776 * efficient, but it is handled through an action code where the 777 * redirection happens only after the eBPF program has returned. 778 * 779 * A call to this helper is susceptible to change the underlying 780 * packet buffer. Therefore, at load time, all checks on pointers 781 * previously done by the verifier are invalidated and must be 782 * performed again, if the helper is used in combination with 783 * direct packet access. 784 * Return 785 * 0 on success, or a negative error in case of failure. Positive 786 * error indicates a potential drop or congestion in the target 787 * device. The particular positive error codes are not defined. 788 * 789 * u64 bpf_get_current_pid_tgid(void) 790 * Return 791 * A 64-bit integer containing the current tgid and pid, and 792 * created as such: 793 * *current_task*\ **->tgid << 32 \|** 794 * *current_task*\ **->pid**. 795 * 796 * u64 bpf_get_current_uid_gid(void) 797 * Return 798 * A 64-bit integer containing the current GID and UID, and 799 * created as such: *current_gid* **<< 32 \|** *current_uid*. 800 * 801 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 802 * Description 803 * Copy the **comm** attribute of the current task into *buf* of 804 * *size_of_buf*. The **comm** attribute contains the name of 805 * the executable (excluding the path) for the current task. The 806 * *size_of_buf* must be strictly positive. On success, the 807 * helper makes sure that the *buf* is NUL-terminated. On failure, 808 * it is filled with zeroes. 809 * Return 810 * 0 on success, or a negative error in case of failure. 811 * 812 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 813 * Description 814 * Retrieve the classid for the current task, i.e. for the net_cls 815 * cgroup to which *skb* belongs. 816 * 817 * This helper can be used on TC egress path, but not on ingress. 818 * 819 * The net_cls cgroup provides an interface to tag network packets 820 * based on a user-provided identifier for all traffic coming from 821 * the tasks belonging to the related cgroup. See also the related 822 * kernel documentation, available from the Linux sources in file 823 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 824 * 825 * The Linux kernel has two versions for cgroups: there are 826 * cgroups v1 and cgroups v2. Both are available to users, who can 827 * use a mixture of them, but note that the net_cls cgroup is for 828 * cgroup v1 only. This makes it incompatible with BPF programs 829 * run on cgroups, which is a cgroup-v2-only feature (a socket can 830 * only hold data for one version of cgroups at a time). 831 * 832 * This helper is only available is the kernel was compiled with 833 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 834 * "**y**" or to "**m**". 835 * Return 836 * The classid, or 0 for the default unconfigured classid. 837 * 838 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 839 * Description 840 * Push a *vlan_tci* (VLAN tag control information) of protocol 841 * *vlan_proto* to the packet associated to *skb*, then update 842 * the checksum. Note that if *vlan_proto* is different from 843 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 844 * be **ETH_P_8021Q**. 845 * 846 * A call to this helper is susceptible to change the underlying 847 * packet buffer. Therefore, at load time, all checks on pointers 848 * previously done by the verifier are invalidated and must be 849 * performed again, if the helper is used in combination with 850 * direct packet access. 851 * Return 852 * 0 on success, or a negative error in case of failure. 853 * 854 * int bpf_skb_vlan_pop(struct sk_buff *skb) 855 * Description 856 * Pop a VLAN header from the packet associated to *skb*. 857 * 858 * A call to this helper is susceptible to change the underlying 859 * packet buffer. Therefore, at load time, all checks on pointers 860 * previously done by the verifier are invalidated and must be 861 * performed again, if the helper is used in combination with 862 * direct packet access. 863 * Return 864 * 0 on success, or a negative error in case of failure. 865 * 866 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 867 * Description 868 * Get tunnel metadata. This helper takes a pointer *key* to an 869 * empty **struct bpf_tunnel_key** of **size**, that will be 870 * filled with tunnel metadata for the packet associated to *skb*. 871 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 872 * indicates that the tunnel is based on IPv6 protocol instead of 873 * IPv4. 874 * 875 * The **struct bpf_tunnel_key** is an object that generalizes the 876 * principal parameters used by various tunneling protocols into a 877 * single struct. This way, it can be used to easily make a 878 * decision based on the contents of the encapsulation header, 879 * "summarized" in this struct. In particular, it holds the IP 880 * address of the remote end (IPv4 or IPv6, depending on the case) 881 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 882 * this struct exposes the *key*\ **->tunnel_id**, which is 883 * generally mapped to a VNI (Virtual Network Identifier), making 884 * it programmable together with the **bpf_skb_set_tunnel_key**\ 885 * () helper. 886 * 887 * Let's imagine that the following code is part of a program 888 * attached to the TC ingress interface, on one end of a GRE 889 * tunnel, and is supposed to filter out all messages coming from 890 * remote ends with IPv4 address other than 10.0.0.1: 891 * 892 * :: 893 * 894 * int ret; 895 * struct bpf_tunnel_key key = {}; 896 * 897 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 898 * if (ret < 0) 899 * return TC_ACT_SHOT; // drop packet 900 * 901 * if (key.remote_ipv4 != 0x0a000001) 902 * return TC_ACT_SHOT; // drop packet 903 * 904 * return TC_ACT_OK; // accept packet 905 * 906 * This interface can also be used with all encapsulation devices 907 * that can operate in "collect metadata" mode: instead of having 908 * one network device per specific configuration, the "collect 909 * metadata" mode only requires a single device where the 910 * configuration can be extracted from this helper. 911 * 912 * This can be used together with various tunnels such as VXLan, 913 * Geneve, GRE or IP in IP (IPIP). 914 * Return 915 * 0 on success, or a negative error in case of failure. 916 * 917 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 918 * Description 919 * Populate tunnel metadata for packet associated to *skb.* The 920 * tunnel metadata is set to the contents of *key*, of *size*. The 921 * *flags* can be set to a combination of the following values: 922 * 923 * **BPF_F_TUNINFO_IPV6** 924 * Indicate that the tunnel is based on IPv6 protocol 925 * instead of IPv4. 926 * **BPF_F_ZERO_CSUM_TX** 927 * For IPv4 packets, add a flag to tunnel metadata 928 * indicating that checksum computation should be skipped 929 * and checksum set to zeroes. 930 * **BPF_F_DONT_FRAGMENT** 931 * Add a flag to tunnel metadata indicating that the 932 * packet should not be fragmented. 933 * **BPF_F_SEQ_NUMBER** 934 * Add a flag to tunnel metadata indicating that a 935 * sequence number should be added to tunnel header before 936 * sending the packet. This flag was added for GRE 937 * encapsulation, but might be used with other protocols 938 * as well in the future. 939 * 940 * Here is a typical usage on the transmit path: 941 * 942 * :: 943 * 944 * struct bpf_tunnel_key key; 945 * populate key ... 946 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 947 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 948 * 949 * See also the description of the **bpf_skb_get_tunnel_key**\ () 950 * helper for additional information. 951 * Return 952 * 0 on success, or a negative error in case of failure. 953 * 954 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 955 * Description 956 * Read the value of a perf event counter. This helper relies on a 957 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 958 * the perf event counter is selected when *map* is updated with 959 * perf event file descriptors. The *map* is an array whose size 960 * is the number of available CPUs, and each cell contains a value 961 * relative to one CPU. The value to retrieve is indicated by 962 * *flags*, that contains the index of the CPU to look up, masked 963 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 964 * **BPF_F_CURRENT_CPU** to indicate that the value for the 965 * current CPU should be retrieved. 966 * 967 * Note that before Linux 4.13, only hardware perf event can be 968 * retrieved. 969 * 970 * Also, be aware that the newer helper 971 * **bpf_perf_event_read_value**\ () is recommended over 972 * **bpf_perf_event_read**\ () in general. The latter has some ABI 973 * quirks where error and counter value are used as a return code 974 * (which is wrong to do since ranges may overlap). This issue is 975 * fixed with **bpf_perf_event_read_value**\ (), which at the same 976 * time provides more features over the **bpf_perf_event_read**\ 977 * () interface. Please refer to the description of 978 * **bpf_perf_event_read_value**\ () for details. 979 * Return 980 * The value of the perf event counter read from the map, or a 981 * negative error code in case of failure. 982 * 983 * int bpf_redirect(u32 ifindex, u64 flags) 984 * Description 985 * Redirect the packet to another net device of index *ifindex*. 986 * This helper is somewhat similar to **bpf_clone_redirect**\ 987 * (), except that the packet is not cloned, which provides 988 * increased performance. 989 * 990 * Except for XDP, both ingress and egress interfaces can be used 991 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 992 * to make the distinction (ingress path is selected if the flag 993 * is present, egress path otherwise). Currently, XDP only 994 * supports redirection to the egress interface, and accepts no 995 * flag at all. 996 * 997 * The same effect can be attained with the more generic 998 * **bpf_redirect_map**\ (), which requires specific maps to be 999 * used but offers better performance. 1000 * Return 1001 * For XDP, the helper returns **XDP_REDIRECT** on success or 1002 * **XDP_ABORTED** on error. For other program types, the values 1003 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1004 * error. 1005 * 1006 * u32 bpf_get_route_realm(struct sk_buff *skb) 1007 * Description 1008 * Retrieve the realm or the route, that is to say the 1009 * **tclassid** field of the destination for the *skb*. The 1010 * indentifier retrieved is a user-provided tag, similar to the 1011 * one used with the net_cls cgroup (see description for 1012 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1013 * held by a route (a destination entry), not by a task. 1014 * 1015 * Retrieving this identifier works with the clsact TC egress hook 1016 * (see also **tc-bpf(8)**), or alternatively on conventional 1017 * classful egress qdiscs, but not on TC ingress path. In case of 1018 * clsact TC egress hook, this has the advantage that, internally, 1019 * the destination entry has not been dropped yet in the transmit 1020 * path. Therefore, the destination entry does not need to be 1021 * artificially held via **netif_keep_dst**\ () for a classful 1022 * qdisc until the *skb* is freed. 1023 * 1024 * This helper is available only if the kernel was compiled with 1025 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1026 * Return 1027 * The realm of the route for the packet associated to *skb*, or 0 1028 * if none was found. 1029 * 1030 * int bpf_perf_event_output(struct pt_regs *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1031 * Description 1032 * Write raw *data* blob into a special BPF perf event held by 1033 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1034 * event must have the following attributes: **PERF_SAMPLE_RAW** 1035 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1036 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1037 * 1038 * The *flags* are used to indicate the index in *map* for which 1039 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1040 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1041 * to indicate that the index of the current CPU core should be 1042 * used. 1043 * 1044 * The value to write, of *size*, is passed through eBPF stack and 1045 * pointed by *data*. 1046 * 1047 * The context of the program *ctx* needs also be passed to the 1048 * helper. 1049 * 1050 * On user space, a program willing to read the values needs to 1051 * call **perf_event_open**\ () on the perf event (either for 1052 * one or for all CPUs) and to store the file descriptor into the 1053 * *map*. This must be done before the eBPF program can send data 1054 * into it. An example is available in file 1055 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1056 * tree (the eBPF program counterpart is in 1057 * *samples/bpf/trace_output_kern.c*). 1058 * 1059 * **bpf_perf_event_output**\ () achieves better performance 1060 * than **bpf_trace_printk**\ () for sharing data with user 1061 * space, and is much better suitable for streaming data from eBPF 1062 * programs. 1063 * 1064 * Note that this helper is not restricted to tracing use cases 1065 * and can be used with programs attached to TC or XDP as well, 1066 * where it allows for passing data to user space listeners. Data 1067 * can be: 1068 * 1069 * * Only custom structs, 1070 * * Only the packet payload, or 1071 * * A combination of both. 1072 * Return 1073 * 0 on success, or a negative error in case of failure. 1074 * 1075 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 1076 * Description 1077 * This helper was provided as an easy way to load data from a 1078 * packet. It can be used to load *len* bytes from *offset* from 1079 * the packet associated to *skb*, into the buffer pointed by 1080 * *to*. 1081 * 1082 * Since Linux 4.7, usage of this helper has mostly been replaced 1083 * by "direct packet access", enabling packet data to be 1084 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1085 * pointing respectively to the first byte of packet data and to 1086 * the byte after the last byte of packet data. However, it 1087 * remains useful if one wishes to read large quantities of data 1088 * at once from a packet into the eBPF stack. 1089 * Return 1090 * 0 on success, or a negative error in case of failure. 1091 * 1092 * int bpf_get_stackid(struct pt_regs *ctx, struct bpf_map *map, u64 flags) 1093 * Description 1094 * Walk a user or a kernel stack and return its id. To achieve 1095 * this, the helper needs *ctx*, which is a pointer to the context 1096 * on which the tracing program is executed, and a pointer to a 1097 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1098 * 1099 * The last argument, *flags*, holds the number of stack frames to 1100 * skip (from 0 to 255), masked with 1101 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1102 * a combination of the following flags: 1103 * 1104 * **BPF_F_USER_STACK** 1105 * Collect a user space stack instead of a kernel stack. 1106 * **BPF_F_FAST_STACK_CMP** 1107 * Compare stacks by hash only. 1108 * **BPF_F_REUSE_STACKID** 1109 * If two different stacks hash into the same *stackid*, 1110 * discard the old one. 1111 * 1112 * The stack id retrieved is a 32 bit long integer handle which 1113 * can be further combined with other data (including other stack 1114 * ids) and used as a key into maps. This can be useful for 1115 * generating a variety of graphs (such as flame graphs or off-cpu 1116 * graphs). 1117 * 1118 * For walking a stack, this helper is an improvement over 1119 * **bpf_probe_read**\ (), which can be used with unrolled loops 1120 * but is not efficient and consumes a lot of eBPF instructions. 1121 * Instead, **bpf_get_stackid**\ () can collect up to 1122 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1123 * this limit can be controlled with the **sysctl** program, and 1124 * that it should be manually increased in order to profile long 1125 * user stacks (such as stacks for Java programs). To do so, use: 1126 * 1127 * :: 1128 * 1129 * # sysctl kernel.perf_event_max_stack=<new value> 1130 * Return 1131 * The positive or null stack id on success, or a negative error 1132 * in case of failure. 1133 * 1134 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1135 * Description 1136 * Compute a checksum difference, from the raw buffer pointed by 1137 * *from*, of length *from_size* (that must be a multiple of 4), 1138 * towards the raw buffer pointed by *to*, of size *to_size* 1139 * (same remark). An optional *seed* can be added to the value 1140 * (this can be cascaded, the seed may come from a previous call 1141 * to the helper). 1142 * 1143 * This is flexible enough to be used in several ways: 1144 * 1145 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1146 * checksum, it can be used when pushing new data. 1147 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1148 * checksum, it can be used when removing data from a packet. 1149 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1150 * can be used to compute a diff. Note that *from_size* and 1151 * *to_size* do not need to be equal. 1152 * 1153 * This helper can be used in combination with 1154 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1155 * which one can feed in the difference computed with 1156 * **bpf_csum_diff**\ (). 1157 * Return 1158 * The checksum result, or a negative error code in case of 1159 * failure. 1160 * 1161 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1162 * Description 1163 * Retrieve tunnel options metadata for the packet associated to 1164 * *skb*, and store the raw tunnel option data to the buffer *opt* 1165 * of *size*. 1166 * 1167 * This helper can be used with encapsulation devices that can 1168 * operate in "collect metadata" mode (please refer to the related 1169 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1170 * more details). A particular example where this can be used is 1171 * in combination with the Geneve encapsulation protocol, where it 1172 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1173 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1174 * the eBPF program. This allows for full customization of these 1175 * headers. 1176 * Return 1177 * The size of the option data retrieved. 1178 * 1179 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1180 * Description 1181 * Set tunnel options metadata for the packet associated to *skb* 1182 * to the option data contained in the raw buffer *opt* of *size*. 1183 * 1184 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1185 * helper for additional information. 1186 * Return 1187 * 0 on success, or a negative error in case of failure. 1188 * 1189 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1190 * Description 1191 * Change the protocol of the *skb* to *proto*. Currently 1192 * supported are transition from IPv4 to IPv6, and from IPv6 to 1193 * IPv4. The helper takes care of the groundwork for the 1194 * transition, including resizing the socket buffer. The eBPF 1195 * program is expected to fill the new headers, if any, via 1196 * **skb_store_bytes**\ () and to recompute the checksums with 1197 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1198 * (). The main case for this helper is to perform NAT64 1199 * operations out of an eBPF program. 1200 * 1201 * Internally, the GSO type is marked as dodgy so that headers are 1202 * checked and segments are recalculated by the GSO/GRO engine. 1203 * The size for GSO target is adapted as well. 1204 * 1205 * All values for *flags* are reserved for future usage, and must 1206 * be left at zero. 1207 * 1208 * A call to this helper is susceptible to change the underlying 1209 * packet buffer. Therefore, at load time, all checks on pointers 1210 * previously done by the verifier are invalidated and must be 1211 * performed again, if the helper is used in combination with 1212 * direct packet access. 1213 * Return 1214 * 0 on success, or a negative error in case of failure. 1215 * 1216 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1217 * Description 1218 * Change the packet type for the packet associated to *skb*. This 1219 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1220 * the eBPF program does not have a write access to *skb*\ 1221 * **->pkt_type** beside this helper. Using a helper here allows 1222 * for graceful handling of errors. 1223 * 1224 * The major use case is to change incoming *skb*s to 1225 * **PACKET_HOST** in a programmatic way instead of having to 1226 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1227 * example. 1228 * 1229 * Note that *type* only allows certain values. At this time, they 1230 * are: 1231 * 1232 * **PACKET_HOST** 1233 * Packet is for us. 1234 * **PACKET_BROADCAST** 1235 * Send packet to all. 1236 * **PACKET_MULTICAST** 1237 * Send packet to group. 1238 * **PACKET_OTHERHOST** 1239 * Send packet to someone else. 1240 * Return 1241 * 0 on success, or a negative error in case of failure. 1242 * 1243 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1244 * Description 1245 * Check whether *skb* is a descendant of the cgroup2 held by 1246 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1247 * Return 1248 * The return value depends on the result of the test, and can be: 1249 * 1250 * * 0, if the *skb* failed the cgroup2 descendant test. 1251 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1252 * * A negative error code, if an error occurred. 1253 * 1254 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1255 * Description 1256 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1257 * not set, in particular if the hash was cleared due to mangling, 1258 * recompute this hash. Later accesses to the hash can be done 1259 * directly with *skb*\ **->hash**. 1260 * 1261 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1262 * prototype with **bpf_skb_change_proto**\ (), or calling 1263 * **bpf_skb_store_bytes**\ () with the 1264 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1265 * the hash and to trigger a new computation for the next call to 1266 * **bpf_get_hash_recalc**\ (). 1267 * Return 1268 * The 32-bit hash. 1269 * 1270 * u64 bpf_get_current_task(void) 1271 * Return 1272 * A pointer to the current task struct. 1273 * 1274 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1275 * Description 1276 * Attempt in a safe way to write *len* bytes from the buffer 1277 * *src* to *dst* in memory. It only works for threads that are in 1278 * user context, and *dst* must be a valid user space address. 1279 * 1280 * This helper should not be used to implement any kind of 1281 * security mechanism because of TOC-TOU attacks, but rather to 1282 * debug, divert, and manipulate execution of semi-cooperative 1283 * processes. 1284 * 1285 * Keep in mind that this feature is meant for experiments, and it 1286 * has a risk of crashing the system and running programs. 1287 * Therefore, when an eBPF program using this helper is attached, 1288 * a warning including PID and process name is printed to kernel 1289 * logs. 1290 * Return 1291 * 0 on success, or a negative error in case of failure. 1292 * 1293 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1294 * Description 1295 * Check whether the probe is being run is the context of a given 1296 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1297 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1298 * Return 1299 * The return value depends on the result of the test, and can be: 1300 * 1301 * * 1, if current task belongs to the cgroup2. 1302 * * 0, if current task does not belong to the cgroup2. 1303 * * A negative error code, if an error occurred. 1304 * 1305 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1306 * Description 1307 * Resize (trim or grow) the packet associated to *skb* to the 1308 * new *len*. The *flags* are reserved for future usage, and must 1309 * be left at zero. 1310 * 1311 * The basic idea is that the helper performs the needed work to 1312 * change the size of the packet, then the eBPF program rewrites 1313 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1314 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1315 * and others. This helper is a slow path utility intended for 1316 * replies with control messages. And because it is targeted for 1317 * slow path, the helper itself can afford to be slow: it 1318 * implicitly linearizes, unclones and drops offloads from the 1319 * *skb*. 1320 * 1321 * A call to this helper is susceptible to change the underlying 1322 * packet buffer. Therefore, at load time, all checks on pointers 1323 * previously done by the verifier are invalidated and must be 1324 * performed again, if the helper is used in combination with 1325 * direct packet access. 1326 * Return 1327 * 0 on success, or a negative error in case of failure. 1328 * 1329 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1330 * Description 1331 * Pull in non-linear data in case the *skb* is non-linear and not 1332 * all of *len* are part of the linear section. Make *len* bytes 1333 * from *skb* readable and writable. If a zero value is passed for 1334 * *len*, then the whole length of the *skb* is pulled. 1335 * 1336 * This helper is only needed for reading and writing with direct 1337 * packet access. 1338 * 1339 * For direct packet access, testing that offsets to access 1340 * are within packet boundaries (test on *skb*\ **->data_end**) is 1341 * susceptible to fail if offsets are invalid, or if the requested 1342 * data is in non-linear parts of the *skb*. On failure the 1343 * program can just bail out, or in the case of a non-linear 1344 * buffer, use a helper to make the data available. The 1345 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1346 * the data. Another one consists in using **bpf_skb_pull_data** 1347 * to pull in once the non-linear parts, then retesting and 1348 * eventually access the data. 1349 * 1350 * At the same time, this also makes sure the *skb* is uncloned, 1351 * which is a necessary condition for direct write. As this needs 1352 * to be an invariant for the write part only, the verifier 1353 * detects writes and adds a prologue that is calling 1354 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1355 * the very beginning in case it is indeed cloned. 1356 * 1357 * A call to this helper is susceptible to change the underlying 1358 * packet buffer. Therefore, at load time, all checks on pointers 1359 * previously done by the verifier are invalidated and must be 1360 * performed again, if the helper is used in combination with 1361 * direct packet access. 1362 * Return 1363 * 0 on success, or a negative error in case of failure. 1364 * 1365 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1366 * Description 1367 * Add the checksum *csum* into *skb*\ **->csum** in case the 1368 * driver has supplied a checksum for the entire packet into that 1369 * field. Return an error otherwise. This helper is intended to be 1370 * used in combination with **bpf_csum_diff**\ (), in particular 1371 * when the checksum needs to be updated after data has been 1372 * written into the packet through direct packet access. 1373 * Return 1374 * The checksum on success, or a negative error code in case of 1375 * failure. 1376 * 1377 * void bpf_set_hash_invalid(struct sk_buff *skb) 1378 * Description 1379 * Invalidate the current *skb*\ **->hash**. It can be used after 1380 * mangling on headers through direct packet access, in order to 1381 * indicate that the hash is outdated and to trigger a 1382 * recalculation the next time the kernel tries to access this 1383 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1384 * 1385 * int bpf_get_numa_node_id(void) 1386 * Description 1387 * Return the id of the current NUMA node. The primary use case 1388 * for this helper is the selection of sockets for the local NUMA 1389 * node, when the program is attached to sockets using the 1390 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1391 * but the helper is also available to other eBPF program types, 1392 * similarly to **bpf_get_smp_processor_id**\ (). 1393 * Return 1394 * The id of current NUMA node. 1395 * 1396 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1397 * Description 1398 * Grows headroom of packet associated to *skb* and adjusts the 1399 * offset of the MAC header accordingly, adding *len* bytes of 1400 * space. It automatically extends and reallocates memory as 1401 * required. 1402 * 1403 * This helper can be used on a layer 3 *skb* to push a MAC header 1404 * for redirection into a layer 2 device. 1405 * 1406 * All values for *flags* are reserved for future usage, and must 1407 * be left at zero. 1408 * 1409 * A call to this helper is susceptible to change the underlying 1410 * packet buffer. Therefore, at load time, all checks on pointers 1411 * previously done by the verifier are invalidated and must be 1412 * performed again, if the helper is used in combination with 1413 * direct packet access. 1414 * Return 1415 * 0 on success, or a negative error in case of failure. 1416 * 1417 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1418 * Description 1419 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1420 * it is possible to use a negative value for *delta*. This helper 1421 * can be used to prepare the packet for pushing or popping 1422 * headers. 1423 * 1424 * A call to this helper is susceptible to change the underlying 1425 * packet buffer. Therefore, at load time, all checks on pointers 1426 * previously done by the verifier are invalidated and must be 1427 * performed again, if the helper is used in combination with 1428 * direct packet access. 1429 * Return 1430 * 0 on success, or a negative error in case of failure. 1431 * 1432 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1433 * Description 1434 * Copy a NUL terminated string from an unsafe address 1435 * *unsafe_ptr* to *dst*. The *size* should include the 1436 * terminating NUL byte. In case the string length is smaller than 1437 * *size*, the target is not padded with further NUL bytes. If the 1438 * string length is larger than *size*, just *size*-1 bytes are 1439 * copied and the last byte is set to NUL. 1440 * 1441 * On success, the length of the copied string is returned. This 1442 * makes this helper useful in tracing programs for reading 1443 * strings, and more importantly to get its length at runtime. See 1444 * the following snippet: 1445 * 1446 * :: 1447 * 1448 * SEC("kprobe/sys_open") 1449 * void bpf_sys_open(struct pt_regs *ctx) 1450 * { 1451 * char buf[PATHLEN]; // PATHLEN is defined to 256 1452 * int res = bpf_probe_read_str(buf, sizeof(buf), 1453 * ctx->di); 1454 * 1455 * // Consume buf, for example push it to 1456 * // userspace via bpf_perf_event_output(); we 1457 * // can use res (the string length) as event 1458 * // size, after checking its boundaries. 1459 * } 1460 * 1461 * In comparison, using **bpf_probe_read()** helper here instead 1462 * to read the string would require to estimate the length at 1463 * compile time, and would often result in copying more memory 1464 * than necessary. 1465 * 1466 * Another useful use case is when parsing individual process 1467 * arguments or individual environment variables navigating 1468 * *current*\ **->mm->arg_start** and *current*\ 1469 * **->mm->env_start**: using this helper and the return value, 1470 * one can quickly iterate at the right offset of the memory area. 1471 * Return 1472 * On success, the strictly positive length of the string, 1473 * including the trailing NUL character. On error, a negative 1474 * value. 1475 * 1476 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1477 * Description 1478 * If the **struct sk_buff** pointed by *skb* has a known socket, 1479 * retrieve the cookie (generated by the kernel) of this socket. 1480 * If no cookie has been set yet, generate a new cookie. Once 1481 * generated, the socket cookie remains stable for the life of the 1482 * socket. This helper can be useful for monitoring per socket 1483 * networking traffic statistics as it provides a global socket 1484 * identifier that can be assumed unique. 1485 * Return 1486 * A 8-byte long non-decreasing number on success, or 0 if the 1487 * socket field is missing inside *skb*. 1488 * 1489 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1490 * Description 1491 * Equivalent to bpf_get_socket_cookie() helper that accepts 1492 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1493 * Return 1494 * A 8-byte long non-decreasing number. 1495 * 1496 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1497 * Description 1498 * Equivalent to bpf_get_socket_cookie() helper that accepts 1499 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1500 * Return 1501 * A 8-byte long non-decreasing number. 1502 * 1503 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1504 * Return 1505 * The owner UID of the socket associated to *skb*. If the socket 1506 * is **NULL**, or if it is not a full socket (i.e. if it is a 1507 * time-wait or a request socket instead), **overflowuid** value 1508 * is returned (note that **overflowuid** might also be the actual 1509 * UID value for the socket). 1510 * 1511 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1512 * Description 1513 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1514 * to value *hash*. 1515 * Return 1516 * 0 1517 * 1518 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1519 * Description 1520 * Emulate a call to **setsockopt()** on the socket associated to 1521 * *bpf_socket*, which must be a full socket. The *level* at 1522 * which the option resides and the name *optname* of the option 1523 * must be specified, see **setsockopt(2)** for more information. 1524 * The option value of length *optlen* is pointed by *optval*. 1525 * 1526 * This helper actually implements a subset of **setsockopt()**. 1527 * It supports the following *level*\ s: 1528 * 1529 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1530 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1531 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1532 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1533 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1534 * **TCP_BPF_SNDCWND_CLAMP**. 1535 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1536 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1537 * Return 1538 * 0 on success, or a negative error in case of failure. 1539 * 1540 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1541 * Description 1542 * Grow or shrink the room for data in the packet associated to 1543 * *skb* by *len_diff*, and according to the selected *mode*. 1544 * 1545 * There are two supported modes at this time: 1546 * 1547 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1548 * (room space is added or removed below the layer 2 header). 1549 * 1550 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1551 * (room space is added or removed below the layer 3 header). 1552 * 1553 * The following flags are supported at this time: 1554 * 1555 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1556 * Adjusting mss in this way is not allowed for datagrams. 1557 * 1558 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1559 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1560 * Any new space is reserved to hold a tunnel header. 1561 * Configure skb offsets and other fields accordingly. 1562 * 1563 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1564 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1565 * Use with ENCAP_L3 flags to further specify the tunnel type. 1566 * 1567 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1568 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1569 * type; *len* is the length of the inner MAC header. 1570 * 1571 * A call to this helper is susceptible to change the underlying 1572 * packet buffer. Therefore, at load time, all checks on pointers 1573 * previously done by the verifier are invalidated and must be 1574 * performed again, if the helper is used in combination with 1575 * direct packet access. 1576 * Return 1577 * 0 on success, or a negative error in case of failure. 1578 * 1579 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1580 * Description 1581 * Redirect the packet to the endpoint referenced by *map* at 1582 * index *key*. Depending on its type, this *map* can contain 1583 * references to net devices (for forwarding packets through other 1584 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1585 * but this is only implemented for native XDP (with driver 1586 * support) as of this writing). 1587 * 1588 * The lower two bits of *flags* are used as the return code if 1589 * the map lookup fails. This is so that the return value can be 1590 * one of the XDP program return codes up to XDP_TX, as chosen by 1591 * the caller. Any higher bits in the *flags* argument must be 1592 * unset. 1593 * 1594 * When used to redirect packets to net devices, this helper 1595 * provides a high performance increase over **bpf_redirect**\ (). 1596 * This is due to various implementation details of the underlying 1597 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1598 * () tries to send packet as a "bulk" to the device. 1599 * Return 1600 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1601 * 1602 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1603 * Description 1604 * Redirect the packet to the socket referenced by *map* (of type 1605 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1606 * egress interfaces can be used for redirection. The 1607 * **BPF_F_INGRESS** value in *flags* is used to make the 1608 * distinction (ingress path is selected if the flag is present, 1609 * egress path otherwise). This is the only flag supported for now. 1610 * Return 1611 * **SK_PASS** on success, or **SK_DROP** on error. 1612 * 1613 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1614 * Description 1615 * Add an entry to, or update a *map* referencing sockets. The 1616 * *skops* is used as a new value for the entry associated to 1617 * *key*. *flags* is one of: 1618 * 1619 * **BPF_NOEXIST** 1620 * The entry for *key* must not exist in the map. 1621 * **BPF_EXIST** 1622 * The entry for *key* must already exist in the map. 1623 * **BPF_ANY** 1624 * No condition on the existence of the entry for *key*. 1625 * 1626 * If the *map* has eBPF programs (parser and verdict), those will 1627 * be inherited by the socket being added. If the socket is 1628 * already attached to eBPF programs, this results in an error. 1629 * Return 1630 * 0 on success, or a negative error in case of failure. 1631 * 1632 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1633 * Description 1634 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1635 * *delta* (which can be positive or negative). Note that this 1636 * operation modifies the address stored in *xdp_md*\ **->data**, 1637 * so the latter must be loaded only after the helper has been 1638 * called. 1639 * 1640 * The use of *xdp_md*\ **->data_meta** is optional and programs 1641 * are not required to use it. The rationale is that when the 1642 * packet is processed with XDP (e.g. as DoS filter), it is 1643 * possible to push further meta data along with it before passing 1644 * to the stack, and to give the guarantee that an ingress eBPF 1645 * program attached as a TC classifier on the same device can pick 1646 * this up for further post-processing. Since TC works with socket 1647 * buffers, it remains possible to set from XDP the **mark** or 1648 * **priority** pointers, or other pointers for the socket buffer. 1649 * Having this scratch space generic and programmable allows for 1650 * more flexibility as the user is free to store whatever meta 1651 * data they need. 1652 * 1653 * A call to this helper is susceptible to change the underlying 1654 * packet buffer. Therefore, at load time, all checks on pointers 1655 * previously done by the verifier are invalidated and must be 1656 * performed again, if the helper is used in combination with 1657 * direct packet access. 1658 * Return 1659 * 0 on success, or a negative error in case of failure. 1660 * 1661 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1662 * Description 1663 * Read the value of a perf event counter, and store it into *buf* 1664 * of size *buf_size*. This helper relies on a *map* of type 1665 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1666 * counter is selected when *map* is updated with perf event file 1667 * descriptors. The *map* is an array whose size is the number of 1668 * available CPUs, and each cell contains a value relative to one 1669 * CPU. The value to retrieve is indicated by *flags*, that 1670 * contains the index of the CPU to look up, masked with 1671 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1672 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1673 * current CPU should be retrieved. 1674 * 1675 * This helper behaves in a way close to 1676 * **bpf_perf_event_read**\ () helper, save that instead of 1677 * just returning the value observed, it fills the *buf* 1678 * structure. This allows for additional data to be retrieved: in 1679 * particular, the enabled and running times (in *buf*\ 1680 * **->enabled** and *buf*\ **->running**, respectively) are 1681 * copied. In general, **bpf_perf_event_read_value**\ () is 1682 * recommended over **bpf_perf_event_read**\ (), which has some 1683 * ABI issues and provides fewer functionalities. 1684 * 1685 * These values are interesting, because hardware PMU (Performance 1686 * Monitoring Unit) counters are limited resources. When there are 1687 * more PMU based perf events opened than available counters, 1688 * kernel will multiplex these events so each event gets certain 1689 * percentage (but not all) of the PMU time. In case that 1690 * multiplexing happens, the number of samples or counter value 1691 * will not reflect the case compared to when no multiplexing 1692 * occurs. This makes comparison between different runs difficult. 1693 * Typically, the counter value should be normalized before 1694 * comparing to other experiments. The usual normalization is done 1695 * as follows. 1696 * 1697 * :: 1698 * 1699 * normalized_counter = counter * t_enabled / t_running 1700 * 1701 * Where t_enabled is the time enabled for event and t_running is 1702 * the time running for event since last normalization. The 1703 * enabled and running times are accumulated since the perf event 1704 * open. To achieve scaling factor between two invocations of an 1705 * eBPF program, users can can use CPU id as the key (which is 1706 * typical for perf array usage model) to remember the previous 1707 * value and do the calculation inside the eBPF program. 1708 * Return 1709 * 0 on success, or a negative error in case of failure. 1710 * 1711 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1712 * Description 1713 * For en eBPF program attached to a perf event, retrieve the 1714 * value of the event counter associated to *ctx* and store it in 1715 * the structure pointed by *buf* and of size *buf_size*. Enabled 1716 * and running times are also stored in the structure (see 1717 * description of helper **bpf_perf_event_read_value**\ () for 1718 * more details). 1719 * Return 1720 * 0 on success, or a negative error in case of failure. 1721 * 1722 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1723 * Description 1724 * Emulate a call to **getsockopt()** on the socket associated to 1725 * *bpf_socket*, which must be a full socket. The *level* at 1726 * which the option resides and the name *optname* of the option 1727 * must be specified, see **getsockopt(2)** for more information. 1728 * The retrieved value is stored in the structure pointed by 1729 * *opval* and of length *optlen*. 1730 * 1731 * This helper actually implements a subset of **getsockopt()**. 1732 * It supports the following *level*\ s: 1733 * 1734 * * **IPPROTO_TCP**, which supports *optname* 1735 * **TCP_CONGESTION**. 1736 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1737 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1738 * Return 1739 * 0 on success, or a negative error in case of failure. 1740 * 1741 * int bpf_override_return(struct pt_regs *regs, u64 rc) 1742 * Description 1743 * Used for error injection, this helper uses kprobes to override 1744 * the return value of the probed function, and to set it to *rc*. 1745 * The first argument is the context *regs* on which the kprobe 1746 * works. 1747 * 1748 * This helper works by setting setting the PC (program counter) 1749 * to an override function which is run in place of the original 1750 * probed function. This means the probed function is not run at 1751 * all. The replacement function just returns with the required 1752 * value. 1753 * 1754 * This helper has security implications, and thus is subject to 1755 * restrictions. It is only available if the kernel was compiled 1756 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1757 * option, and in this case it only works on functions tagged with 1758 * **ALLOW_ERROR_INJECTION** in the kernel code. 1759 * 1760 * Also, the helper is only available for the architectures having 1761 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1762 * x86 architecture is the only one to support this feature. 1763 * Return 1764 * 0 1765 * 1766 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1767 * Description 1768 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1769 * for the full TCP socket associated to *bpf_sock_ops* to 1770 * *argval*. 1771 * 1772 * The primary use of this field is to determine if there should 1773 * be calls to eBPF programs of type 1774 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1775 * code. A program of the same type can change its value, per 1776 * connection and as necessary, when the connection is 1777 * established. This field is directly accessible for reading, but 1778 * this helper must be used for updates in order to return an 1779 * error if an eBPF program tries to set a callback that is not 1780 * supported in the current kernel. 1781 * 1782 * *argval* is a flag array which can combine these flags: 1783 * 1784 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1785 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1786 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1787 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1788 * 1789 * Therefore, this function can be used to clear a callback flag by 1790 * setting the appropriate bit to zero. e.g. to disable the RTO 1791 * callback: 1792 * 1793 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1794 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1795 * 1796 * Here are some examples of where one could call such eBPF 1797 * program: 1798 * 1799 * * When RTO fires. 1800 * * When a packet is retransmitted. 1801 * * When the connection terminates. 1802 * * When a packet is sent. 1803 * * When a packet is received. 1804 * Return 1805 * Code **-EINVAL** if the socket is not a full TCP socket; 1806 * otherwise, a positive number containing the bits that could not 1807 * be set is returned (which comes down to 0 if all bits were set 1808 * as required). 1809 * 1810 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1811 * Description 1812 * This helper is used in programs implementing policies at the 1813 * socket level. If the message *msg* is allowed to pass (i.e. if 1814 * the verdict eBPF program returns **SK_PASS**), redirect it to 1815 * the socket referenced by *map* (of type 1816 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1817 * egress interfaces can be used for redirection. The 1818 * **BPF_F_INGRESS** value in *flags* is used to make the 1819 * distinction (ingress path is selected if the flag is present, 1820 * egress path otherwise). This is the only flag supported for now. 1821 * Return 1822 * **SK_PASS** on success, or **SK_DROP** on error. 1823 * 1824 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1825 * Description 1826 * For socket policies, apply the verdict of the eBPF program to 1827 * the next *bytes* (number of bytes) of message *msg*. 1828 * 1829 * For example, this helper can be used in the following cases: 1830 * 1831 * * A single **sendmsg**\ () or **sendfile**\ () system call 1832 * contains multiple logical messages that the eBPF program is 1833 * supposed to read and for which it should apply a verdict. 1834 * * An eBPF program only cares to read the first *bytes* of a 1835 * *msg*. If the message has a large payload, then setting up 1836 * and calling the eBPF program repeatedly for all bytes, even 1837 * though the verdict is already known, would create unnecessary 1838 * overhead. 1839 * 1840 * When called from within an eBPF program, the helper sets a 1841 * counter internal to the BPF infrastructure, that is used to 1842 * apply the last verdict to the next *bytes*. If *bytes* is 1843 * smaller than the current data being processed from a 1844 * **sendmsg**\ () or **sendfile**\ () system call, the first 1845 * *bytes* will be sent and the eBPF program will be re-run with 1846 * the pointer for start of data pointing to byte number *bytes* 1847 * **+ 1**. If *bytes* is larger than the current data being 1848 * processed, then the eBPF verdict will be applied to multiple 1849 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1850 * consumed. 1851 * 1852 * Note that if a socket closes with the internal counter holding 1853 * a non-zero value, this is not a problem because data is not 1854 * being buffered for *bytes* and is sent as it is received. 1855 * Return 1856 * 0 1857 * 1858 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1859 * Description 1860 * For socket policies, prevent the execution of the verdict eBPF 1861 * program for message *msg* until *bytes* (byte number) have been 1862 * accumulated. 1863 * 1864 * This can be used when one needs a specific number of bytes 1865 * before a verdict can be assigned, even if the data spans 1866 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1867 * case would be a user calling **sendmsg**\ () repeatedly with 1868 * 1-byte long message segments. Obviously, this is bad for 1869 * performance, but it is still valid. If the eBPF program needs 1870 * *bytes* bytes to validate a header, this helper can be used to 1871 * prevent the eBPF program to be called again until *bytes* have 1872 * been accumulated. 1873 * Return 1874 * 0 1875 * 1876 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1877 * Description 1878 * For socket policies, pull in non-linear data from user space 1879 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1880 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1881 * respectively. 1882 * 1883 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1884 * *msg* it can only parse data that the (**data**, **data_end**) 1885 * pointers have already consumed. For **sendmsg**\ () hooks this 1886 * is likely the first scatterlist element. But for calls relying 1887 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1888 * be the range (**0**, **0**) because the data is shared with 1889 * user space and by default the objective is to avoid allowing 1890 * user space to modify data while (or after) eBPF verdict is 1891 * being decided. This helper can be used to pull in data and to 1892 * set the start and end pointer to given values. Data will be 1893 * copied if necessary (i.e. if data was not linear and if start 1894 * and end pointers do not point to the same chunk). 1895 * 1896 * A call to this helper is susceptible to change the underlying 1897 * packet buffer. Therefore, at load time, all checks on pointers 1898 * previously done by the verifier are invalidated and must be 1899 * performed again, if the helper is used in combination with 1900 * direct packet access. 1901 * 1902 * All values for *flags* are reserved for future usage, and must 1903 * be left at zero. 1904 * Return 1905 * 0 on success, or a negative error in case of failure. 1906 * 1907 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1908 * Description 1909 * Bind the socket associated to *ctx* to the address pointed by 1910 * *addr*, of length *addr_len*. This allows for making outgoing 1911 * connection from the desired IP address, which can be useful for 1912 * example when all processes inside a cgroup should use one 1913 * single IP address on a host that has multiple IP configured. 1914 * 1915 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1916 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1917 * **AF_INET6**). Looking for a free port to bind to can be 1918 * expensive, therefore binding to port is not permitted by the 1919 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1920 * must be set to zero. 1921 * Return 1922 * 0 on success, or a negative error in case of failure. 1923 * 1924 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1925 * Description 1926 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1927 * only possible to shrink the packet as of this writing, 1928 * therefore *delta* must be a negative integer. 1929 * 1930 * A call to this helper is susceptible to change the underlying 1931 * packet buffer. Therefore, at load time, all checks on pointers 1932 * previously done by the verifier are invalidated and must be 1933 * performed again, if the helper is used in combination with 1934 * direct packet access. 1935 * Return 1936 * 0 on success, or a negative error in case of failure. 1937 * 1938 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1939 * Description 1940 * Retrieve the XFRM state (IP transform framework, see also 1941 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1942 * 1943 * The retrieved value is stored in the **struct bpf_xfrm_state** 1944 * pointed by *xfrm_state* and of length *size*. 1945 * 1946 * All values for *flags* are reserved for future usage, and must 1947 * be left at zero. 1948 * 1949 * This helper is available only if the kernel was compiled with 1950 * **CONFIG_XFRM** configuration option. 1951 * Return 1952 * 0 on success, or a negative error in case of failure. 1953 * 1954 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1955 * Description 1956 * Return a user or a kernel stack in bpf program provided buffer. 1957 * To achieve this, the helper needs *ctx*, which is a pointer 1958 * to the context on which the tracing program is executed. 1959 * To store the stacktrace, the bpf program provides *buf* with 1960 * a nonnegative *size*. 1961 * 1962 * The last argument, *flags*, holds the number of stack frames to 1963 * skip (from 0 to 255), masked with 1964 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1965 * the following flags: 1966 * 1967 * **BPF_F_USER_STACK** 1968 * Collect a user space stack instead of a kernel stack. 1969 * **BPF_F_USER_BUILD_ID** 1970 * Collect buildid+offset instead of ips for user stack, 1971 * only valid if **BPF_F_USER_STACK** is also specified. 1972 * 1973 * **bpf_get_stack**\ () can collect up to 1974 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1975 * to sufficient large buffer size. Note that 1976 * this limit can be controlled with the **sysctl** program, and 1977 * that it should be manually increased in order to profile long 1978 * user stacks (such as stacks for Java programs). To do so, use: 1979 * 1980 * :: 1981 * 1982 * # sysctl kernel.perf_event_max_stack=<new value> 1983 * Return 1984 * A non-negative value equal to or less than *size* on success, 1985 * or a negative error in case of failure. 1986 * 1987 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1988 * Description 1989 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1990 * it provides an easy way to load *len* bytes from *offset* 1991 * from the packet associated to *skb*, into the buffer pointed 1992 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1993 * a fifth argument *start_header* exists in order to select a 1994 * base offset to start from. *start_header* can be one of: 1995 * 1996 * **BPF_HDR_START_MAC** 1997 * Base offset to load data from is *skb*'s mac header. 1998 * **BPF_HDR_START_NET** 1999 * Base offset to load data from is *skb*'s network header. 2000 * 2001 * In general, "direct packet access" is the preferred method to 2002 * access packet data, however, this helper is in particular useful 2003 * in socket filters where *skb*\ **->data** does not always point 2004 * to the start of the mac header and where "direct packet access" 2005 * is not available. 2006 * Return 2007 * 0 on success, or a negative error in case of failure. 2008 * 2009 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2010 * Description 2011 * Do FIB lookup in kernel tables using parameters in *params*. 2012 * If lookup is successful and result shows packet is to be 2013 * forwarded, the neighbor tables are searched for the nexthop. 2014 * If successful (ie., FIB lookup shows forwarding and nexthop 2015 * is resolved), the nexthop address is returned in ipv4_dst 2016 * or ipv6_dst based on family, smac is set to mac address of 2017 * egress device, dmac is set to nexthop mac address, rt_metric 2018 * is set to metric from route (IPv4/IPv6 only), and ifindex 2019 * is set to the device index of the nexthop from the FIB lookup. 2020 * 2021 * *plen* argument is the size of the passed in struct. 2022 * *flags* argument can be a combination of one or more of the 2023 * following values: 2024 * 2025 * **BPF_FIB_LOOKUP_DIRECT** 2026 * Do a direct table lookup vs full lookup using FIB 2027 * rules. 2028 * **BPF_FIB_LOOKUP_OUTPUT** 2029 * Perform lookup from an egress perspective (default is 2030 * ingress). 2031 * 2032 * *ctx* is either **struct xdp_md** for XDP programs or 2033 * **struct sk_buff** tc cls_act programs. 2034 * Return 2035 * * < 0 if any input argument is invalid 2036 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2037 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2038 * packet is not forwarded or needs assist from full stack 2039 * 2040 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 2041 * Description 2042 * Add an entry to, or update a sockhash *map* referencing sockets. 2043 * The *skops* is used as a new value for the entry associated to 2044 * *key*. *flags* is one of: 2045 * 2046 * **BPF_NOEXIST** 2047 * The entry for *key* must not exist in the map. 2048 * **BPF_EXIST** 2049 * The entry for *key* must already exist in the map. 2050 * **BPF_ANY** 2051 * No condition on the existence of the entry for *key*. 2052 * 2053 * If the *map* has eBPF programs (parser and verdict), those will 2054 * be inherited by the socket being added. If the socket is 2055 * already attached to eBPF programs, this results in an error. 2056 * Return 2057 * 0 on success, or a negative error in case of failure. 2058 * 2059 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2060 * Description 2061 * This helper is used in programs implementing policies at the 2062 * socket level. If the message *msg* is allowed to pass (i.e. if 2063 * the verdict eBPF program returns **SK_PASS**), redirect it to 2064 * the socket referenced by *map* (of type 2065 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2066 * egress interfaces can be used for redirection. The 2067 * **BPF_F_INGRESS** value in *flags* is used to make the 2068 * distinction (ingress path is selected if the flag is present, 2069 * egress path otherwise). This is the only flag supported for now. 2070 * Return 2071 * **SK_PASS** on success, or **SK_DROP** on error. 2072 * 2073 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2074 * Description 2075 * This helper is used in programs implementing policies at the 2076 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2077 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2078 * to the socket referenced by *map* (of type 2079 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2080 * egress interfaces can be used for redirection. The 2081 * **BPF_F_INGRESS** value in *flags* is used to make the 2082 * distinction (ingress path is selected if the flag is present, 2083 * egress otherwise). This is the only flag supported for now. 2084 * Return 2085 * **SK_PASS** on success, or **SK_DROP** on error. 2086 * 2087 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2088 * Description 2089 * Encapsulate the packet associated to *skb* within a Layer 3 2090 * protocol header. This header is provided in the buffer at 2091 * address *hdr*, with *len* its size in bytes. *type* indicates 2092 * the protocol of the header and can be one of: 2093 * 2094 * **BPF_LWT_ENCAP_SEG6** 2095 * IPv6 encapsulation with Segment Routing Header 2096 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2097 * the IPv6 header is computed by the kernel. 2098 * **BPF_LWT_ENCAP_SEG6_INLINE** 2099 * Only works if *skb* contains an IPv6 packet. Insert a 2100 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2101 * the IPv6 header. 2102 * **BPF_LWT_ENCAP_IP** 2103 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2104 * must be IPv4 or IPv6, followed by zero or more 2105 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2106 * total bytes in all prepended headers. Please note that 2107 * if **skb_is_gso**\ (*skb*) is true, no more than two 2108 * headers can be prepended, and the inner header, if 2109 * present, should be either GRE or UDP/GUE. 2110 * 2111 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2112 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2113 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2114 * **BPF_PROG_TYPE_LWT_XMIT**. 2115 * 2116 * A call to this helper is susceptible to change the underlying 2117 * packet buffer. Therefore, at load time, all checks on pointers 2118 * previously done by the verifier are invalidated and must be 2119 * performed again, if the helper is used in combination with 2120 * direct packet access. 2121 * Return 2122 * 0 on success, or a negative error in case of failure. 2123 * 2124 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2125 * Description 2126 * Store *len* bytes from address *from* into the packet 2127 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2128 * inside the outermost IPv6 Segment Routing Header can be 2129 * modified through this helper. 2130 * 2131 * A call to this helper is susceptible to change the underlying 2132 * packet buffer. Therefore, at load time, all checks on pointers 2133 * previously done by the verifier are invalidated and must be 2134 * performed again, if the helper is used in combination with 2135 * direct packet access. 2136 * Return 2137 * 0 on success, or a negative error in case of failure. 2138 * 2139 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2140 * Description 2141 * Adjust the size allocated to TLVs in the outermost IPv6 2142 * Segment Routing Header contained in the packet associated to 2143 * *skb*, at position *offset* by *delta* bytes. Only offsets 2144 * after the segments are accepted. *delta* can be as well 2145 * positive (growing) as negative (shrinking). 2146 * 2147 * A call to this helper is susceptible to change the underlying 2148 * packet buffer. Therefore, at load time, all checks on pointers 2149 * previously done by the verifier are invalidated and must be 2150 * performed again, if the helper is used in combination with 2151 * direct packet access. 2152 * Return 2153 * 0 on success, or a negative error in case of failure. 2154 * 2155 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2156 * Description 2157 * Apply an IPv6 Segment Routing action of type *action* to the 2158 * packet associated to *skb*. Each action takes a parameter 2159 * contained at address *param*, and of length *param_len* bytes. 2160 * *action* can be one of: 2161 * 2162 * **SEG6_LOCAL_ACTION_END_X** 2163 * End.X action: Endpoint with Layer-3 cross-connect. 2164 * Type of *param*: **struct in6_addr**. 2165 * **SEG6_LOCAL_ACTION_END_T** 2166 * End.T action: Endpoint with specific IPv6 table lookup. 2167 * Type of *param*: **int**. 2168 * **SEG6_LOCAL_ACTION_END_B6** 2169 * End.B6 action: Endpoint bound to an SRv6 policy. 2170 * Type of *param*: **struct ipv6_sr_hdr**. 2171 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2172 * End.B6.Encap action: Endpoint bound to an SRv6 2173 * encapsulation policy. 2174 * Type of *param*: **struct ipv6_sr_hdr**. 2175 * 2176 * A call to this helper is susceptible to change the underlying 2177 * packet buffer. Therefore, at load time, all checks on pointers 2178 * previously done by the verifier are invalidated and must be 2179 * performed again, if the helper is used in combination with 2180 * direct packet access. 2181 * Return 2182 * 0 on success, or a negative error in case of failure. 2183 * 2184 * int bpf_rc_repeat(void *ctx) 2185 * Description 2186 * This helper is used in programs implementing IR decoding, to 2187 * report a successfully decoded repeat key message. This delays 2188 * the generation of a key up event for previously generated 2189 * key down event. 2190 * 2191 * Some IR protocols like NEC have a special IR message for 2192 * repeating last button, for when a button is held down. 2193 * 2194 * The *ctx* should point to the lirc sample as passed into 2195 * the program. 2196 * 2197 * This helper is only available is the kernel was compiled with 2198 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2199 * "**y**". 2200 * Return 2201 * 0 2202 * 2203 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2204 * Description 2205 * This helper is used in programs implementing IR decoding, to 2206 * report a successfully decoded key press with *scancode*, 2207 * *toggle* value in the given *protocol*. The scancode will be 2208 * translated to a keycode using the rc keymap, and reported as 2209 * an input key down event. After a period a key up event is 2210 * generated. This period can be extended by calling either 2211 * **bpf_rc_keydown**\ () again with the same values, or calling 2212 * **bpf_rc_repeat**\ (). 2213 * 2214 * Some protocols include a toggle bit, in case the button was 2215 * released and pressed again between consecutive scancodes. 2216 * 2217 * The *ctx* should point to the lirc sample as passed into 2218 * the program. 2219 * 2220 * The *protocol* is the decoded protocol number (see 2221 * **enum rc_proto** for some predefined values). 2222 * 2223 * This helper is only available is the kernel was compiled with 2224 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2225 * "**y**". 2226 * Return 2227 * 0 2228 * 2229 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2230 * Description 2231 * Return the cgroup v2 id of the socket associated with the *skb*. 2232 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2233 * helper for cgroup v1 by providing a tag resp. identifier that 2234 * can be matched on or used for map lookups e.g. to implement 2235 * policy. The cgroup v2 id of a given path in the hierarchy is 2236 * exposed in user space through the f_handle API in order to get 2237 * to the same 64-bit id. 2238 * 2239 * This helper can be used on TC egress path, but not on ingress, 2240 * and is available only if the kernel was compiled with the 2241 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2242 * Return 2243 * The id is returned or 0 in case the id could not be retrieved. 2244 * 2245 * u64 bpf_get_current_cgroup_id(void) 2246 * Return 2247 * A 64-bit integer containing the current cgroup id based 2248 * on the cgroup within which the current task is running. 2249 * 2250 * void *bpf_get_local_storage(void *map, u64 flags) 2251 * Description 2252 * Get the pointer to the local storage area. 2253 * The type and the size of the local storage is defined 2254 * by the *map* argument. 2255 * The *flags* meaning is specific for each map type, 2256 * and has to be 0 for cgroup local storage. 2257 * 2258 * Depending on the BPF program type, a local storage area 2259 * can be shared between multiple instances of the BPF program, 2260 * running simultaneously. 2261 * 2262 * A user should care about the synchronization by himself. 2263 * For example, by using the **BPF_STX_XADD** instruction to alter 2264 * the shared data. 2265 * Return 2266 * A pointer to the local storage area. 2267 * 2268 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2269 * Description 2270 * Select a **SO_REUSEPORT** socket from a 2271 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 2272 * It checks the selected socket is matching the incoming 2273 * request in the socket buffer. 2274 * Return 2275 * 0 on success, or a negative error in case of failure. 2276 * 2277 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2278 * Description 2279 * Return id of cgroup v2 that is ancestor of cgroup associated 2280 * with the *skb* at the *ancestor_level*. The root cgroup is at 2281 * *ancestor_level* zero and each step down the hierarchy 2282 * increments the level. If *ancestor_level* == level of cgroup 2283 * associated with *skb*, then return value will be same as that 2284 * of **bpf_skb_cgroup_id**\ (). 2285 * 2286 * The helper is useful to implement policies based on cgroups 2287 * that are upper in hierarchy than immediate cgroup associated 2288 * with *skb*. 2289 * 2290 * The format of returned id and helper limitations are same as in 2291 * **bpf_skb_cgroup_id**\ (). 2292 * Return 2293 * The id is returned or 0 in case the id could not be retrieved. 2294 * 2295 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2296 * Description 2297 * Look for TCP socket matching *tuple*, optionally in a child 2298 * network namespace *netns*. The return value must be checked, 2299 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2300 * 2301 * The *ctx* should point to the context of the program, such as 2302 * the skb or socket (depending on the hook in use). This is used 2303 * to determine the base network namespace for the lookup. 2304 * 2305 * *tuple_size* must be one of: 2306 * 2307 * **sizeof**\ (*tuple*\ **->ipv4**) 2308 * Look for an IPv4 socket. 2309 * **sizeof**\ (*tuple*\ **->ipv6**) 2310 * Look for an IPv6 socket. 2311 * 2312 * If the *netns* is a negative signed 32-bit integer, then the 2313 * socket lookup table in the netns associated with the *ctx* will 2314 * will be used. For the TC hooks, this is the netns of the device 2315 * in the skb. For socket hooks, this is the netns of the socket. 2316 * If *netns* is any other signed 32-bit value greater than or 2317 * equal to zero then it specifies the ID of the netns relative to 2318 * the netns associated with the *ctx*. *netns* values beyond the 2319 * range of 32-bit integers are reserved for future use. 2320 * 2321 * All values for *flags* are reserved for future usage, and must 2322 * be left at zero. 2323 * 2324 * This helper is available only if the kernel was compiled with 2325 * **CONFIG_NET** configuration option. 2326 * Return 2327 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2328 * For sockets with reuseport option, the **struct bpf_sock** 2329 * result is from *reuse*\ **->socks**\ [] using the hash of the 2330 * tuple. 2331 * 2332 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2333 * Description 2334 * Look for UDP socket matching *tuple*, optionally in a child 2335 * network namespace *netns*. The return value must be checked, 2336 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2337 * 2338 * The *ctx* should point to the context of the program, such as 2339 * the skb or socket (depending on the hook in use). This is used 2340 * to determine the base network namespace for the lookup. 2341 * 2342 * *tuple_size* must be one of: 2343 * 2344 * **sizeof**\ (*tuple*\ **->ipv4**) 2345 * Look for an IPv4 socket. 2346 * **sizeof**\ (*tuple*\ **->ipv6**) 2347 * Look for an IPv6 socket. 2348 * 2349 * If the *netns* is a negative signed 32-bit integer, then the 2350 * socket lookup table in the netns associated with the *ctx* will 2351 * will be used. For the TC hooks, this is the netns of the device 2352 * in the skb. For socket hooks, this is the netns of the socket. 2353 * If *netns* is any other signed 32-bit value greater than or 2354 * equal to zero then it specifies the ID of the netns relative to 2355 * the netns associated with the *ctx*. *netns* values beyond the 2356 * range of 32-bit integers are reserved for future use. 2357 * 2358 * All values for *flags* are reserved for future usage, and must 2359 * be left at zero. 2360 * 2361 * This helper is available only if the kernel was compiled with 2362 * **CONFIG_NET** configuration option. 2363 * Return 2364 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2365 * For sockets with reuseport option, the **struct bpf_sock** 2366 * result is from *reuse*\ **->socks**\ [] using the hash of the 2367 * tuple. 2368 * 2369 * int bpf_sk_release(struct bpf_sock *sock) 2370 * Description 2371 * Release the reference held by *sock*. *sock* must be a 2372 * non-**NULL** pointer that was returned from 2373 * **bpf_sk_lookup_xxx**\ (). 2374 * Return 2375 * 0 on success, or a negative error in case of failure. 2376 * 2377 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2378 * Description 2379 * Push an element *value* in *map*. *flags* is one of: 2380 * 2381 * **BPF_EXIST** 2382 * If the queue/stack is full, the oldest element is 2383 * removed to make room for this. 2384 * Return 2385 * 0 on success, or a negative error in case of failure. 2386 * 2387 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2388 * Description 2389 * Pop an element from *map*. 2390 * Return 2391 * 0 on success, or a negative error in case of failure. 2392 * 2393 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2394 * Description 2395 * Get an element from *map* without removing it. 2396 * Return 2397 * 0 on success, or a negative error in case of failure. 2398 * 2399 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 2400 * Description 2401 * For socket policies, insert *len* bytes into *msg* at offset 2402 * *start*. 2403 * 2404 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2405 * *msg* it may want to insert metadata or options into the *msg*. 2406 * This can later be read and used by any of the lower layer BPF 2407 * hooks. 2408 * 2409 * This helper may fail if under memory pressure (a malloc 2410 * fails) in these cases BPF programs will get an appropriate 2411 * error and BPF programs will need to handle them. 2412 * Return 2413 * 0 on success, or a negative error in case of failure. 2414 * 2415 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags) 2416 * Description 2417 * Will remove *pop* bytes from a *msg* starting at byte *start*. 2418 * This may result in **ENOMEM** errors under certain situations if 2419 * an allocation and copy are required due to a full ring buffer. 2420 * However, the helper will try to avoid doing the allocation 2421 * if possible. Other errors can occur if input parameters are 2422 * invalid either due to *start* byte not being valid part of *msg* 2423 * payload and/or *pop* value being to large. 2424 * Return 2425 * 0 on success, or a negative error in case of failure. 2426 * 2427 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2428 * Description 2429 * This helper is used in programs implementing IR decoding, to 2430 * report a successfully decoded pointer movement. 2431 * 2432 * The *ctx* should point to the lirc sample as passed into 2433 * the program. 2434 * 2435 * This helper is only available is the kernel was compiled with 2436 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2437 * "**y**". 2438 * Return 2439 * 0 2440 * 2441 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2442 * Description 2443 * Acquire a spinlock represented by the pointer *lock*, which is 2444 * stored as part of a value of a map. Taking the lock allows to 2445 * safely update the rest of the fields in that value. The 2446 * spinlock can (and must) later be released with a call to 2447 * **bpf_spin_unlock**\ (\ *lock*\ ). 2448 * 2449 * Spinlocks in BPF programs come with a number of restrictions 2450 * and constraints: 2451 * 2452 * * **bpf_spin_lock** objects are only allowed inside maps of 2453 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2454 * list could be extended in the future). 2455 * * BTF description of the map is mandatory. 2456 * * The BPF program can take ONE lock at a time, since taking two 2457 * or more could cause dead locks. 2458 * * Only one **struct bpf_spin_lock** is allowed per map element. 2459 * * When the lock is taken, calls (either BPF to BPF or helpers) 2460 * are not allowed. 2461 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2462 * allowed inside a spinlock-ed region. 2463 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2464 * the lock, on all execution paths, before it returns. 2465 * * The BPF program can access **struct bpf_spin_lock** only via 2466 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2467 * helpers. Loading or storing data into the **struct 2468 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2469 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2470 * of the map value must be a struct and have **struct 2471 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2472 * Nested lock inside another struct is not allowed. 2473 * * The **struct bpf_spin_lock** *lock* field in a map value must 2474 * be aligned on a multiple of 4 bytes in that value. 2475 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2476 * the **bpf_spin_lock** field to user space. 2477 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2478 * a BPF program, do not update the **bpf_spin_lock** field. 2479 * * **bpf_spin_lock** cannot be on the stack or inside a 2480 * networking packet (it can only be inside of a map values). 2481 * * **bpf_spin_lock** is available to root only. 2482 * * Tracing programs and socket filter programs cannot use 2483 * **bpf_spin_lock**\ () due to insufficient preemption checks 2484 * (but this may change in the future). 2485 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2486 * Return 2487 * 0 2488 * 2489 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2490 * Description 2491 * Release the *lock* previously locked by a call to 2492 * **bpf_spin_lock**\ (\ *lock*\ ). 2493 * Return 2494 * 0 2495 * 2496 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2497 * Description 2498 * This helper gets a **struct bpf_sock** pointer such 2499 * that all the fields in this **bpf_sock** can be accessed. 2500 * Return 2501 * A **struct bpf_sock** pointer on success, or **NULL** in 2502 * case of failure. 2503 * 2504 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2505 * Description 2506 * This helper gets a **struct bpf_tcp_sock** pointer from a 2507 * **struct bpf_sock** pointer. 2508 * Return 2509 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2510 * case of failure. 2511 * 2512 * int bpf_skb_ecn_set_ce(struct sk_buf *skb) 2513 * Description 2514 * Set ECN (Explicit Congestion Notification) field of IP header 2515 * to **CE** (Congestion Encountered) if current value is **ECT** 2516 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2517 * and IPv4. 2518 * Return 2519 * 1 if the **CE** flag is set (either by the current helper call 2520 * or because it was already present), 0 if it is not set. 2521 * 2522 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2523 * Description 2524 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2525 * **bpf_sk_release**\ () is unnecessary and not allowed. 2526 * Return 2527 * A **struct bpf_sock** pointer on success, or **NULL** in 2528 * case of failure. 2529 * 2530 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2531 * Description 2532 * Look for TCP socket matching *tuple*, optionally in a child 2533 * network namespace *netns*. The return value must be checked, 2534 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2535 * 2536 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2537 * that it also returns timewait or request sockets. Use 2538 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2539 * full structure. 2540 * 2541 * This helper is available only if the kernel was compiled with 2542 * **CONFIG_NET** configuration option. 2543 * Return 2544 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2545 * For sockets with reuseport option, the **struct bpf_sock** 2546 * result is from *reuse*\ **->socks**\ [] using the hash of the 2547 * tuple. 2548 * 2549 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2550 * Description 2551 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2552 * the listening socket in *sk*. 2553 * 2554 * *iph* points to the start of the IPv4 or IPv6 header, while 2555 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2556 * **sizeof**\ (**struct ip6hdr**). 2557 * 2558 * *th* points to the start of the TCP header, while *th_len* 2559 * contains **sizeof**\ (**struct tcphdr**). 2560 * 2561 * Return 2562 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2563 * error otherwise. 2564 * 2565 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2566 * Description 2567 * Get name of sysctl in /proc/sys/ and copy it into provided by 2568 * program buffer *buf* of size *buf_len*. 2569 * 2570 * The buffer is always NUL terminated, unless it's zero-sized. 2571 * 2572 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2573 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2574 * only (e.g. "tcp_mem"). 2575 * Return 2576 * Number of character copied (not including the trailing NUL). 2577 * 2578 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2579 * truncated name in this case). 2580 * 2581 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2582 * Description 2583 * Get current value of sysctl as it is presented in /proc/sys 2584 * (incl. newline, etc), and copy it as a string into provided 2585 * by program buffer *buf* of size *buf_len*. 2586 * 2587 * The whole value is copied, no matter what file position user 2588 * space issued e.g. sys_read at. 2589 * 2590 * The buffer is always NUL terminated, unless it's zero-sized. 2591 * Return 2592 * Number of character copied (not including the trailing NUL). 2593 * 2594 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2595 * truncated name in this case). 2596 * 2597 * **-EINVAL** if current value was unavailable, e.g. because 2598 * sysctl is uninitialized and read returns -EIO for it. 2599 * 2600 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2601 * Description 2602 * Get new value being written by user space to sysctl (before 2603 * the actual write happens) and copy it as a string into 2604 * provided by program buffer *buf* of size *buf_len*. 2605 * 2606 * User space may write new value at file position > 0. 2607 * 2608 * The buffer is always NUL terminated, unless it's zero-sized. 2609 * Return 2610 * Number of character copied (not including the trailing NUL). 2611 * 2612 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2613 * truncated name in this case). 2614 * 2615 * **-EINVAL** if sysctl is being read. 2616 * 2617 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2618 * Description 2619 * Override new value being written by user space to sysctl with 2620 * value provided by program in buffer *buf* of size *buf_len*. 2621 * 2622 * *buf* should contain a string in same form as provided by user 2623 * space on sysctl write. 2624 * 2625 * User space may write new value at file position > 0. To override 2626 * the whole sysctl value file position should be set to zero. 2627 * Return 2628 * 0 on success. 2629 * 2630 * **-E2BIG** if the *buf_len* is too big. 2631 * 2632 * **-EINVAL** if sysctl is being read. 2633 * 2634 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2635 * Description 2636 * Convert the initial part of the string from buffer *buf* of 2637 * size *buf_len* to a long integer according to the given base 2638 * and save the result in *res*. 2639 * 2640 * The string may begin with an arbitrary amount of white space 2641 * (as determined by **isspace**\ (3)) followed by a single 2642 * optional '**-**' sign. 2643 * 2644 * Five least significant bits of *flags* encode base, other bits 2645 * are currently unused. 2646 * 2647 * Base must be either 8, 10, 16 or 0 to detect it automatically 2648 * similar to user space **strtol**\ (3). 2649 * Return 2650 * Number of characters consumed on success. Must be positive but 2651 * no more than *buf_len*. 2652 * 2653 * **-EINVAL** if no valid digits were found or unsupported base 2654 * was provided. 2655 * 2656 * **-ERANGE** if resulting value was out of range. 2657 * 2658 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2659 * Description 2660 * Convert the initial part of the string from buffer *buf* of 2661 * size *buf_len* to an unsigned long integer according to the 2662 * given base and save the result in *res*. 2663 * 2664 * The string may begin with an arbitrary amount of white space 2665 * (as determined by **isspace**\ (3)). 2666 * 2667 * Five least significant bits of *flags* encode base, other bits 2668 * are currently unused. 2669 * 2670 * Base must be either 8, 10, 16 or 0 to detect it automatically 2671 * similar to user space **strtoul**\ (3). 2672 * Return 2673 * Number of characters consumed on success. Must be positive but 2674 * no more than *buf_len*. 2675 * 2676 * **-EINVAL** if no valid digits were found or unsupported base 2677 * was provided. 2678 * 2679 * **-ERANGE** if resulting value was out of range. 2680 * 2681 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2682 * Description 2683 * Get a bpf-local-storage from a *sk*. 2684 * 2685 * Logically, it could be thought of getting the value from 2686 * a *map* with *sk* as the **key**. From this 2687 * perspective, the usage is not much different from 2688 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2689 * helper enforces the key must be a full socket and the map must 2690 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2691 * 2692 * Underneath, the value is stored locally at *sk* instead of 2693 * the *map*. The *map* is used as the bpf-local-storage 2694 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2695 * searched against all bpf-local-storages residing at *sk*. 2696 * 2697 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2698 * used such that a new bpf-local-storage will be 2699 * created if one does not exist. *value* can be used 2700 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2701 * the initial value of a bpf-local-storage. If *value* is 2702 * **NULL**, the new bpf-local-storage will be zero initialized. 2703 * Return 2704 * A bpf-local-storage pointer is returned on success. 2705 * 2706 * **NULL** if not found or there was an error in adding 2707 * a new bpf-local-storage. 2708 * 2709 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2710 * Description 2711 * Delete a bpf-local-storage from a *sk*. 2712 * Return 2713 * 0 on success. 2714 * 2715 * **-ENOENT** if the bpf-local-storage cannot be found. 2716 * 2717 * int bpf_send_signal(u32 sig) 2718 * Description 2719 * Send signal *sig* to the current task. 2720 * Return 2721 * 0 on success or successfully queued. 2722 * 2723 * **-EBUSY** if work queue under nmi is full. 2724 * 2725 * **-EINVAL** if *sig* is invalid. 2726 * 2727 * **-EPERM** if no permission to send the *sig*. 2728 * 2729 * **-EAGAIN** if bpf program can try again. 2730 * 2731 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2732 * Description 2733 * Try to issue a SYN cookie for the packet with corresponding 2734 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2735 * 2736 * *iph* points to the start of the IPv4 or IPv6 header, while 2737 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2738 * **sizeof**\ (**struct ip6hdr**). 2739 * 2740 * *th* points to the start of the TCP header, while *th_len* 2741 * contains the length of the TCP header. 2742 * 2743 * Return 2744 * On success, lower 32 bits hold the generated SYN cookie in 2745 * followed by 16 bits which hold the MSS value for that cookie, 2746 * and the top 16 bits are unused. 2747 * 2748 * On failure, the returned value is one of the following: 2749 * 2750 * **-EINVAL** SYN cookie cannot be issued due to error 2751 * 2752 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2753 * 2754 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2755 * 2756 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2757 * 2758 * u64 bpf_ktime_get_boot_ns(void) 2759 * Description 2760 * Return the time elapsed since system boot, in nanoseconds. 2761 * Does include the time the system was suspended. 2762 * See: clock_gettime(CLOCK_BOOTTIME) 2763 * Return 2764 * Current *ktime*. 2765 */ 2766 #define __BPF_FUNC_MAPPER(FN) \ 2767 FN(unspec), \ 2768 FN(map_lookup_elem), \ 2769 FN(map_update_elem), \ 2770 FN(map_delete_elem), \ 2771 FN(probe_read), \ 2772 FN(ktime_get_ns), \ 2773 FN(trace_printk), \ 2774 FN(get_prandom_u32), \ 2775 FN(get_smp_processor_id), \ 2776 FN(skb_store_bytes), \ 2777 FN(l3_csum_replace), \ 2778 FN(l4_csum_replace), \ 2779 FN(tail_call), \ 2780 FN(clone_redirect), \ 2781 FN(get_current_pid_tgid), \ 2782 FN(get_current_uid_gid), \ 2783 FN(get_current_comm), \ 2784 FN(get_cgroup_classid), \ 2785 FN(skb_vlan_push), \ 2786 FN(skb_vlan_pop), \ 2787 FN(skb_get_tunnel_key), \ 2788 FN(skb_set_tunnel_key), \ 2789 FN(perf_event_read), \ 2790 FN(redirect), \ 2791 FN(get_route_realm), \ 2792 FN(perf_event_output), \ 2793 FN(skb_load_bytes), \ 2794 FN(get_stackid), \ 2795 FN(csum_diff), \ 2796 FN(skb_get_tunnel_opt), \ 2797 FN(skb_set_tunnel_opt), \ 2798 FN(skb_change_proto), \ 2799 FN(skb_change_type), \ 2800 FN(skb_under_cgroup), \ 2801 FN(get_hash_recalc), \ 2802 FN(get_current_task), \ 2803 FN(probe_write_user), \ 2804 FN(current_task_under_cgroup), \ 2805 FN(skb_change_tail), \ 2806 FN(skb_pull_data), \ 2807 FN(csum_update), \ 2808 FN(set_hash_invalid), \ 2809 FN(get_numa_node_id), \ 2810 FN(skb_change_head), \ 2811 FN(xdp_adjust_head), \ 2812 FN(probe_read_str), \ 2813 FN(get_socket_cookie), \ 2814 FN(get_socket_uid), \ 2815 FN(set_hash), \ 2816 FN(setsockopt), \ 2817 FN(skb_adjust_room), \ 2818 FN(redirect_map), \ 2819 FN(sk_redirect_map), \ 2820 FN(sock_map_update), \ 2821 FN(xdp_adjust_meta), \ 2822 FN(perf_event_read_value), \ 2823 FN(perf_prog_read_value), \ 2824 FN(getsockopt), \ 2825 FN(override_return), \ 2826 FN(sock_ops_cb_flags_set), \ 2827 FN(msg_redirect_map), \ 2828 FN(msg_apply_bytes), \ 2829 FN(msg_cork_bytes), \ 2830 FN(msg_pull_data), \ 2831 FN(bind), \ 2832 FN(xdp_adjust_tail), \ 2833 FN(skb_get_xfrm_state), \ 2834 FN(get_stack), \ 2835 FN(skb_load_bytes_relative), \ 2836 FN(fib_lookup), \ 2837 FN(sock_hash_update), \ 2838 FN(msg_redirect_hash), \ 2839 FN(sk_redirect_hash), \ 2840 FN(lwt_push_encap), \ 2841 FN(lwt_seg6_store_bytes), \ 2842 FN(lwt_seg6_adjust_srh), \ 2843 FN(lwt_seg6_action), \ 2844 FN(rc_repeat), \ 2845 FN(rc_keydown), \ 2846 FN(skb_cgroup_id), \ 2847 FN(get_current_cgroup_id), \ 2848 FN(get_local_storage), \ 2849 FN(sk_select_reuseport), \ 2850 FN(skb_ancestor_cgroup_id), \ 2851 FN(sk_lookup_tcp), \ 2852 FN(sk_lookup_udp), \ 2853 FN(sk_release), \ 2854 FN(map_push_elem), \ 2855 FN(map_pop_elem), \ 2856 FN(map_peek_elem), \ 2857 FN(msg_push_data), \ 2858 FN(msg_pop_data), \ 2859 FN(rc_pointer_rel), \ 2860 FN(spin_lock), \ 2861 FN(spin_unlock), \ 2862 FN(sk_fullsock), \ 2863 FN(tcp_sock), \ 2864 FN(skb_ecn_set_ce), \ 2865 FN(get_listener_sock), \ 2866 FN(skc_lookup_tcp), \ 2867 FN(tcp_check_syncookie), \ 2868 FN(sysctl_get_name), \ 2869 FN(sysctl_get_current_value), \ 2870 FN(sysctl_get_new_value), \ 2871 FN(sysctl_set_new_value), \ 2872 FN(strtol), \ 2873 FN(strtoul), \ 2874 FN(sk_storage_get), \ 2875 FN(sk_storage_delete), \ 2876 FN(send_signal), \ 2877 FN(tcp_gen_syncookie), \ 2878 FN(skb_output), \ 2879 FN(probe_read_user), \ 2880 FN(probe_read_kernel), \ 2881 FN(probe_read_user_str), \ 2882 FN(probe_read_kernel_str), \ 2883 FN(tcp_send_ack), \ 2884 FN(send_signal_thread), \ 2885 FN(jiffies64), \ 2886 FN(read_branch_records), \ 2887 FN(get_ns_current_pid_tgid), \ 2888 FN(xdp_output), \ 2889 FN(get_netns_cookie), \ 2890 FN(get_current_ancestor_cgroup_id), \ 2891 FN(sk_assign), \ 2892 FN(ktime_get_boot_ns), 2893 2894 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2895 * function eBPF program intends to call 2896 */ 2897 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2898 enum bpf_func_id { 2899 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2900 __BPF_FUNC_MAX_ID, 2901 }; 2902 #undef __BPF_ENUM_FN 2903 2904 /* All flags used by eBPF helper functions, placed here. */ 2905 2906 /* BPF_FUNC_skb_store_bytes flags. */ 2907 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2908 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2909 2910 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2911 * First 4 bits are for passing the header field size. 2912 */ 2913 #define BPF_F_HDR_FIELD_MASK 0xfULL 2914 2915 /* BPF_FUNC_l4_csum_replace flags. */ 2916 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2917 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2918 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2919 2920 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2921 #define BPF_F_INGRESS (1ULL << 0) 2922 2923 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2924 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2925 2926 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2927 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2928 #define BPF_F_USER_STACK (1ULL << 8) 2929 /* flags used by BPF_FUNC_get_stackid only. */ 2930 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2931 #define BPF_F_REUSE_STACKID (1ULL << 10) 2932 /* flags used by BPF_FUNC_get_stack only. */ 2933 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2934 2935 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2936 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2937 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2938 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2939 2940 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2941 * BPF_FUNC_perf_event_read_value flags. 2942 */ 2943 #define BPF_F_INDEX_MASK 0xffffffffULL 2944 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2945 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2946 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2947 2948 /* Current network namespace */ 2949 #define BPF_F_CURRENT_NETNS (-1L) 2950 2951 /* BPF_FUNC_skb_adjust_room flags. */ 2952 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 2953 2954 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 2955 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 2956 2957 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 2958 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 2959 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 2960 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 2961 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 2962 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 2963 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 2964 2965 /* BPF_FUNC_sysctl_get_name flags. */ 2966 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0) 2967 2968 /* BPF_FUNC_sk_storage_get flags */ 2969 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0) 2970 2971 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 2972 enum bpf_adj_room_mode { 2973 BPF_ADJ_ROOM_NET, 2974 BPF_ADJ_ROOM_MAC, 2975 }; 2976 2977 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 2978 enum bpf_hdr_start_off { 2979 BPF_HDR_START_MAC, 2980 BPF_HDR_START_NET, 2981 }; 2982 2983 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2984 enum bpf_lwt_encap_mode { 2985 BPF_LWT_ENCAP_SEG6, 2986 BPF_LWT_ENCAP_SEG6_INLINE, 2987 BPF_LWT_ENCAP_IP, 2988 }; 2989 2990 #define __bpf_md_ptr(type, name) \ 2991 union { \ 2992 type name; \ 2993 __u64 :64; \ 2994 } __attribute__((aligned(8))) 2995 2996 /* user accessible mirror of in-kernel sk_buff. 2997 * new fields can only be added to the end of this structure 2998 */ 2999 struct __sk_buff { 3000 __u32 len; 3001 __u32 pkt_type; 3002 __u32 mark; 3003 __u32 queue_mapping; 3004 __u32 protocol; 3005 __u32 vlan_present; 3006 __u32 vlan_tci; 3007 __u32 vlan_proto; 3008 __u32 priority; 3009 __u32 ingress_ifindex; 3010 __u32 ifindex; 3011 __u32 tc_index; 3012 __u32 cb[5]; 3013 __u32 hash; 3014 __u32 tc_classid; 3015 __u32 data; 3016 __u32 data_end; 3017 __u32 napi_id; 3018 3019 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3020 __u32 family; 3021 __u32 remote_ip4; /* Stored in network byte order */ 3022 __u32 local_ip4; /* Stored in network byte order */ 3023 __u32 remote_ip6[4]; /* Stored in network byte order */ 3024 __u32 local_ip6[4]; /* Stored in network byte order */ 3025 __u32 remote_port; /* Stored in network byte order */ 3026 __u32 local_port; /* stored in host byte order */ 3027 /* ... here. */ 3028 3029 __u32 data_meta; 3030 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3031 __u64 tstamp; 3032 __u32 wire_len; 3033 __u32 gso_segs; 3034 __bpf_md_ptr(struct bpf_sock *, sk); 3035 }; 3036 3037 struct bpf_tunnel_key { 3038 __u32 tunnel_id; 3039 union { 3040 __u32 remote_ipv4; 3041 __u32 remote_ipv6[4]; 3042 }; 3043 __u8 tunnel_tos; 3044 __u8 tunnel_ttl; 3045 __u16 tunnel_ext; /* Padding, future use. */ 3046 __u32 tunnel_label; 3047 }; 3048 3049 /* user accessible mirror of in-kernel xfrm_state. 3050 * new fields can only be added to the end of this structure 3051 */ 3052 struct bpf_xfrm_state { 3053 __u32 reqid; 3054 __u32 spi; /* Stored in network byte order */ 3055 __u16 family; 3056 __u16 ext; /* Padding, future use. */ 3057 union { 3058 __u32 remote_ipv4; /* Stored in network byte order */ 3059 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3060 }; 3061 }; 3062 3063 /* Generic BPF return codes which all BPF program types may support. 3064 * The values are binary compatible with their TC_ACT_* counter-part to 3065 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3066 * programs. 3067 * 3068 * XDP is handled seprately, see XDP_*. 3069 */ 3070 enum bpf_ret_code { 3071 BPF_OK = 0, 3072 /* 1 reserved */ 3073 BPF_DROP = 2, 3074 /* 3-6 reserved */ 3075 BPF_REDIRECT = 7, 3076 /* >127 are reserved for prog type specific return codes. 3077 * 3078 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3079 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3080 * changed and should be routed based on its new L3 header. 3081 * (This is an L3 redirect, as opposed to L2 redirect 3082 * represented by BPF_REDIRECT above). 3083 */ 3084 BPF_LWT_REROUTE = 128, 3085 }; 3086 3087 struct bpf_sock { 3088 __u32 bound_dev_if; 3089 __u32 family; 3090 __u32 type; 3091 __u32 protocol; 3092 __u32 mark; 3093 __u32 priority; 3094 /* IP address also allows 1 and 2 bytes access */ 3095 __u32 src_ip4; 3096 __u32 src_ip6[4]; 3097 __u32 src_port; /* host byte order */ 3098 __be16 dst_port; /* network byte order */ 3099 __u16 :16; /* zero padding */ 3100 __u32 dst_ip4; 3101 __u32 dst_ip6[4]; 3102 __u32 state; 3103 }; 3104 3105 struct bpf_tcp_sock { 3106 __u32 snd_cwnd; /* Sending congestion window */ 3107 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3108 __u32 rtt_min; 3109 __u32 snd_ssthresh; /* Slow start size threshold */ 3110 __u32 rcv_nxt; /* What we want to receive next */ 3111 __u32 snd_nxt; /* Next sequence we send */ 3112 __u32 snd_una; /* First byte we want an ack for */ 3113 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3114 __u32 ecn_flags; /* ECN status bits. */ 3115 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3116 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3117 __u32 packets_out; /* Packets which are "in flight" */ 3118 __u32 retrans_out; /* Retransmitted packets out */ 3119 __u32 total_retrans; /* Total retransmits for entire connection */ 3120 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3121 * total number of segments in. 3122 */ 3123 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3124 * total number of data segments in. 3125 */ 3126 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3127 * The total number of segments sent. 3128 */ 3129 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3130 * total number of data segments sent. 3131 */ 3132 __u32 lost_out; /* Lost packets */ 3133 __u32 sacked_out; /* SACK'd packets */ 3134 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3135 * sum(delta(rcv_nxt)), or how many bytes 3136 * were acked. 3137 */ 3138 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3139 * sum(delta(snd_una)), or how many bytes 3140 * were acked. 3141 */ 3142 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3143 * total number of DSACK blocks received 3144 */ 3145 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3146 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3147 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3148 }; 3149 3150 struct bpf_sock_tuple { 3151 union { 3152 struct { 3153 __be32 saddr; 3154 __be32 daddr; 3155 __be16 sport; 3156 __be16 dport; 3157 } ipv4; 3158 struct { 3159 __be32 saddr[4]; 3160 __be32 daddr[4]; 3161 __be16 sport; 3162 __be16 dport; 3163 } ipv6; 3164 }; 3165 }; 3166 3167 struct bpf_xdp_sock { 3168 __u32 queue_id; 3169 }; 3170 3171 #define XDP_PACKET_HEADROOM 256 3172 3173 /* User return codes for XDP prog type. 3174 * A valid XDP program must return one of these defined values. All other 3175 * return codes are reserved for future use. Unknown return codes will 3176 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3177 */ 3178 enum xdp_action { 3179 XDP_ABORTED = 0, 3180 XDP_DROP, 3181 XDP_PASS, 3182 XDP_TX, 3183 XDP_REDIRECT, 3184 }; 3185 3186 /* user accessible metadata for XDP packet hook 3187 * new fields must be added to the end of this structure 3188 */ 3189 struct xdp_md { 3190 __u32 data; 3191 __u32 data_end; 3192 __u32 data_meta; 3193 /* Below access go through struct xdp_rxq_info */ 3194 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3195 __u32 rx_queue_index; /* rxq->queue_index */ 3196 }; 3197 3198 enum sk_action { 3199 SK_DROP = 0, 3200 SK_PASS, 3201 }; 3202 3203 /* user accessible metadata for SK_MSG packet hook, new fields must 3204 * be added to the end of this structure 3205 */ 3206 struct sk_msg_md { 3207 __bpf_md_ptr(void *, data); 3208 __bpf_md_ptr(void *, data_end); 3209 3210 __u32 family; 3211 __u32 remote_ip4; /* Stored in network byte order */ 3212 __u32 local_ip4; /* Stored in network byte order */ 3213 __u32 remote_ip6[4]; /* Stored in network byte order */ 3214 __u32 local_ip6[4]; /* Stored in network byte order */ 3215 __u32 remote_port; /* Stored in network byte order */ 3216 __u32 local_port; /* stored in host byte order */ 3217 __u32 size; /* Total size of sk_msg */ 3218 }; 3219 3220 struct sk_reuseport_md { 3221 /* 3222 * Start of directly accessible data. It begins from 3223 * the tcp/udp header. 3224 */ 3225 __bpf_md_ptr(void *, data); 3226 /* End of directly accessible data */ 3227 __bpf_md_ptr(void *, data_end); 3228 /* 3229 * Total length of packet (starting from the tcp/udp header). 3230 * Note that the directly accessible bytes (data_end - data) 3231 * could be less than this "len". Those bytes could be 3232 * indirectly read by a helper "bpf_skb_load_bytes()". 3233 */ 3234 __u32 len; 3235 /* 3236 * Eth protocol in the mac header (network byte order). e.g. 3237 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3238 */ 3239 __u32 eth_protocol; 3240 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3241 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3242 __u32 hash; /* A hash of the packet 4 tuples */ 3243 }; 3244 3245 #define BPF_TAG_SIZE 8 3246 3247 struct bpf_prog_info { 3248 __u32 type; 3249 __u32 id; 3250 __u8 tag[BPF_TAG_SIZE]; 3251 __u32 jited_prog_len; 3252 __u32 xlated_prog_len; 3253 __aligned_u64 jited_prog_insns; 3254 __aligned_u64 xlated_prog_insns; 3255 __u64 load_time; /* ns since boottime */ 3256 __u32 created_by_uid; 3257 __u32 nr_map_ids; 3258 __aligned_u64 map_ids; 3259 char name[BPF_OBJ_NAME_LEN]; 3260 __u32 ifindex; 3261 __u32 gpl_compatible:1; 3262 __u32 :31; /* alignment pad */ 3263 __u64 netns_dev; 3264 __u64 netns_ino; 3265 __u32 nr_jited_ksyms; 3266 __u32 nr_jited_func_lens; 3267 __aligned_u64 jited_ksyms; 3268 __aligned_u64 jited_func_lens; 3269 __u32 btf_id; 3270 __u32 func_info_rec_size; 3271 __aligned_u64 func_info; 3272 __u32 nr_func_info; 3273 __u32 nr_line_info; 3274 __aligned_u64 line_info; 3275 __aligned_u64 jited_line_info; 3276 __u32 nr_jited_line_info; 3277 __u32 line_info_rec_size; 3278 __u32 jited_line_info_rec_size; 3279 __u32 nr_prog_tags; 3280 __aligned_u64 prog_tags; 3281 __u64 run_time_ns; 3282 __u64 run_cnt; 3283 } __attribute__((aligned(8))); 3284 3285 struct bpf_map_info { 3286 __u32 type; 3287 __u32 id; 3288 __u32 key_size; 3289 __u32 value_size; 3290 __u32 max_entries; 3291 __u32 map_flags; 3292 char name[BPF_OBJ_NAME_LEN]; 3293 __u32 ifindex; 3294 __u32 :32; 3295 __u64 netns_dev; 3296 __u64 netns_ino; 3297 __u32 btf_id; 3298 __u32 btf_key_type_id; 3299 __u32 btf_value_type_id; 3300 } __attribute__((aligned(8))); 3301 3302 struct bpf_btf_info { 3303 __aligned_u64 btf; 3304 __u32 btf_size; 3305 __u32 id; 3306 } __attribute__((aligned(8))); 3307 3308 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3309 * by user and intended to be used by socket (e.g. to bind to, depends on 3310 * attach attach type). 3311 */ 3312 struct bpf_sock_addr { 3313 __u32 user_family; /* Allows 4-byte read, but no write. */ 3314 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3315 * Stored in network byte order. 3316 */ 3317 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3318 * Stored in network byte order. 3319 */ 3320 __u32 user_port; /* Allows 4-byte read and write. 3321 * Stored in network byte order 3322 */ 3323 __u32 family; /* Allows 4-byte read, but no write */ 3324 __u32 type; /* Allows 4-byte read, but no write */ 3325 __u32 protocol; /* Allows 4-byte read, but no write */ 3326 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3327 * Stored in network byte order. 3328 */ 3329 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3330 * Stored in network byte order. 3331 */ 3332 __bpf_md_ptr(struct bpf_sock *, sk); 3333 }; 3334 3335 /* User bpf_sock_ops struct to access socket values and specify request ops 3336 * and their replies. 3337 * Some of this fields are in network (bigendian) byte order and may need 3338 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3339 * New fields can only be added at the end of this structure 3340 */ 3341 struct bpf_sock_ops { 3342 __u32 op; 3343 union { 3344 __u32 args[4]; /* Optionally passed to bpf program */ 3345 __u32 reply; /* Returned by bpf program */ 3346 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3347 }; 3348 __u32 family; 3349 __u32 remote_ip4; /* Stored in network byte order */ 3350 __u32 local_ip4; /* Stored in network byte order */ 3351 __u32 remote_ip6[4]; /* Stored in network byte order */ 3352 __u32 local_ip6[4]; /* Stored in network byte order */ 3353 __u32 remote_port; /* Stored in network byte order */ 3354 __u32 local_port; /* stored in host byte order */ 3355 __u32 is_fullsock; /* Some TCP fields are only valid if 3356 * there is a full socket. If not, the 3357 * fields read as zero. 3358 */ 3359 __u32 snd_cwnd; 3360 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3361 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3362 __u32 state; 3363 __u32 rtt_min; 3364 __u32 snd_ssthresh; 3365 __u32 rcv_nxt; 3366 __u32 snd_nxt; 3367 __u32 snd_una; 3368 __u32 mss_cache; 3369 __u32 ecn_flags; 3370 __u32 rate_delivered; 3371 __u32 rate_interval_us; 3372 __u32 packets_out; 3373 __u32 retrans_out; 3374 __u32 total_retrans; 3375 __u32 segs_in; 3376 __u32 data_segs_in; 3377 __u32 segs_out; 3378 __u32 data_segs_out; 3379 __u32 lost_out; 3380 __u32 sacked_out; 3381 __u32 sk_txhash; 3382 __u64 bytes_received; 3383 __u64 bytes_acked; 3384 __bpf_md_ptr(struct bpf_sock *, sk); 3385 }; 3386 3387 /* Definitions for bpf_sock_ops_cb_flags */ 3388 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3389 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3390 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3391 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3) 3392 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently 3393 * supported cb flags 3394 */ 3395 3396 /* List of known BPF sock_ops operators. 3397 * New entries can only be added at the end 3398 */ 3399 enum { 3400 BPF_SOCK_OPS_VOID, 3401 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3402 * -1 if default value should be used 3403 */ 3404 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3405 * window (in packets) or -1 if default 3406 * value should be used 3407 */ 3408 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3409 * active connection is initialized 3410 */ 3411 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3412 * active connection is 3413 * established 3414 */ 3415 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3416 * passive connection is 3417 * established 3418 */ 3419 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3420 * needs ECN 3421 */ 3422 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3423 * based on the path and may be 3424 * dependent on the congestion control 3425 * algorithm. In general it indicates 3426 * a congestion threshold. RTTs above 3427 * this indicate congestion 3428 */ 3429 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3430 * Arg1: value of icsk_retransmits 3431 * Arg2: value of icsk_rto 3432 * Arg3: whether RTO has expired 3433 */ 3434 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3435 * Arg1: sequence number of 1st byte 3436 * Arg2: # segments 3437 * Arg3: return value of 3438 * tcp_transmit_skb (0 => success) 3439 */ 3440 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3441 * Arg1: old_state 3442 * Arg2: new_state 3443 */ 3444 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3445 * socket transition to LISTEN state. 3446 */ 3447 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 3448 */ 3449 }; 3450 3451 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3452 * changes between the TCP and BPF versions. Ideally this should never happen. 3453 * If it does, we need to add code to convert them before calling 3454 * the BPF sock_ops function. 3455 */ 3456 enum { 3457 BPF_TCP_ESTABLISHED = 1, 3458 BPF_TCP_SYN_SENT, 3459 BPF_TCP_SYN_RECV, 3460 BPF_TCP_FIN_WAIT1, 3461 BPF_TCP_FIN_WAIT2, 3462 BPF_TCP_TIME_WAIT, 3463 BPF_TCP_CLOSE, 3464 BPF_TCP_CLOSE_WAIT, 3465 BPF_TCP_LAST_ACK, 3466 BPF_TCP_LISTEN, 3467 BPF_TCP_CLOSING, /* Now a valid state */ 3468 BPF_TCP_NEW_SYN_RECV, 3469 3470 BPF_TCP_MAX_STATES /* Leave at the end! */ 3471 }; 3472 3473 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3474 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3475 3476 struct bpf_perf_event_value { 3477 __u64 counter; 3478 __u64 enabled; 3479 __u64 running; 3480 }; 3481 3482 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3483 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3484 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3485 3486 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3487 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3488 3489 struct bpf_cgroup_dev_ctx { 3490 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3491 __u32 access_type; 3492 __u32 major; 3493 __u32 minor; 3494 }; 3495 3496 struct bpf_raw_tracepoint_args { 3497 __u64 args[0]; 3498 }; 3499 3500 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3501 * OUTPUT: Do lookup from egress perspective; default is ingress 3502 */ 3503 #define BPF_FIB_LOOKUP_DIRECT (1U << 0) 3504 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1) 3505 3506 enum { 3507 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3508 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3509 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3510 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3511 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3512 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3513 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3514 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3515 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3516 }; 3517 3518 struct bpf_fib_lookup { 3519 /* input: network family for lookup (AF_INET, AF_INET6) 3520 * output: network family of egress nexthop 3521 */ 3522 __u8 family; 3523 3524 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3525 __u8 l4_protocol; 3526 __be16 sport; 3527 __be16 dport; 3528 3529 /* total length of packet from network header - used for MTU check */ 3530 __u16 tot_len; 3531 3532 /* input: L3 device index for lookup 3533 * output: device index from FIB lookup 3534 */ 3535 __u32 ifindex; 3536 3537 union { 3538 /* inputs to lookup */ 3539 __u8 tos; /* AF_INET */ 3540 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3541 3542 /* output: metric of fib result (IPv4/IPv6 only) */ 3543 __u32 rt_metric; 3544 }; 3545 3546 union { 3547 __be32 ipv4_src; 3548 __u32 ipv6_src[4]; /* in6_addr; network order */ 3549 }; 3550 3551 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3552 * network header. output: bpf_fib_lookup sets to gateway address 3553 * if FIB lookup returns gateway route 3554 */ 3555 union { 3556 __be32 ipv4_dst; 3557 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3558 }; 3559 3560 /* output */ 3561 __be16 h_vlan_proto; 3562 __be16 h_vlan_TCI; 3563 __u8 smac[6]; /* ETH_ALEN */ 3564 __u8 dmac[6]; /* ETH_ALEN */ 3565 }; 3566 3567 enum bpf_task_fd_type { 3568 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3569 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3570 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3571 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3572 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3573 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3574 }; 3575 3576 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0) 3577 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1) 3578 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2) 3579 3580 struct bpf_flow_keys { 3581 __u16 nhoff; 3582 __u16 thoff; 3583 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3584 __u8 is_frag; 3585 __u8 is_first_frag; 3586 __u8 is_encap; 3587 __u8 ip_proto; 3588 __be16 n_proto; 3589 __be16 sport; 3590 __be16 dport; 3591 union { 3592 struct { 3593 __be32 ipv4_src; 3594 __be32 ipv4_dst; 3595 }; 3596 struct { 3597 __u32 ipv6_src[4]; /* in6_addr; network order */ 3598 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3599 }; 3600 }; 3601 __u32 flags; 3602 __be32 flow_label; 3603 }; 3604 3605 struct bpf_func_info { 3606 __u32 insn_off; 3607 __u32 type_id; 3608 }; 3609 3610 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3611 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3612 3613 struct bpf_line_info { 3614 __u32 insn_off; 3615 __u32 file_name_off; 3616 __u32 line_off; 3617 __u32 line_col; 3618 }; 3619 3620 struct bpf_spin_lock { 3621 __u32 val; 3622 }; 3623 3624 struct bpf_sysctl { 3625 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3626 * Allows 1,2,4-byte read, but no write. 3627 */ 3628 __u32 file_pos; /* Sysctl file position to read from, write to. 3629 * Allows 1,2,4-byte read an 4-byte write. 3630 */ 3631 }; 3632 3633 struct bpf_sockopt { 3634 __bpf_md_ptr(struct bpf_sock *, sk); 3635 __bpf_md_ptr(void *, optval); 3636 __bpf_md_ptr(void *, optval_end); 3637 3638 __s32 level; 3639 __s32 optname; 3640 __s32 optlen; 3641 __s32 retval; 3642 }; 3643 3644 #endif /* _UAPI__LINUX_BPF_H__ */ 3645