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 BPF_PROG_TYPE_TRACING, 177 }; 178 179 enum bpf_attach_type { 180 BPF_CGROUP_INET_INGRESS, 181 BPF_CGROUP_INET_EGRESS, 182 BPF_CGROUP_INET_SOCK_CREATE, 183 BPF_CGROUP_SOCK_OPS, 184 BPF_SK_SKB_STREAM_PARSER, 185 BPF_SK_SKB_STREAM_VERDICT, 186 BPF_CGROUP_DEVICE, 187 BPF_SK_MSG_VERDICT, 188 BPF_CGROUP_INET4_BIND, 189 BPF_CGROUP_INET6_BIND, 190 BPF_CGROUP_INET4_CONNECT, 191 BPF_CGROUP_INET6_CONNECT, 192 BPF_CGROUP_INET4_POST_BIND, 193 BPF_CGROUP_INET6_POST_BIND, 194 BPF_CGROUP_UDP4_SENDMSG, 195 BPF_CGROUP_UDP6_SENDMSG, 196 BPF_LIRC_MODE2, 197 BPF_FLOW_DISSECTOR, 198 BPF_CGROUP_SYSCTL, 199 BPF_CGROUP_UDP4_RECVMSG, 200 BPF_CGROUP_UDP6_RECVMSG, 201 BPF_CGROUP_GETSOCKOPT, 202 BPF_CGROUP_SETSOCKOPT, 203 BPF_TRACE_RAW_TP, 204 __MAX_BPF_ATTACH_TYPE 205 }; 206 207 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 208 209 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 210 * 211 * NONE(default): No further bpf programs allowed in the subtree. 212 * 213 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 214 * the program in this cgroup yields to sub-cgroup program. 215 * 216 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 217 * that cgroup program gets run in addition to the program in this cgroup. 218 * 219 * Only one program is allowed to be attached to a cgroup with 220 * NONE or BPF_F_ALLOW_OVERRIDE flag. 221 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 222 * release old program and attach the new one. Attach flags has to match. 223 * 224 * Multiple programs are allowed to be attached to a cgroup with 225 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 226 * (those that were attached first, run first) 227 * The programs of sub-cgroup are executed first, then programs of 228 * this cgroup and then programs of parent cgroup. 229 * When children program makes decision (like picking TCP CA or sock bind) 230 * parent program has a chance to override it. 231 * 232 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 233 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 234 * Ex1: 235 * cgrp1 (MULTI progs A, B) -> 236 * cgrp2 (OVERRIDE prog C) -> 237 * cgrp3 (MULTI prog D) -> 238 * cgrp4 (OVERRIDE prog E) -> 239 * cgrp5 (NONE prog F) 240 * the event in cgrp5 triggers execution of F,D,A,B in that order. 241 * if prog F is detached, the execution is E,D,A,B 242 * if prog F and D are detached, the execution is E,A,B 243 * if prog F, E and D are detached, the execution is C,A,B 244 * 245 * All eligible programs are executed regardless of return code from 246 * earlier programs. 247 */ 248 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 249 #define BPF_F_ALLOW_MULTI (1U << 1) 250 251 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 252 * verifier will perform strict alignment checking as if the kernel 253 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 254 * and NET_IP_ALIGN defined to 2. 255 */ 256 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 257 258 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 259 * verifier will allow any alignment whatsoever. On platforms 260 * with strict alignment requirements for loads ands stores (such 261 * as sparc and mips) the verifier validates that all loads and 262 * stores provably follow this requirement. This flag turns that 263 * checking and enforcement off. 264 * 265 * It is mostly used for testing when we want to validate the 266 * context and memory access aspects of the verifier, but because 267 * of an unaligned access the alignment check would trigger before 268 * the one we are interested in. 269 */ 270 #define BPF_F_ANY_ALIGNMENT (1U << 1) 271 272 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 273 * Verifier does sub-register def/use analysis and identifies instructions whose 274 * def only matters for low 32-bit, high 32-bit is never referenced later 275 * through implicit zero extension. Therefore verifier notifies JIT back-ends 276 * that it is safe to ignore clearing high 32-bit for these instructions. This 277 * saves some back-ends a lot of code-gen. However such optimization is not 278 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 279 * hence hasn't used verifier's analysis result. But, we really want to have a 280 * way to be able to verify the correctness of the described optimization on 281 * x86_64 on which testsuites are frequently exercised. 282 * 283 * So, this flag is introduced. Once it is set, verifier will randomize high 284 * 32-bit for those instructions who has been identified as safe to ignore them. 285 * Then, if verifier is not doing correct analysis, such randomization will 286 * regress tests to expose bugs. 287 */ 288 #define BPF_F_TEST_RND_HI32 (1U << 2) 289 290 /* The verifier internal test flag. Behavior is undefined */ 291 #define BPF_F_TEST_STATE_FREQ (1U << 3) 292 293 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 294 * two extensions: 295 * 296 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 297 * insn[0].imm: map fd map fd 298 * insn[1].imm: 0 offset into value 299 * insn[0].off: 0 0 300 * insn[1].off: 0 0 301 * ldimm64 rewrite: address of map address of map[0]+offset 302 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 303 */ 304 #define BPF_PSEUDO_MAP_FD 1 305 #define BPF_PSEUDO_MAP_VALUE 2 306 307 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 308 * offset to another bpf function 309 */ 310 #define BPF_PSEUDO_CALL 1 311 312 /* flags for BPF_MAP_UPDATE_ELEM command */ 313 #define BPF_ANY 0 /* create new element or update existing */ 314 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 315 #define BPF_EXIST 2 /* update existing element */ 316 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 317 318 /* flags for BPF_MAP_CREATE command */ 319 #define BPF_F_NO_PREALLOC (1U << 0) 320 /* Instead of having one common LRU list in the 321 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 322 * which can scale and perform better. 323 * Note, the LRU nodes (including free nodes) cannot be moved 324 * across different LRU lists. 325 */ 326 #define BPF_F_NO_COMMON_LRU (1U << 1) 327 /* Specify numa node during map creation */ 328 #define BPF_F_NUMA_NODE (1U << 2) 329 330 #define BPF_OBJ_NAME_LEN 16U 331 332 /* Flags for accessing BPF object from syscall side. */ 333 #define BPF_F_RDONLY (1U << 3) 334 #define BPF_F_WRONLY (1U << 4) 335 336 /* Flag for stack_map, store build_id+offset instead of pointer */ 337 #define BPF_F_STACK_BUILD_ID (1U << 5) 338 339 /* Zero-initialize hash function seed. This should only be used for testing. */ 340 #define BPF_F_ZERO_SEED (1U << 6) 341 342 /* Flags for accessing BPF object from program side. */ 343 #define BPF_F_RDONLY_PROG (1U << 7) 344 #define BPF_F_WRONLY_PROG (1U << 8) 345 346 /* Clone map from listener for newly accepted socket */ 347 #define BPF_F_CLONE (1U << 9) 348 349 /* flags for BPF_PROG_QUERY */ 350 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 351 352 enum bpf_stack_build_id_status { 353 /* user space need an empty entry to identify end of a trace */ 354 BPF_STACK_BUILD_ID_EMPTY = 0, 355 /* with valid build_id and offset */ 356 BPF_STACK_BUILD_ID_VALID = 1, 357 /* couldn't get build_id, fallback to ip */ 358 BPF_STACK_BUILD_ID_IP = 2, 359 }; 360 361 #define BPF_BUILD_ID_SIZE 20 362 struct bpf_stack_build_id { 363 __s32 status; 364 unsigned char build_id[BPF_BUILD_ID_SIZE]; 365 union { 366 __u64 offset; 367 __u64 ip; 368 }; 369 }; 370 371 union bpf_attr { 372 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 373 __u32 map_type; /* one of enum bpf_map_type */ 374 __u32 key_size; /* size of key in bytes */ 375 __u32 value_size; /* size of value in bytes */ 376 __u32 max_entries; /* max number of entries in a map */ 377 __u32 map_flags; /* BPF_MAP_CREATE related 378 * flags defined above. 379 */ 380 __u32 inner_map_fd; /* fd pointing to the inner map */ 381 __u32 numa_node; /* numa node (effective only if 382 * BPF_F_NUMA_NODE is set). 383 */ 384 char map_name[BPF_OBJ_NAME_LEN]; 385 __u32 map_ifindex; /* ifindex of netdev to create on */ 386 __u32 btf_fd; /* fd pointing to a BTF type data */ 387 __u32 btf_key_type_id; /* BTF type_id of the key */ 388 __u32 btf_value_type_id; /* BTF type_id of the value */ 389 }; 390 391 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 392 __u32 map_fd; 393 __aligned_u64 key; 394 union { 395 __aligned_u64 value; 396 __aligned_u64 next_key; 397 }; 398 __u64 flags; 399 }; 400 401 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 402 __u32 prog_type; /* one of enum bpf_prog_type */ 403 __u32 insn_cnt; 404 __aligned_u64 insns; 405 __aligned_u64 license; 406 __u32 log_level; /* verbosity level of verifier */ 407 __u32 log_size; /* size of user buffer */ 408 __aligned_u64 log_buf; /* user supplied buffer */ 409 __u32 kern_version; /* not used */ 410 __u32 prog_flags; 411 char prog_name[BPF_OBJ_NAME_LEN]; 412 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 413 /* For some prog types expected attach type must be known at 414 * load time to verify attach type specific parts of prog 415 * (context accesses, allowed helpers, etc). 416 */ 417 __u32 expected_attach_type; 418 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 419 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 420 __aligned_u64 func_info; /* func info */ 421 __u32 func_info_cnt; /* number of bpf_func_info records */ 422 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 423 __aligned_u64 line_info; /* line info */ 424 __u32 line_info_cnt; /* number of bpf_line_info records */ 425 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 426 }; 427 428 struct { /* anonymous struct used by BPF_OBJ_* commands */ 429 __aligned_u64 pathname; 430 __u32 bpf_fd; 431 __u32 file_flags; 432 }; 433 434 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 435 __u32 target_fd; /* container object to attach to */ 436 __u32 attach_bpf_fd; /* eBPF program to attach */ 437 __u32 attach_type; 438 __u32 attach_flags; 439 }; 440 441 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 442 __u32 prog_fd; 443 __u32 retval; 444 __u32 data_size_in; /* input: len of data_in */ 445 __u32 data_size_out; /* input/output: len of data_out 446 * returns ENOSPC if data_out 447 * is too small. 448 */ 449 __aligned_u64 data_in; 450 __aligned_u64 data_out; 451 __u32 repeat; 452 __u32 duration; 453 __u32 ctx_size_in; /* input: len of ctx_in */ 454 __u32 ctx_size_out; /* input/output: len of ctx_out 455 * returns ENOSPC if ctx_out 456 * is too small. 457 */ 458 __aligned_u64 ctx_in; 459 __aligned_u64 ctx_out; 460 } test; 461 462 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 463 union { 464 __u32 start_id; 465 __u32 prog_id; 466 __u32 map_id; 467 __u32 btf_id; 468 }; 469 __u32 next_id; 470 __u32 open_flags; 471 }; 472 473 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 474 __u32 bpf_fd; 475 __u32 info_len; 476 __aligned_u64 info; 477 } info; 478 479 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 480 __u32 target_fd; /* container object to query */ 481 __u32 attach_type; 482 __u32 query_flags; 483 __u32 attach_flags; 484 __aligned_u64 prog_ids; 485 __u32 prog_cnt; 486 } query; 487 488 struct { 489 __u64 name; 490 __u32 prog_fd; 491 } raw_tracepoint; 492 493 struct { /* anonymous struct for BPF_BTF_LOAD */ 494 __aligned_u64 btf; 495 __aligned_u64 btf_log_buf; 496 __u32 btf_size; 497 __u32 btf_log_size; 498 __u32 btf_log_level; 499 }; 500 501 struct { 502 __u32 pid; /* input: pid */ 503 __u32 fd; /* input: fd */ 504 __u32 flags; /* input: flags */ 505 __u32 buf_len; /* input/output: buf len */ 506 __aligned_u64 buf; /* input/output: 507 * tp_name for tracepoint 508 * symbol for kprobe 509 * filename for uprobe 510 */ 511 __u32 prog_id; /* output: prod_id */ 512 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 513 __u64 probe_offset; /* output: probe_offset */ 514 __u64 probe_addr; /* output: probe_addr */ 515 } task_fd_query; 516 } __attribute__((aligned(8))); 517 518 /* The description below is an attempt at providing documentation to eBPF 519 * developers about the multiple available eBPF helper functions. It can be 520 * parsed and used to produce a manual page. The workflow is the following, 521 * and requires the rst2man utility: 522 * 523 * $ ./scripts/bpf_helpers_doc.py \ 524 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 525 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 526 * $ man /tmp/bpf-helpers.7 527 * 528 * Note that in order to produce this external documentation, some RST 529 * formatting is used in the descriptions to get "bold" and "italics" in 530 * manual pages. Also note that the few trailing white spaces are 531 * intentional, removing them would break paragraphs for rst2man. 532 * 533 * Start of BPF helper function descriptions: 534 * 535 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 536 * Description 537 * Perform a lookup in *map* for an entry associated to *key*. 538 * Return 539 * Map value associated to *key*, or **NULL** if no entry was 540 * found. 541 * 542 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 543 * Description 544 * Add or update the value of the entry associated to *key* in 545 * *map* with *value*. *flags* is one of: 546 * 547 * **BPF_NOEXIST** 548 * The entry for *key* must not exist in the map. 549 * **BPF_EXIST** 550 * The entry for *key* must already exist in the map. 551 * **BPF_ANY** 552 * No condition on the existence of the entry for *key*. 553 * 554 * Flag value **BPF_NOEXIST** cannot be used for maps of types 555 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 556 * elements always exist), the helper would return an error. 557 * Return 558 * 0 on success, or a negative error in case of failure. 559 * 560 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 561 * Description 562 * Delete entry with *key* from *map*. 563 * Return 564 * 0 on success, or a negative error in case of failure. 565 * 566 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 567 * Description 568 * For tracing programs, safely attempt to read *size* bytes from 569 * kernel space address *unsafe_ptr* and store the data in *dst*. 570 * 571 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel() 572 * instead. 573 * Return 574 * 0 on success, or a negative error in case of failure. 575 * 576 * u64 bpf_ktime_get_ns(void) 577 * Description 578 * Return the time elapsed since system boot, in nanoseconds. 579 * Return 580 * Current *ktime*. 581 * 582 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 583 * Description 584 * This helper is a "printk()-like" facility for debugging. It 585 * prints a message defined by format *fmt* (of size *fmt_size*) 586 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 587 * available. It can take up to three additional **u64** 588 * arguments (as an eBPF helpers, the total number of arguments is 589 * limited to five). 590 * 591 * Each time the helper is called, it appends a line to the trace. 592 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 593 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 594 * The format of the trace is customizable, and the exact output 595 * one will get depends on the options set in 596 * *\/sys/kernel/debug/tracing/trace_options* (see also the 597 * *README* file under the same directory). However, it usually 598 * defaults to something like: 599 * 600 * :: 601 * 602 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 603 * 604 * In the above: 605 * 606 * * ``telnet`` is the name of the current task. 607 * * ``470`` is the PID of the current task. 608 * * ``001`` is the CPU number on which the task is 609 * running. 610 * * In ``.N..``, each character refers to a set of 611 * options (whether irqs are enabled, scheduling 612 * options, whether hard/softirqs are running, level of 613 * preempt_disabled respectively). **N** means that 614 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 615 * are set. 616 * * ``419421.045894`` is a timestamp. 617 * * ``0x00000001`` is a fake value used by BPF for the 618 * instruction pointer register. 619 * * ``<formatted msg>`` is the message formatted with 620 * *fmt*. 621 * 622 * The conversion specifiers supported by *fmt* are similar, but 623 * more limited than for printk(). They are **%d**, **%i**, 624 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 625 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 626 * of field, padding with zeroes, etc.) is available, and the 627 * helper will return **-EINVAL** (but print nothing) if it 628 * encounters an unknown specifier. 629 * 630 * Also, note that **bpf_trace_printk**\ () is slow, and should 631 * only be used for debugging purposes. For this reason, a notice 632 * bloc (spanning several lines) is printed to kernel logs and 633 * states that the helper should not be used "for production use" 634 * the first time this helper is used (or more precisely, when 635 * **trace_printk**\ () buffers are allocated). For passing values 636 * to user space, perf events should be preferred. 637 * Return 638 * The number of bytes written to the buffer, or a negative error 639 * in case of failure. 640 * 641 * u32 bpf_get_prandom_u32(void) 642 * Description 643 * Get a pseudo-random number. 644 * 645 * From a security point of view, this helper uses its own 646 * pseudo-random internal state, and cannot be used to infer the 647 * seed of other random functions in the kernel. However, it is 648 * essential to note that the generator used by the helper is not 649 * cryptographically secure. 650 * Return 651 * A random 32-bit unsigned value. 652 * 653 * u32 bpf_get_smp_processor_id(void) 654 * Description 655 * Get the SMP (symmetric multiprocessing) processor id. Note that 656 * all programs run with preemption disabled, which means that the 657 * SMP processor id is stable during all the execution of the 658 * program. 659 * Return 660 * The SMP id of the processor running the program. 661 * 662 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 663 * Description 664 * Store *len* bytes from address *from* into the packet 665 * associated to *skb*, at *offset*. *flags* are a combination of 666 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 667 * checksum for the packet after storing the bytes) and 668 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 669 * **->swhash** and *skb*\ **->l4hash** to 0). 670 * 671 * A call to this helper is susceptible to change the underlying 672 * packet buffer. Therefore, at load time, all checks on pointers 673 * previously done by the verifier are invalidated and must be 674 * performed again, if the helper is used in combination with 675 * direct packet access. 676 * Return 677 * 0 on success, or a negative error in case of failure. 678 * 679 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 680 * Description 681 * Recompute the layer 3 (e.g. IP) checksum for the packet 682 * associated to *skb*. Computation is incremental, so the helper 683 * must know the former value of the header field that was 684 * modified (*from*), the new value of this field (*to*), and the 685 * number of bytes (2 or 4) for this field, stored in *size*. 686 * Alternatively, it is possible to store the difference between 687 * the previous and the new values of the header field in *to*, by 688 * setting *from* and *size* to 0. For both methods, *offset* 689 * indicates the location of the IP checksum within the packet. 690 * 691 * This helper works in combination with **bpf_csum_diff**\ (), 692 * which does not update the checksum in-place, but offers more 693 * flexibility and can handle sizes larger than 2 or 4 for the 694 * checksum to update. 695 * 696 * A call to this helper is susceptible to change the underlying 697 * packet buffer. Therefore, at load time, all checks on pointers 698 * previously done by the verifier are invalidated and must be 699 * performed again, if the helper is used in combination with 700 * direct packet access. 701 * Return 702 * 0 on success, or a negative error in case of failure. 703 * 704 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 705 * Description 706 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 707 * packet associated to *skb*. Computation is incremental, so the 708 * helper must know the former value of the header field that was 709 * modified (*from*), the new value of this field (*to*), and the 710 * number of bytes (2 or 4) for this field, stored on the lowest 711 * four bits of *flags*. Alternatively, it is possible to store 712 * the difference between the previous and the new values of the 713 * header field in *to*, by setting *from* and the four lowest 714 * bits of *flags* to 0. For both methods, *offset* indicates the 715 * location of the IP checksum within the packet. In addition to 716 * the size of the field, *flags* can be added (bitwise OR) actual 717 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 718 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 719 * for updates resulting in a null checksum the value is set to 720 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 721 * the checksum is to be computed against a pseudo-header. 722 * 723 * This helper works in combination with **bpf_csum_diff**\ (), 724 * which does not update the checksum in-place, but offers more 725 * flexibility and can handle sizes larger than 2 or 4 for the 726 * checksum to update. 727 * 728 * A call to this helper is susceptible to change the underlying 729 * packet buffer. Therefore, at load time, all checks on pointers 730 * previously done by the verifier are invalidated and must be 731 * performed again, if the helper is used in combination with 732 * direct packet access. 733 * Return 734 * 0 on success, or a negative error in case of failure. 735 * 736 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 737 * Description 738 * This special helper is used to trigger a "tail call", or in 739 * other words, to jump into another eBPF program. The same stack 740 * frame is used (but values on stack and in registers for the 741 * caller are not accessible to the callee). This mechanism allows 742 * for program chaining, either for raising the maximum number of 743 * available eBPF instructions, or to execute given programs in 744 * conditional blocks. For security reasons, there is an upper 745 * limit to the number of successive tail calls that can be 746 * performed. 747 * 748 * Upon call of this helper, the program attempts to jump into a 749 * program referenced at index *index* in *prog_array_map*, a 750 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 751 * *ctx*, a pointer to the context. 752 * 753 * If the call succeeds, the kernel immediately runs the first 754 * instruction of the new program. This is not a function call, 755 * and it never returns to the previous program. If the call 756 * fails, then the helper has no effect, and the caller continues 757 * to run its subsequent instructions. A call can fail if the 758 * destination program for the jump does not exist (i.e. *index* 759 * is superior to the number of entries in *prog_array_map*), or 760 * if the maximum number of tail calls has been reached for this 761 * chain of programs. This limit is defined in the kernel by the 762 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 763 * which is currently set to 32. 764 * Return 765 * 0 on success, or a negative error in case of failure. 766 * 767 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 768 * Description 769 * Clone and redirect the packet associated to *skb* to another 770 * net device of index *ifindex*. Both ingress and egress 771 * interfaces can be used for redirection. The **BPF_F_INGRESS** 772 * value in *flags* is used to make the distinction (ingress path 773 * is selected if the flag is present, egress path otherwise). 774 * This is the only flag supported for now. 775 * 776 * In comparison with **bpf_redirect**\ () helper, 777 * **bpf_clone_redirect**\ () has the associated cost of 778 * duplicating the packet buffer, but this can be executed out of 779 * the eBPF program. Conversely, **bpf_redirect**\ () is more 780 * efficient, but it is handled through an action code where the 781 * redirection happens only after the eBPF program has returned. 782 * 783 * A call to this helper is susceptible to change the underlying 784 * packet buffer. Therefore, at load time, all checks on pointers 785 * previously done by the verifier are invalidated and must be 786 * performed again, if the helper is used in combination with 787 * direct packet access. 788 * Return 789 * 0 on success, or a negative error in case of failure. 790 * 791 * u64 bpf_get_current_pid_tgid(void) 792 * Return 793 * A 64-bit integer containing the current tgid and pid, and 794 * created as such: 795 * *current_task*\ **->tgid << 32 \|** 796 * *current_task*\ **->pid**. 797 * 798 * u64 bpf_get_current_uid_gid(void) 799 * Return 800 * A 64-bit integer containing the current GID and UID, and 801 * created as such: *current_gid* **<< 32 \|** *current_uid*. 802 * 803 * int bpf_get_current_comm(void *buf, u32 size_of_buf) 804 * Description 805 * Copy the **comm** attribute of the current task into *buf* of 806 * *size_of_buf*. The **comm** attribute contains the name of 807 * the executable (excluding the path) for the current task. The 808 * *size_of_buf* must be strictly positive. On success, the 809 * helper makes sure that the *buf* is NUL-terminated. On failure, 810 * it is filled with zeroes. 811 * Return 812 * 0 on success, or a negative error in case of failure. 813 * 814 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 815 * Description 816 * Retrieve the classid for the current task, i.e. for the net_cls 817 * cgroup to which *skb* belongs. 818 * 819 * This helper can be used on TC egress path, but not on ingress. 820 * 821 * The net_cls cgroup provides an interface to tag network packets 822 * based on a user-provided identifier for all traffic coming from 823 * the tasks belonging to the related cgroup. See also the related 824 * kernel documentation, available from the Linux sources in file 825 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 826 * 827 * The Linux kernel has two versions for cgroups: there are 828 * cgroups v1 and cgroups v2. Both are available to users, who can 829 * use a mixture of them, but note that the net_cls cgroup is for 830 * cgroup v1 only. This makes it incompatible with BPF programs 831 * run on cgroups, which is a cgroup-v2-only feature (a socket can 832 * only hold data for one version of cgroups at a time). 833 * 834 * This helper is only available is the kernel was compiled with 835 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 836 * "**y**" or to "**m**". 837 * Return 838 * The classid, or 0 for the default unconfigured classid. 839 * 840 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 841 * Description 842 * Push a *vlan_tci* (VLAN tag control information) of protocol 843 * *vlan_proto* to the packet associated to *skb*, then update 844 * the checksum. Note that if *vlan_proto* is different from 845 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 846 * be **ETH_P_8021Q**. 847 * 848 * A call to this helper is susceptible to change the underlying 849 * packet buffer. Therefore, at load time, all checks on pointers 850 * previously done by the verifier are invalidated and must be 851 * performed again, if the helper is used in combination with 852 * direct packet access. 853 * Return 854 * 0 on success, or a negative error in case of failure. 855 * 856 * int bpf_skb_vlan_pop(struct sk_buff *skb) 857 * Description 858 * Pop a VLAN header from the packet associated to *skb*. 859 * 860 * A call to this helper is susceptible to change the underlying 861 * packet buffer. Therefore, at load time, all checks on pointers 862 * previously done by the verifier are invalidated and must be 863 * performed again, if the helper is used in combination with 864 * direct packet access. 865 * Return 866 * 0 on success, or a negative error in case of failure. 867 * 868 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 869 * Description 870 * Get tunnel metadata. This helper takes a pointer *key* to an 871 * empty **struct bpf_tunnel_key** of **size**, that will be 872 * filled with tunnel metadata for the packet associated to *skb*. 873 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 874 * indicates that the tunnel is based on IPv6 protocol instead of 875 * IPv4. 876 * 877 * The **struct bpf_tunnel_key** is an object that generalizes the 878 * principal parameters used by various tunneling protocols into a 879 * single struct. This way, it can be used to easily make a 880 * decision based on the contents of the encapsulation header, 881 * "summarized" in this struct. In particular, it holds the IP 882 * address of the remote end (IPv4 or IPv6, depending on the case) 883 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 884 * this struct exposes the *key*\ **->tunnel_id**, which is 885 * generally mapped to a VNI (Virtual Network Identifier), making 886 * it programmable together with the **bpf_skb_set_tunnel_key**\ 887 * () helper. 888 * 889 * Let's imagine that the following code is part of a program 890 * attached to the TC ingress interface, on one end of a GRE 891 * tunnel, and is supposed to filter out all messages coming from 892 * remote ends with IPv4 address other than 10.0.0.1: 893 * 894 * :: 895 * 896 * int ret; 897 * struct bpf_tunnel_key key = {}; 898 * 899 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 900 * if (ret < 0) 901 * return TC_ACT_SHOT; // drop packet 902 * 903 * if (key.remote_ipv4 != 0x0a000001) 904 * return TC_ACT_SHOT; // drop packet 905 * 906 * return TC_ACT_OK; // accept packet 907 * 908 * This interface can also be used with all encapsulation devices 909 * that can operate in "collect metadata" mode: instead of having 910 * one network device per specific configuration, the "collect 911 * metadata" mode only requires a single device where the 912 * configuration can be extracted from this helper. 913 * 914 * This can be used together with various tunnels such as VXLan, 915 * Geneve, GRE or IP in IP (IPIP). 916 * Return 917 * 0 on success, or a negative error in case of failure. 918 * 919 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 920 * Description 921 * Populate tunnel metadata for packet associated to *skb.* The 922 * tunnel metadata is set to the contents of *key*, of *size*. The 923 * *flags* can be set to a combination of the following values: 924 * 925 * **BPF_F_TUNINFO_IPV6** 926 * Indicate that the tunnel is based on IPv6 protocol 927 * instead of IPv4. 928 * **BPF_F_ZERO_CSUM_TX** 929 * For IPv4 packets, add a flag to tunnel metadata 930 * indicating that checksum computation should be skipped 931 * and checksum set to zeroes. 932 * **BPF_F_DONT_FRAGMENT** 933 * Add a flag to tunnel metadata indicating that the 934 * packet should not be fragmented. 935 * **BPF_F_SEQ_NUMBER** 936 * Add a flag to tunnel metadata indicating that a 937 * sequence number should be added to tunnel header before 938 * sending the packet. This flag was added for GRE 939 * encapsulation, but might be used with other protocols 940 * as well in the future. 941 * 942 * Here is a typical usage on the transmit path: 943 * 944 * :: 945 * 946 * struct bpf_tunnel_key key; 947 * populate key ... 948 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 949 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 950 * 951 * See also the description of the **bpf_skb_get_tunnel_key**\ () 952 * helper for additional information. 953 * Return 954 * 0 on success, or a negative error in case of failure. 955 * 956 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 957 * Description 958 * Read the value of a perf event counter. This helper relies on a 959 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 960 * the perf event counter is selected when *map* is updated with 961 * perf event file descriptors. The *map* is an array whose size 962 * is the number of available CPUs, and each cell contains a value 963 * relative to one CPU. The value to retrieve is indicated by 964 * *flags*, that contains the index of the CPU to look up, masked 965 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 966 * **BPF_F_CURRENT_CPU** to indicate that the value for the 967 * current CPU should be retrieved. 968 * 969 * Note that before Linux 4.13, only hardware perf event can be 970 * retrieved. 971 * 972 * Also, be aware that the newer helper 973 * **bpf_perf_event_read_value**\ () is recommended over 974 * **bpf_perf_event_read**\ () in general. The latter has some ABI 975 * quirks where error and counter value are used as a return code 976 * (which is wrong to do since ranges may overlap). This issue is 977 * fixed with **bpf_perf_event_read_value**\ (), which at the same 978 * time provides more features over the **bpf_perf_event_read**\ 979 * () interface. Please refer to the description of 980 * **bpf_perf_event_read_value**\ () for details. 981 * Return 982 * The value of the perf event counter read from the map, or a 983 * negative error code in case of failure. 984 * 985 * int bpf_redirect(u32 ifindex, u64 flags) 986 * Description 987 * Redirect the packet to another net device of index *ifindex*. 988 * This helper is somewhat similar to **bpf_clone_redirect**\ 989 * (), except that the packet is not cloned, which provides 990 * increased performance. 991 * 992 * Except for XDP, both ingress and egress interfaces can be used 993 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 994 * to make the distinction (ingress path is selected if the flag 995 * is present, egress path otherwise). Currently, XDP only 996 * supports redirection to the egress interface, and accepts no 997 * flag at all. 998 * 999 * The same effect can be attained with the more generic 1000 * **bpf_redirect_map**\ (), which requires specific maps to be 1001 * used but offers better performance. 1002 * Return 1003 * For XDP, the helper returns **XDP_REDIRECT** on success or 1004 * **XDP_ABORTED** on error. For other program types, the values 1005 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1006 * error. 1007 * 1008 * u32 bpf_get_route_realm(struct sk_buff *skb) 1009 * Description 1010 * Retrieve the realm or the route, that is to say the 1011 * **tclassid** field of the destination for the *skb*. The 1012 * indentifier retrieved is a user-provided tag, similar to the 1013 * one used with the net_cls cgroup (see description for 1014 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1015 * held by a route (a destination entry), not by a task. 1016 * 1017 * Retrieving this identifier works with the clsact TC egress hook 1018 * (see also **tc-bpf(8)**), or alternatively on conventional 1019 * classful egress qdiscs, but not on TC ingress path. In case of 1020 * clsact TC egress hook, this has the advantage that, internally, 1021 * the destination entry has not been dropped yet in the transmit 1022 * path. Therefore, the destination entry does not need to be 1023 * artificially held via **netif_keep_dst**\ () for a classful 1024 * qdisc until the *skb* is freed. 1025 * 1026 * This helper is available only if the kernel was compiled with 1027 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1028 * Return 1029 * The realm of the route for the packet associated to *skb*, or 0 1030 * if none was found. 1031 * 1032 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1033 * Description 1034 * Write raw *data* blob into a special BPF perf event held by 1035 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1036 * event must have the following attributes: **PERF_SAMPLE_RAW** 1037 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1038 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1039 * 1040 * The *flags* are used to indicate the index in *map* for which 1041 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1042 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1043 * to indicate that the index of the current CPU core should be 1044 * used. 1045 * 1046 * The value to write, of *size*, is passed through eBPF stack and 1047 * pointed by *data*. 1048 * 1049 * The context of the program *ctx* needs also be passed to the 1050 * helper. 1051 * 1052 * On user space, a program willing to read the values needs to 1053 * call **perf_event_open**\ () on the perf event (either for 1054 * one or for all CPUs) and to store the file descriptor into the 1055 * *map*. This must be done before the eBPF program can send data 1056 * into it. An example is available in file 1057 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1058 * tree (the eBPF program counterpart is in 1059 * *samples/bpf/trace_output_kern.c*). 1060 * 1061 * **bpf_perf_event_output**\ () achieves better performance 1062 * than **bpf_trace_printk**\ () for sharing data with user 1063 * space, and is much better suitable for streaming data from eBPF 1064 * programs. 1065 * 1066 * Note that this helper is not restricted to tracing use cases 1067 * and can be used with programs attached to TC or XDP as well, 1068 * where it allows for passing data to user space listeners. Data 1069 * can be: 1070 * 1071 * * Only custom structs, 1072 * * Only the packet payload, or 1073 * * A combination of both. 1074 * Return 1075 * 0 on success, or a negative error in case of failure. 1076 * 1077 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1078 * Description 1079 * This helper was provided as an easy way to load data from a 1080 * packet. It can be used to load *len* bytes from *offset* from 1081 * the packet associated to *skb*, into the buffer pointed by 1082 * *to*. 1083 * 1084 * Since Linux 4.7, usage of this helper has mostly been replaced 1085 * by "direct packet access", enabling packet data to be 1086 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1087 * pointing respectively to the first byte of packet data and to 1088 * the byte after the last byte of packet data. However, it 1089 * remains useful if one wishes to read large quantities of data 1090 * at once from a packet into the eBPF stack. 1091 * Return 1092 * 0 on success, or a negative error in case of failure. 1093 * 1094 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1095 * Description 1096 * Walk a user or a kernel stack and return its id. To achieve 1097 * this, the helper needs *ctx*, which is a pointer to the context 1098 * on which the tracing program is executed, and a pointer to a 1099 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1100 * 1101 * The last argument, *flags*, holds the number of stack frames to 1102 * skip (from 0 to 255), masked with 1103 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1104 * a combination of the following flags: 1105 * 1106 * **BPF_F_USER_STACK** 1107 * Collect a user space stack instead of a kernel stack. 1108 * **BPF_F_FAST_STACK_CMP** 1109 * Compare stacks by hash only. 1110 * **BPF_F_REUSE_STACKID** 1111 * If two different stacks hash into the same *stackid*, 1112 * discard the old one. 1113 * 1114 * The stack id retrieved is a 32 bit long integer handle which 1115 * can be further combined with other data (including other stack 1116 * ids) and used as a key into maps. This can be useful for 1117 * generating a variety of graphs (such as flame graphs or off-cpu 1118 * graphs). 1119 * 1120 * For walking a stack, this helper is an improvement over 1121 * **bpf_probe_read**\ (), which can be used with unrolled loops 1122 * but is not efficient and consumes a lot of eBPF instructions. 1123 * Instead, **bpf_get_stackid**\ () can collect up to 1124 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1125 * this limit can be controlled with the **sysctl** program, and 1126 * that it should be manually increased in order to profile long 1127 * user stacks (such as stacks for Java programs). To do so, use: 1128 * 1129 * :: 1130 * 1131 * # sysctl kernel.perf_event_max_stack=<new value> 1132 * Return 1133 * The positive or null stack id on success, or a negative error 1134 * in case of failure. 1135 * 1136 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1137 * Description 1138 * Compute a checksum difference, from the raw buffer pointed by 1139 * *from*, of length *from_size* (that must be a multiple of 4), 1140 * towards the raw buffer pointed by *to*, of size *to_size* 1141 * (same remark). An optional *seed* can be added to the value 1142 * (this can be cascaded, the seed may come from a previous call 1143 * to the helper). 1144 * 1145 * This is flexible enough to be used in several ways: 1146 * 1147 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1148 * checksum, it can be used when pushing new data. 1149 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1150 * checksum, it can be used when removing data from a packet. 1151 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1152 * can be used to compute a diff. Note that *from_size* and 1153 * *to_size* do not need to be equal. 1154 * 1155 * This helper can be used in combination with 1156 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1157 * which one can feed in the difference computed with 1158 * **bpf_csum_diff**\ (). 1159 * Return 1160 * The checksum result, or a negative error code in case of 1161 * failure. 1162 * 1163 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1164 * Description 1165 * Retrieve tunnel options metadata for the packet associated to 1166 * *skb*, and store the raw tunnel option data to the buffer *opt* 1167 * of *size*. 1168 * 1169 * This helper can be used with encapsulation devices that can 1170 * operate in "collect metadata" mode (please refer to the related 1171 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1172 * more details). A particular example where this can be used is 1173 * in combination with the Geneve encapsulation protocol, where it 1174 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1175 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1176 * the eBPF program. This allows for full customization of these 1177 * headers. 1178 * Return 1179 * The size of the option data retrieved. 1180 * 1181 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1182 * Description 1183 * Set tunnel options metadata for the packet associated to *skb* 1184 * to the option data contained in the raw buffer *opt* of *size*. 1185 * 1186 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1187 * helper for additional information. 1188 * Return 1189 * 0 on success, or a negative error in case of failure. 1190 * 1191 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1192 * Description 1193 * Change the protocol of the *skb* to *proto*. Currently 1194 * supported are transition from IPv4 to IPv6, and from IPv6 to 1195 * IPv4. The helper takes care of the groundwork for the 1196 * transition, including resizing the socket buffer. The eBPF 1197 * program is expected to fill the new headers, if any, via 1198 * **skb_store_bytes**\ () and to recompute the checksums with 1199 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1200 * (). The main case for this helper is to perform NAT64 1201 * operations out of an eBPF program. 1202 * 1203 * Internally, the GSO type is marked as dodgy so that headers are 1204 * checked and segments are recalculated by the GSO/GRO engine. 1205 * The size for GSO target is adapted as well. 1206 * 1207 * All values for *flags* are reserved for future usage, and must 1208 * be left at zero. 1209 * 1210 * A call to this helper is susceptible to change the underlying 1211 * packet buffer. Therefore, at load time, all checks on pointers 1212 * previously done by the verifier are invalidated and must be 1213 * performed again, if the helper is used in combination with 1214 * direct packet access. 1215 * Return 1216 * 0 on success, or a negative error in case of failure. 1217 * 1218 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1219 * Description 1220 * Change the packet type for the packet associated to *skb*. This 1221 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1222 * the eBPF program does not have a write access to *skb*\ 1223 * **->pkt_type** beside this helper. Using a helper here allows 1224 * for graceful handling of errors. 1225 * 1226 * The major use case is to change incoming *skb*s to 1227 * **PACKET_HOST** in a programmatic way instead of having to 1228 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1229 * example. 1230 * 1231 * Note that *type* only allows certain values. At this time, they 1232 * are: 1233 * 1234 * **PACKET_HOST** 1235 * Packet is for us. 1236 * **PACKET_BROADCAST** 1237 * Send packet to all. 1238 * **PACKET_MULTICAST** 1239 * Send packet to group. 1240 * **PACKET_OTHERHOST** 1241 * Send packet to someone else. 1242 * Return 1243 * 0 on success, or a negative error in case of failure. 1244 * 1245 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1246 * Description 1247 * Check whether *skb* is a descendant of the cgroup2 held by 1248 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1249 * Return 1250 * The return value depends on the result of the test, and can be: 1251 * 1252 * * 0, if the *skb* failed the cgroup2 descendant test. 1253 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1254 * * A negative error code, if an error occurred. 1255 * 1256 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1257 * Description 1258 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1259 * not set, in particular if the hash was cleared due to mangling, 1260 * recompute this hash. Later accesses to the hash can be done 1261 * directly with *skb*\ **->hash**. 1262 * 1263 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1264 * prototype with **bpf_skb_change_proto**\ (), or calling 1265 * **bpf_skb_store_bytes**\ () with the 1266 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1267 * the hash and to trigger a new computation for the next call to 1268 * **bpf_get_hash_recalc**\ (). 1269 * Return 1270 * The 32-bit hash. 1271 * 1272 * u64 bpf_get_current_task(void) 1273 * Return 1274 * A pointer to the current task struct. 1275 * 1276 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1277 * Description 1278 * Attempt in a safe way to write *len* bytes from the buffer 1279 * *src* to *dst* in memory. It only works for threads that are in 1280 * user context, and *dst* must be a valid user space address. 1281 * 1282 * This helper should not be used to implement any kind of 1283 * security mechanism because of TOC-TOU attacks, but rather to 1284 * debug, divert, and manipulate execution of semi-cooperative 1285 * processes. 1286 * 1287 * Keep in mind that this feature is meant for experiments, and it 1288 * has a risk of crashing the system and running programs. 1289 * Therefore, when an eBPF program using this helper is attached, 1290 * a warning including PID and process name is printed to kernel 1291 * logs. 1292 * Return 1293 * 0 on success, or a negative error in case of failure. 1294 * 1295 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1296 * Description 1297 * Check whether the probe is being run is the context of a given 1298 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1299 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1300 * Return 1301 * The return value depends on the result of the test, and can be: 1302 * 1303 * * 0, if the *skb* task belongs to the cgroup2. 1304 * * 1, if the *skb* task does not belong to the cgroup2. 1305 * * A negative error code, if an error occurred. 1306 * 1307 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1308 * Description 1309 * Resize (trim or grow) the packet associated to *skb* to the 1310 * new *len*. The *flags* are reserved for future usage, and must 1311 * be left at zero. 1312 * 1313 * The basic idea is that the helper performs the needed work to 1314 * change the size of the packet, then the eBPF program rewrites 1315 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1316 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1317 * and others. This helper is a slow path utility intended for 1318 * replies with control messages. And because it is targeted for 1319 * slow path, the helper itself can afford to be slow: it 1320 * implicitly linearizes, unclones and drops offloads from the 1321 * *skb*. 1322 * 1323 * A call to this helper is susceptible to change the underlying 1324 * packet buffer. Therefore, at load time, all checks on pointers 1325 * previously done by the verifier are invalidated and must be 1326 * performed again, if the helper is used in combination with 1327 * direct packet access. 1328 * Return 1329 * 0 on success, or a negative error in case of failure. 1330 * 1331 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1332 * Description 1333 * Pull in non-linear data in case the *skb* is non-linear and not 1334 * all of *len* are part of the linear section. Make *len* bytes 1335 * from *skb* readable and writable. If a zero value is passed for 1336 * *len*, then the whole length of the *skb* is pulled. 1337 * 1338 * This helper is only needed for reading and writing with direct 1339 * packet access. 1340 * 1341 * For direct packet access, testing that offsets to access 1342 * are within packet boundaries (test on *skb*\ **->data_end**) is 1343 * susceptible to fail if offsets are invalid, or if the requested 1344 * data is in non-linear parts of the *skb*. On failure the 1345 * program can just bail out, or in the case of a non-linear 1346 * buffer, use a helper to make the data available. The 1347 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1348 * the data. Another one consists in using **bpf_skb_pull_data** 1349 * to pull in once the non-linear parts, then retesting and 1350 * eventually access the data. 1351 * 1352 * At the same time, this also makes sure the *skb* is uncloned, 1353 * which is a necessary condition for direct write. As this needs 1354 * to be an invariant for the write part only, the verifier 1355 * detects writes and adds a prologue that is calling 1356 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1357 * the very beginning in case it is indeed cloned. 1358 * 1359 * A call to this helper is susceptible to change the underlying 1360 * packet buffer. Therefore, at load time, all checks on pointers 1361 * previously done by the verifier are invalidated and must be 1362 * performed again, if the helper is used in combination with 1363 * direct packet access. 1364 * Return 1365 * 0 on success, or a negative error in case of failure. 1366 * 1367 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1368 * Description 1369 * Add the checksum *csum* into *skb*\ **->csum** in case the 1370 * driver has supplied a checksum for the entire packet into that 1371 * field. Return an error otherwise. This helper is intended to be 1372 * used in combination with **bpf_csum_diff**\ (), in particular 1373 * when the checksum needs to be updated after data has been 1374 * written into the packet through direct packet access. 1375 * Return 1376 * The checksum on success, or a negative error code in case of 1377 * failure. 1378 * 1379 * void bpf_set_hash_invalid(struct sk_buff *skb) 1380 * Description 1381 * Invalidate the current *skb*\ **->hash**. It can be used after 1382 * mangling on headers through direct packet access, in order to 1383 * indicate that the hash is outdated and to trigger a 1384 * recalculation the next time the kernel tries to access this 1385 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1386 * 1387 * int bpf_get_numa_node_id(void) 1388 * Description 1389 * Return the id of the current NUMA node. The primary use case 1390 * for this helper is the selection of sockets for the local NUMA 1391 * node, when the program is attached to sockets using the 1392 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1393 * but the helper is also available to other eBPF program types, 1394 * similarly to **bpf_get_smp_processor_id**\ (). 1395 * Return 1396 * The id of current NUMA node. 1397 * 1398 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1399 * Description 1400 * Grows headroom of packet associated to *skb* and adjusts the 1401 * offset of the MAC header accordingly, adding *len* bytes of 1402 * space. It automatically extends and reallocates memory as 1403 * required. 1404 * 1405 * This helper can be used on a layer 3 *skb* to push a MAC header 1406 * for redirection into a layer 2 device. 1407 * 1408 * All values for *flags* are reserved for future usage, and must 1409 * be left at zero. 1410 * 1411 * A call to this helper is susceptible to change the underlying 1412 * packet buffer. Therefore, at load time, all checks on pointers 1413 * previously done by the verifier are invalidated and must be 1414 * performed again, if the helper is used in combination with 1415 * direct packet access. 1416 * Return 1417 * 0 on success, or a negative error in case of failure. 1418 * 1419 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1420 * Description 1421 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1422 * it is possible to use a negative value for *delta*. This helper 1423 * can be used to prepare the packet for pushing or popping 1424 * headers. 1425 * 1426 * A call to this helper is susceptible to change the underlying 1427 * packet buffer. Therefore, at load time, all checks on pointers 1428 * previously done by the verifier are invalidated and must be 1429 * performed again, if the helper is used in combination with 1430 * direct packet access. 1431 * Return 1432 * 0 on success, or a negative error in case of failure. 1433 * 1434 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1435 * Description 1436 * Copy a NUL terminated string from an unsafe kernel address 1437 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for 1438 * more details. 1439 * 1440 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str() 1441 * instead. 1442 * Return 1443 * On success, the strictly positive length of the string, 1444 * including the trailing NUL character. On error, a negative 1445 * value. 1446 * 1447 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1448 * Description 1449 * If the **struct sk_buff** pointed by *skb* has a known socket, 1450 * retrieve the cookie (generated by the kernel) of this socket. 1451 * If no cookie has been set yet, generate a new cookie. Once 1452 * generated, the socket cookie remains stable for the life of the 1453 * socket. This helper can be useful for monitoring per socket 1454 * networking traffic statistics as it provides a global socket 1455 * identifier that can be assumed unique. 1456 * Return 1457 * A 8-byte long non-decreasing number on success, or 0 if the 1458 * socket field is missing inside *skb*. 1459 * 1460 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1461 * Description 1462 * Equivalent to bpf_get_socket_cookie() helper that accepts 1463 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1464 * Return 1465 * A 8-byte long non-decreasing number. 1466 * 1467 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1468 * Description 1469 * Equivalent to bpf_get_socket_cookie() helper that accepts 1470 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1471 * Return 1472 * A 8-byte long non-decreasing number. 1473 * 1474 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1475 * Return 1476 * The owner UID of the socket associated to *skb*. If the socket 1477 * is **NULL**, or if it is not a full socket (i.e. if it is a 1478 * time-wait or a request socket instead), **overflowuid** value 1479 * is returned (note that **overflowuid** might also be the actual 1480 * UID value for the socket). 1481 * 1482 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1483 * Description 1484 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1485 * to value *hash*. 1486 * Return 1487 * 0 1488 * 1489 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1490 * Description 1491 * Emulate a call to **setsockopt()** on the socket associated to 1492 * *bpf_socket*, which must be a full socket. The *level* at 1493 * which the option resides and the name *optname* of the option 1494 * must be specified, see **setsockopt(2)** for more information. 1495 * The option value of length *optlen* is pointed by *optval*. 1496 * 1497 * This helper actually implements a subset of **setsockopt()**. 1498 * It supports the following *level*\ s: 1499 * 1500 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1501 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1502 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1503 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1504 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1505 * **TCP_BPF_SNDCWND_CLAMP**. 1506 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1507 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1508 * Return 1509 * 0 on success, or a negative error in case of failure. 1510 * 1511 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1512 * Description 1513 * Grow or shrink the room for data in the packet associated to 1514 * *skb* by *len_diff*, and according to the selected *mode*. 1515 * 1516 * There are two supported modes at this time: 1517 * 1518 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1519 * (room space is added or removed below the layer 2 header). 1520 * 1521 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1522 * (room space is added or removed below the layer 3 header). 1523 * 1524 * The following flags are supported at this time: 1525 * 1526 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1527 * Adjusting mss in this way is not allowed for datagrams. 1528 * 1529 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1530 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1531 * Any new space is reserved to hold a tunnel header. 1532 * Configure skb offsets and other fields accordingly. 1533 * 1534 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1535 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1536 * Use with ENCAP_L3 flags to further specify the tunnel type. 1537 * 1538 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1539 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1540 * type; *len* is the length of the inner MAC header. 1541 * 1542 * A call to this helper is susceptible to change the underlying 1543 * packet buffer. Therefore, at load time, all checks on pointers 1544 * previously done by the verifier are invalidated and must be 1545 * performed again, if the helper is used in combination with 1546 * direct packet access. 1547 * Return 1548 * 0 on success, or a negative error in case of failure. 1549 * 1550 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1551 * Description 1552 * Redirect the packet to the endpoint referenced by *map* at 1553 * index *key*. Depending on its type, this *map* can contain 1554 * references to net devices (for forwarding packets through other 1555 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1556 * but this is only implemented for native XDP (with driver 1557 * support) as of this writing). 1558 * 1559 * The lower two bits of *flags* are used as the return code if 1560 * the map lookup fails. This is so that the return value can be 1561 * one of the XDP program return codes up to XDP_TX, as chosen by 1562 * the caller. Any higher bits in the *flags* argument must be 1563 * unset. 1564 * 1565 * When used to redirect packets to net devices, this helper 1566 * provides a high performance increase over **bpf_redirect**\ (). 1567 * This is due to various implementation details of the underlying 1568 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1569 * () tries to send packet as a "bulk" to the device. 1570 * Return 1571 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1572 * 1573 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1574 * Description 1575 * Redirect the packet to the socket referenced by *map* (of type 1576 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1577 * egress interfaces can be used for redirection. The 1578 * **BPF_F_INGRESS** value in *flags* is used to make the 1579 * distinction (ingress path is selected if the flag is present, 1580 * egress path otherwise). This is the only flag supported for now. 1581 * Return 1582 * **SK_PASS** on success, or **SK_DROP** on error. 1583 * 1584 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1585 * Description 1586 * Add an entry to, or update a *map* referencing sockets. The 1587 * *skops* is used as a new value for the entry associated to 1588 * *key*. *flags* is one of: 1589 * 1590 * **BPF_NOEXIST** 1591 * The entry for *key* must not exist in the map. 1592 * **BPF_EXIST** 1593 * The entry for *key* must already exist in the map. 1594 * **BPF_ANY** 1595 * No condition on the existence of the entry for *key*. 1596 * 1597 * If the *map* has eBPF programs (parser and verdict), those will 1598 * be inherited by the socket being added. If the socket is 1599 * already attached to eBPF programs, this results in an error. 1600 * Return 1601 * 0 on success, or a negative error in case of failure. 1602 * 1603 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1604 * Description 1605 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1606 * *delta* (which can be positive or negative). Note that this 1607 * operation modifies the address stored in *xdp_md*\ **->data**, 1608 * so the latter must be loaded only after the helper has been 1609 * called. 1610 * 1611 * The use of *xdp_md*\ **->data_meta** is optional and programs 1612 * are not required to use it. The rationale is that when the 1613 * packet is processed with XDP (e.g. as DoS filter), it is 1614 * possible to push further meta data along with it before passing 1615 * to the stack, and to give the guarantee that an ingress eBPF 1616 * program attached as a TC classifier on the same device can pick 1617 * this up for further post-processing. Since TC works with socket 1618 * buffers, it remains possible to set from XDP the **mark** or 1619 * **priority** pointers, or other pointers for the socket buffer. 1620 * Having this scratch space generic and programmable allows for 1621 * more flexibility as the user is free to store whatever meta 1622 * data they need. 1623 * 1624 * A call to this helper is susceptible to change the underlying 1625 * packet buffer. Therefore, at load time, all checks on pointers 1626 * previously done by the verifier are invalidated and must be 1627 * performed again, if the helper is used in combination with 1628 * direct packet access. 1629 * Return 1630 * 0 on success, or a negative error in case of failure. 1631 * 1632 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1633 * Description 1634 * Read the value of a perf event counter, and store it into *buf* 1635 * of size *buf_size*. This helper relies on a *map* of type 1636 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1637 * counter is selected when *map* is updated with perf event file 1638 * descriptors. The *map* is an array whose size is the number of 1639 * available CPUs, and each cell contains a value relative to one 1640 * CPU. The value to retrieve is indicated by *flags*, that 1641 * contains the index of the CPU to look up, masked with 1642 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1643 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1644 * current CPU should be retrieved. 1645 * 1646 * This helper behaves in a way close to 1647 * **bpf_perf_event_read**\ () helper, save that instead of 1648 * just returning the value observed, it fills the *buf* 1649 * structure. This allows for additional data to be retrieved: in 1650 * particular, the enabled and running times (in *buf*\ 1651 * **->enabled** and *buf*\ **->running**, respectively) are 1652 * copied. In general, **bpf_perf_event_read_value**\ () is 1653 * recommended over **bpf_perf_event_read**\ (), which has some 1654 * ABI issues and provides fewer functionalities. 1655 * 1656 * These values are interesting, because hardware PMU (Performance 1657 * Monitoring Unit) counters are limited resources. When there are 1658 * more PMU based perf events opened than available counters, 1659 * kernel will multiplex these events so each event gets certain 1660 * percentage (but not all) of the PMU time. In case that 1661 * multiplexing happens, the number of samples or counter value 1662 * will not reflect the case compared to when no multiplexing 1663 * occurs. This makes comparison between different runs difficult. 1664 * Typically, the counter value should be normalized before 1665 * comparing to other experiments. The usual normalization is done 1666 * as follows. 1667 * 1668 * :: 1669 * 1670 * normalized_counter = counter * t_enabled / t_running 1671 * 1672 * Where t_enabled is the time enabled for event and t_running is 1673 * the time running for event since last normalization. The 1674 * enabled and running times are accumulated since the perf event 1675 * open. To achieve scaling factor between two invocations of an 1676 * eBPF program, users can can use CPU id as the key (which is 1677 * typical for perf array usage model) to remember the previous 1678 * value and do the calculation inside the eBPF program. 1679 * Return 1680 * 0 on success, or a negative error in case of failure. 1681 * 1682 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1683 * Description 1684 * For en eBPF program attached to a perf event, retrieve the 1685 * value of the event counter associated to *ctx* and store it in 1686 * the structure pointed by *buf* and of size *buf_size*. Enabled 1687 * and running times are also stored in the structure (see 1688 * description of helper **bpf_perf_event_read_value**\ () for 1689 * more details). 1690 * Return 1691 * 0 on success, or a negative error in case of failure. 1692 * 1693 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1694 * Description 1695 * Emulate a call to **getsockopt()** on the socket associated to 1696 * *bpf_socket*, which must be a full socket. The *level* at 1697 * which the option resides and the name *optname* of the option 1698 * must be specified, see **getsockopt(2)** for more information. 1699 * The retrieved value is stored in the structure pointed by 1700 * *opval* and of length *optlen*. 1701 * 1702 * This helper actually implements a subset of **getsockopt()**. 1703 * It supports the following *level*\ s: 1704 * 1705 * * **IPPROTO_TCP**, which supports *optname* 1706 * **TCP_CONGESTION**. 1707 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1708 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1709 * Return 1710 * 0 on success, or a negative error in case of failure. 1711 * 1712 * int bpf_override_return(struct pt_regs *regs, u64 rc) 1713 * Description 1714 * Used for error injection, this helper uses kprobes to override 1715 * the return value of the probed function, and to set it to *rc*. 1716 * The first argument is the context *regs* on which the kprobe 1717 * works. 1718 * 1719 * This helper works by setting setting the PC (program counter) 1720 * to an override function which is run in place of the original 1721 * probed function. This means the probed function is not run at 1722 * all. The replacement function just returns with the required 1723 * value. 1724 * 1725 * This helper has security implications, and thus is subject to 1726 * restrictions. It is only available if the kernel was compiled 1727 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1728 * option, and in this case it only works on functions tagged with 1729 * **ALLOW_ERROR_INJECTION** in the kernel code. 1730 * 1731 * Also, the helper is only available for the architectures having 1732 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1733 * x86 architecture is the only one to support this feature. 1734 * Return 1735 * 0 1736 * 1737 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1738 * Description 1739 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1740 * for the full TCP socket associated to *bpf_sock_ops* to 1741 * *argval*. 1742 * 1743 * The primary use of this field is to determine if there should 1744 * be calls to eBPF programs of type 1745 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1746 * code. A program of the same type can change its value, per 1747 * connection and as necessary, when the connection is 1748 * established. This field is directly accessible for reading, but 1749 * this helper must be used for updates in order to return an 1750 * error if an eBPF program tries to set a callback that is not 1751 * supported in the current kernel. 1752 * 1753 * *argval* is a flag array which can combine these flags: 1754 * 1755 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1756 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1757 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1758 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1759 * 1760 * Therefore, this function can be used to clear a callback flag by 1761 * setting the appropriate bit to zero. e.g. to disable the RTO 1762 * callback: 1763 * 1764 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1765 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1766 * 1767 * Here are some examples of where one could call such eBPF 1768 * program: 1769 * 1770 * * When RTO fires. 1771 * * When a packet is retransmitted. 1772 * * When the connection terminates. 1773 * * When a packet is sent. 1774 * * When a packet is received. 1775 * Return 1776 * Code **-EINVAL** if the socket is not a full TCP socket; 1777 * otherwise, a positive number containing the bits that could not 1778 * be set is returned (which comes down to 0 if all bits were set 1779 * as required). 1780 * 1781 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1782 * Description 1783 * This helper is used in programs implementing policies at the 1784 * socket level. If the message *msg* is allowed to pass (i.e. if 1785 * the verdict eBPF program returns **SK_PASS**), redirect it to 1786 * the socket referenced by *map* (of type 1787 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1788 * egress interfaces can be used for redirection. The 1789 * **BPF_F_INGRESS** value in *flags* is used to make the 1790 * distinction (ingress path is selected if the flag is present, 1791 * egress path otherwise). This is the only flag supported for now. 1792 * Return 1793 * **SK_PASS** on success, or **SK_DROP** on error. 1794 * 1795 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1796 * Description 1797 * For socket policies, apply the verdict of the eBPF program to 1798 * the next *bytes* (number of bytes) of message *msg*. 1799 * 1800 * For example, this helper can be used in the following cases: 1801 * 1802 * * A single **sendmsg**\ () or **sendfile**\ () system call 1803 * contains multiple logical messages that the eBPF program is 1804 * supposed to read and for which it should apply a verdict. 1805 * * An eBPF program only cares to read the first *bytes* of a 1806 * *msg*. If the message has a large payload, then setting up 1807 * and calling the eBPF program repeatedly for all bytes, even 1808 * though the verdict is already known, would create unnecessary 1809 * overhead. 1810 * 1811 * When called from within an eBPF program, the helper sets a 1812 * counter internal to the BPF infrastructure, that is used to 1813 * apply the last verdict to the next *bytes*. If *bytes* is 1814 * smaller than the current data being processed from a 1815 * **sendmsg**\ () or **sendfile**\ () system call, the first 1816 * *bytes* will be sent and the eBPF program will be re-run with 1817 * the pointer for start of data pointing to byte number *bytes* 1818 * **+ 1**. If *bytes* is larger than the current data being 1819 * processed, then the eBPF verdict will be applied to multiple 1820 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1821 * consumed. 1822 * 1823 * Note that if a socket closes with the internal counter holding 1824 * a non-zero value, this is not a problem because data is not 1825 * being buffered for *bytes* and is sent as it is received. 1826 * Return 1827 * 0 1828 * 1829 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1830 * Description 1831 * For socket policies, prevent the execution of the verdict eBPF 1832 * program for message *msg* until *bytes* (byte number) have been 1833 * accumulated. 1834 * 1835 * This can be used when one needs a specific number of bytes 1836 * before a verdict can be assigned, even if the data spans 1837 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1838 * case would be a user calling **sendmsg**\ () repeatedly with 1839 * 1-byte long message segments. Obviously, this is bad for 1840 * performance, but it is still valid. If the eBPF program needs 1841 * *bytes* bytes to validate a header, this helper can be used to 1842 * prevent the eBPF program to be called again until *bytes* have 1843 * been accumulated. 1844 * Return 1845 * 0 1846 * 1847 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1848 * Description 1849 * For socket policies, pull in non-linear data from user space 1850 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1851 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1852 * respectively. 1853 * 1854 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1855 * *msg* it can only parse data that the (**data**, **data_end**) 1856 * pointers have already consumed. For **sendmsg**\ () hooks this 1857 * is likely the first scatterlist element. But for calls relying 1858 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1859 * be the range (**0**, **0**) because the data is shared with 1860 * user space and by default the objective is to avoid allowing 1861 * user space to modify data while (or after) eBPF verdict is 1862 * being decided. This helper can be used to pull in data and to 1863 * set the start and end pointer to given values. Data will be 1864 * copied if necessary (i.e. if data was not linear and if start 1865 * and end pointers do not point to the same chunk). 1866 * 1867 * A call to this helper is susceptible to change the underlying 1868 * packet buffer. Therefore, at load time, all checks on pointers 1869 * previously done by the verifier are invalidated and must be 1870 * performed again, if the helper is used in combination with 1871 * direct packet access. 1872 * 1873 * All values for *flags* are reserved for future usage, and must 1874 * be left at zero. 1875 * Return 1876 * 0 on success, or a negative error in case of failure. 1877 * 1878 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1879 * Description 1880 * Bind the socket associated to *ctx* to the address pointed by 1881 * *addr*, of length *addr_len*. This allows for making outgoing 1882 * connection from the desired IP address, which can be useful for 1883 * example when all processes inside a cgroup should use one 1884 * single IP address on a host that has multiple IP configured. 1885 * 1886 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1887 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1888 * **AF_INET6**). Looking for a free port to bind to can be 1889 * expensive, therefore binding to port is not permitted by the 1890 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1891 * must be set to zero. 1892 * Return 1893 * 0 on success, or a negative error in case of failure. 1894 * 1895 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1896 * Description 1897 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1898 * only possible to shrink the packet as of this writing, 1899 * therefore *delta* must be a negative integer. 1900 * 1901 * A call to this helper is susceptible to change the underlying 1902 * packet buffer. Therefore, at load time, all checks on pointers 1903 * previously done by the verifier are invalidated and must be 1904 * performed again, if the helper is used in combination with 1905 * direct packet access. 1906 * Return 1907 * 0 on success, or a negative error in case of failure. 1908 * 1909 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1910 * Description 1911 * Retrieve the XFRM state (IP transform framework, see also 1912 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1913 * 1914 * The retrieved value is stored in the **struct bpf_xfrm_state** 1915 * pointed by *xfrm_state* and of length *size*. 1916 * 1917 * All values for *flags* are reserved for future usage, and must 1918 * be left at zero. 1919 * 1920 * This helper is available only if the kernel was compiled with 1921 * **CONFIG_XFRM** configuration option. 1922 * Return 1923 * 0 on success, or a negative error in case of failure. 1924 * 1925 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 1926 * Description 1927 * Return a user or a kernel stack in bpf program provided buffer. 1928 * To achieve this, the helper needs *ctx*, which is a pointer 1929 * to the context on which the tracing program is executed. 1930 * To store the stacktrace, the bpf program provides *buf* with 1931 * a nonnegative *size*. 1932 * 1933 * The last argument, *flags*, holds the number of stack frames to 1934 * skip (from 0 to 255), masked with 1935 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1936 * the following flags: 1937 * 1938 * **BPF_F_USER_STACK** 1939 * Collect a user space stack instead of a kernel stack. 1940 * **BPF_F_USER_BUILD_ID** 1941 * Collect buildid+offset instead of ips for user stack, 1942 * only valid if **BPF_F_USER_STACK** is also specified. 1943 * 1944 * **bpf_get_stack**\ () can collect up to 1945 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1946 * to sufficient large buffer size. Note that 1947 * this limit can be controlled with the **sysctl** program, and 1948 * that it should be manually increased in order to profile long 1949 * user stacks (such as stacks for Java programs). To do so, use: 1950 * 1951 * :: 1952 * 1953 * # sysctl kernel.perf_event_max_stack=<new value> 1954 * Return 1955 * A non-negative value equal to or less than *size* on success, 1956 * or a negative error in case of failure. 1957 * 1958 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 1959 * Description 1960 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1961 * it provides an easy way to load *len* bytes from *offset* 1962 * from the packet associated to *skb*, into the buffer pointed 1963 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1964 * a fifth argument *start_header* exists in order to select a 1965 * base offset to start from. *start_header* can be one of: 1966 * 1967 * **BPF_HDR_START_MAC** 1968 * Base offset to load data from is *skb*'s mac header. 1969 * **BPF_HDR_START_NET** 1970 * Base offset to load data from is *skb*'s network header. 1971 * 1972 * In general, "direct packet access" is the preferred method to 1973 * access packet data, however, this helper is in particular useful 1974 * in socket filters where *skb*\ **->data** does not always point 1975 * to the start of the mac header and where "direct packet access" 1976 * is not available. 1977 * Return 1978 * 0 on success, or a negative error in case of failure. 1979 * 1980 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 1981 * Description 1982 * Do FIB lookup in kernel tables using parameters in *params*. 1983 * If lookup is successful and result shows packet is to be 1984 * forwarded, the neighbor tables are searched for the nexthop. 1985 * If successful (ie., FIB lookup shows forwarding and nexthop 1986 * is resolved), the nexthop address is returned in ipv4_dst 1987 * or ipv6_dst based on family, smac is set to mac address of 1988 * egress device, dmac is set to nexthop mac address, rt_metric 1989 * is set to metric from route (IPv4/IPv6 only), and ifindex 1990 * is set to the device index of the nexthop from the FIB lookup. 1991 * 1992 * *plen* argument is the size of the passed in struct. 1993 * *flags* argument can be a combination of one or more of the 1994 * following values: 1995 * 1996 * **BPF_FIB_LOOKUP_DIRECT** 1997 * Do a direct table lookup vs full lookup using FIB 1998 * rules. 1999 * **BPF_FIB_LOOKUP_OUTPUT** 2000 * Perform lookup from an egress perspective (default is 2001 * ingress). 2002 * 2003 * *ctx* is either **struct xdp_md** for XDP programs or 2004 * **struct sk_buff** tc cls_act programs. 2005 * Return 2006 * * < 0 if any input argument is invalid 2007 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2008 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2009 * packet is not forwarded or needs assist from full stack 2010 * 2011 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2012 * Description 2013 * Add an entry to, or update a sockhash *map* referencing sockets. 2014 * The *skops* is used as a new value for the entry associated to 2015 * *key*. *flags* is one of: 2016 * 2017 * **BPF_NOEXIST** 2018 * The entry for *key* must not exist in the map. 2019 * **BPF_EXIST** 2020 * The entry for *key* must already exist in the map. 2021 * **BPF_ANY** 2022 * No condition on the existence of the entry for *key*. 2023 * 2024 * If the *map* has eBPF programs (parser and verdict), those will 2025 * be inherited by the socket being added. If the socket is 2026 * already attached to eBPF programs, this results in an error. 2027 * Return 2028 * 0 on success, or a negative error in case of failure. 2029 * 2030 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2031 * Description 2032 * This helper is used in programs implementing policies at the 2033 * socket level. If the message *msg* is allowed to pass (i.e. if 2034 * the verdict eBPF program returns **SK_PASS**), redirect it to 2035 * the socket referenced by *map* (of type 2036 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2037 * egress interfaces can be used for redirection. The 2038 * **BPF_F_INGRESS** value in *flags* is used to make the 2039 * distinction (ingress path is selected if the flag is present, 2040 * egress path otherwise). This is the only flag supported for now. 2041 * Return 2042 * **SK_PASS** on success, or **SK_DROP** on error. 2043 * 2044 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2045 * Description 2046 * This helper is used in programs implementing policies at the 2047 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2048 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2049 * to the socket referenced by *map* (of type 2050 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2051 * egress interfaces can be used for redirection. The 2052 * **BPF_F_INGRESS** value in *flags* is used to make the 2053 * distinction (ingress path is selected if the flag is present, 2054 * egress otherwise). This is the only flag supported for now. 2055 * Return 2056 * **SK_PASS** on success, or **SK_DROP** on error. 2057 * 2058 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2059 * Description 2060 * Encapsulate the packet associated to *skb* within a Layer 3 2061 * protocol header. This header is provided in the buffer at 2062 * address *hdr*, with *len* its size in bytes. *type* indicates 2063 * the protocol of the header and can be one of: 2064 * 2065 * **BPF_LWT_ENCAP_SEG6** 2066 * IPv6 encapsulation with Segment Routing Header 2067 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2068 * the IPv6 header is computed by the kernel. 2069 * **BPF_LWT_ENCAP_SEG6_INLINE** 2070 * Only works if *skb* contains an IPv6 packet. Insert a 2071 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2072 * the IPv6 header. 2073 * **BPF_LWT_ENCAP_IP** 2074 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2075 * must be IPv4 or IPv6, followed by zero or more 2076 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2077 * total bytes in all prepended headers. Please note that 2078 * if **skb_is_gso**\ (*skb*) is true, no more than two 2079 * headers can be prepended, and the inner header, if 2080 * present, should be either GRE or UDP/GUE. 2081 * 2082 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2083 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2084 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2085 * **BPF_PROG_TYPE_LWT_XMIT**. 2086 * 2087 * A call to this helper is susceptible to change the underlying 2088 * packet buffer. Therefore, at load time, all checks on pointers 2089 * previously done by the verifier are invalidated and must be 2090 * performed again, if the helper is used in combination with 2091 * direct packet access. 2092 * Return 2093 * 0 on success, or a negative error in case of failure. 2094 * 2095 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2096 * Description 2097 * Store *len* bytes from address *from* into the packet 2098 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2099 * inside the outermost IPv6 Segment Routing Header can be 2100 * modified through this helper. 2101 * 2102 * A call to this helper is susceptible to change the underlying 2103 * packet buffer. Therefore, at load time, all checks on pointers 2104 * previously done by the verifier are invalidated and must be 2105 * performed again, if the helper is used in combination with 2106 * direct packet access. 2107 * Return 2108 * 0 on success, or a negative error in case of failure. 2109 * 2110 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2111 * Description 2112 * Adjust the size allocated to TLVs in the outermost IPv6 2113 * Segment Routing Header contained in the packet associated to 2114 * *skb*, at position *offset* by *delta* bytes. Only offsets 2115 * after the segments are accepted. *delta* can be as well 2116 * positive (growing) as negative (shrinking). 2117 * 2118 * A call to this helper is susceptible to change the underlying 2119 * packet buffer. Therefore, at load time, all checks on pointers 2120 * previously done by the verifier are invalidated and must be 2121 * performed again, if the helper is used in combination with 2122 * direct packet access. 2123 * Return 2124 * 0 on success, or a negative error in case of failure. 2125 * 2126 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2127 * Description 2128 * Apply an IPv6 Segment Routing action of type *action* to the 2129 * packet associated to *skb*. Each action takes a parameter 2130 * contained at address *param*, and of length *param_len* bytes. 2131 * *action* can be one of: 2132 * 2133 * **SEG6_LOCAL_ACTION_END_X** 2134 * End.X action: Endpoint with Layer-3 cross-connect. 2135 * Type of *param*: **struct in6_addr**. 2136 * **SEG6_LOCAL_ACTION_END_T** 2137 * End.T action: Endpoint with specific IPv6 table lookup. 2138 * Type of *param*: **int**. 2139 * **SEG6_LOCAL_ACTION_END_B6** 2140 * End.B6 action: Endpoint bound to an SRv6 policy. 2141 * Type of *param*: **struct ipv6_sr_hdr**. 2142 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2143 * End.B6.Encap action: Endpoint bound to an SRv6 2144 * encapsulation policy. 2145 * Type of *param*: **struct ipv6_sr_hdr**. 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_rc_repeat(void *ctx) 2156 * Description 2157 * This helper is used in programs implementing IR decoding, to 2158 * report a successfully decoded repeat key message. This delays 2159 * the generation of a key up event for previously generated 2160 * key down event. 2161 * 2162 * Some IR protocols like NEC have a special IR message for 2163 * repeating last button, for when a button is held down. 2164 * 2165 * The *ctx* should point to the lirc sample as passed into 2166 * the program. 2167 * 2168 * This helper is only available is the kernel was compiled with 2169 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2170 * "**y**". 2171 * Return 2172 * 0 2173 * 2174 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2175 * Description 2176 * This helper is used in programs implementing IR decoding, to 2177 * report a successfully decoded key press with *scancode*, 2178 * *toggle* value in the given *protocol*. The scancode will be 2179 * translated to a keycode using the rc keymap, and reported as 2180 * an input key down event. After a period a key up event is 2181 * generated. This period can be extended by calling either 2182 * **bpf_rc_keydown**\ () again with the same values, or calling 2183 * **bpf_rc_repeat**\ (). 2184 * 2185 * Some protocols include a toggle bit, in case the button was 2186 * released and pressed again between consecutive scancodes. 2187 * 2188 * The *ctx* should point to the lirc sample as passed into 2189 * the program. 2190 * 2191 * The *protocol* is the decoded protocol number (see 2192 * **enum rc_proto** for some predefined values). 2193 * 2194 * This helper is only available is the kernel was compiled with 2195 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2196 * "**y**". 2197 * Return 2198 * 0 2199 * 2200 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2201 * Description 2202 * Return the cgroup v2 id of the socket associated with the *skb*. 2203 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2204 * helper for cgroup v1 by providing a tag resp. identifier that 2205 * can be matched on or used for map lookups e.g. to implement 2206 * policy. The cgroup v2 id of a given path in the hierarchy is 2207 * exposed in user space through the f_handle API in order to get 2208 * to the same 64-bit id. 2209 * 2210 * This helper can be used on TC egress path, but not on ingress, 2211 * and is available only if the kernel was compiled with the 2212 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2213 * Return 2214 * The id is returned or 0 in case the id could not be retrieved. 2215 * 2216 * u64 bpf_get_current_cgroup_id(void) 2217 * Return 2218 * A 64-bit integer containing the current cgroup id based 2219 * on the cgroup within which the current task is running. 2220 * 2221 * void *bpf_get_local_storage(void *map, u64 flags) 2222 * Description 2223 * Get the pointer to the local storage area. 2224 * The type and the size of the local storage is defined 2225 * by the *map* argument. 2226 * The *flags* meaning is specific for each map type, 2227 * and has to be 0 for cgroup local storage. 2228 * 2229 * Depending on the BPF program type, a local storage area 2230 * can be shared between multiple instances of the BPF program, 2231 * running simultaneously. 2232 * 2233 * A user should care about the synchronization by himself. 2234 * For example, by using the **BPF_STX_XADD** instruction to alter 2235 * the shared data. 2236 * Return 2237 * A pointer to the local storage area. 2238 * 2239 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2240 * Description 2241 * Select a **SO_REUSEPORT** socket from a 2242 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2243 * It checks the selected socket is matching the incoming 2244 * request in the socket buffer. 2245 * Return 2246 * 0 on success, or a negative error in case of failure. 2247 * 2248 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2249 * Description 2250 * Return id of cgroup v2 that is ancestor of cgroup associated 2251 * with the *skb* at the *ancestor_level*. The root cgroup is at 2252 * *ancestor_level* zero and each step down the hierarchy 2253 * increments the level. If *ancestor_level* == level of cgroup 2254 * associated with *skb*, then return value will be same as that 2255 * of **bpf_skb_cgroup_id**\ (). 2256 * 2257 * The helper is useful to implement policies based on cgroups 2258 * that are upper in hierarchy than immediate cgroup associated 2259 * with *skb*. 2260 * 2261 * The format of returned id and helper limitations are same as in 2262 * **bpf_skb_cgroup_id**\ (). 2263 * Return 2264 * The id is returned or 0 in case the id could not be retrieved. 2265 * 2266 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2267 * Description 2268 * Look for TCP socket matching *tuple*, optionally in a child 2269 * network namespace *netns*. The return value must be checked, 2270 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2271 * 2272 * The *ctx* should point to the context of the program, such as 2273 * the skb or socket (depending on the hook in use). This is used 2274 * to determine the base network namespace for the lookup. 2275 * 2276 * *tuple_size* must be one of: 2277 * 2278 * **sizeof**\ (*tuple*\ **->ipv4**) 2279 * Look for an IPv4 socket. 2280 * **sizeof**\ (*tuple*\ **->ipv6**) 2281 * Look for an IPv6 socket. 2282 * 2283 * If the *netns* is a negative signed 32-bit integer, then the 2284 * socket lookup table in the netns associated with the *ctx* will 2285 * will be used. For the TC hooks, this is the netns of the device 2286 * in the skb. For socket hooks, this is the netns of the socket. 2287 * If *netns* is any other signed 32-bit value greater than or 2288 * equal to zero then it specifies the ID of the netns relative to 2289 * the netns associated with the *ctx*. *netns* values beyond the 2290 * range of 32-bit integers are reserved for future use. 2291 * 2292 * All values for *flags* are reserved for future usage, and must 2293 * be left at zero. 2294 * 2295 * This helper is available only if the kernel was compiled with 2296 * **CONFIG_NET** configuration option. 2297 * Return 2298 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2299 * For sockets with reuseport option, the **struct bpf_sock** 2300 * result is from *reuse*\ **->socks**\ [] using the hash of the 2301 * tuple. 2302 * 2303 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2304 * Description 2305 * Look for UDP socket matching *tuple*, optionally in a child 2306 * network namespace *netns*. The return value must be checked, 2307 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2308 * 2309 * The *ctx* should point to the context of the program, such as 2310 * the skb or socket (depending on the hook in use). This is used 2311 * to determine the base network namespace for the lookup. 2312 * 2313 * *tuple_size* must be one of: 2314 * 2315 * **sizeof**\ (*tuple*\ **->ipv4**) 2316 * Look for an IPv4 socket. 2317 * **sizeof**\ (*tuple*\ **->ipv6**) 2318 * Look for an IPv6 socket. 2319 * 2320 * If the *netns* is a negative signed 32-bit integer, then the 2321 * socket lookup table in the netns associated with the *ctx* will 2322 * will be used. For the TC hooks, this is the netns of the device 2323 * in the skb. For socket hooks, this is the netns of the socket. 2324 * If *netns* is any other signed 32-bit value greater than or 2325 * equal to zero then it specifies the ID of the netns relative to 2326 * the netns associated with the *ctx*. *netns* values beyond the 2327 * range of 32-bit integers are reserved for future use. 2328 * 2329 * All values for *flags* are reserved for future usage, and must 2330 * be left at zero. 2331 * 2332 * This helper is available only if the kernel was compiled with 2333 * **CONFIG_NET** configuration option. 2334 * Return 2335 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2336 * For sockets with reuseport option, the **struct bpf_sock** 2337 * result is from *reuse*\ **->socks**\ [] using the hash of the 2338 * tuple. 2339 * 2340 * int bpf_sk_release(struct bpf_sock *sock) 2341 * Description 2342 * Release the reference held by *sock*. *sock* must be a 2343 * non-**NULL** pointer that was returned from 2344 * **bpf_sk_lookup_xxx**\ (). 2345 * Return 2346 * 0 on success, or a negative error in case of failure. 2347 * 2348 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2349 * Description 2350 * Push an element *value* in *map*. *flags* is one of: 2351 * 2352 * **BPF_EXIST** 2353 * If the queue/stack is full, the oldest element is 2354 * removed to make room for this. 2355 * Return 2356 * 0 on success, or a negative error in case of failure. 2357 * 2358 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2359 * Description 2360 * Pop an element from *map*. 2361 * Return 2362 * 0 on success, or a negative error in case of failure. 2363 * 2364 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2365 * Description 2366 * Get an element from *map* without removing it. 2367 * Return 2368 * 0 on success, or a negative error in case of failure. 2369 * 2370 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2371 * Description 2372 * For socket policies, insert *len* bytes into *msg* at offset 2373 * *start*. 2374 * 2375 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2376 * *msg* it may want to insert metadata or options into the *msg*. 2377 * This can later be read and used by any of the lower layer BPF 2378 * hooks. 2379 * 2380 * This helper may fail if under memory pressure (a malloc 2381 * fails) in these cases BPF programs will get an appropriate 2382 * error and BPF programs will need to handle them. 2383 * Return 2384 * 0 on success, or a negative error in case of failure. 2385 * 2386 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2387 * Description 2388 * Will remove *len* bytes from a *msg* starting at byte *start*. 2389 * This may result in **ENOMEM** errors under certain situations if 2390 * an allocation and copy are required due to a full ring buffer. 2391 * However, the helper will try to avoid doing the allocation 2392 * if possible. Other errors can occur if input parameters are 2393 * invalid either due to *start* byte not being valid part of *msg* 2394 * payload and/or *pop* value being to large. 2395 * Return 2396 * 0 on success, or a negative error in case of failure. 2397 * 2398 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2399 * Description 2400 * This helper is used in programs implementing IR decoding, to 2401 * report a successfully decoded pointer movement. 2402 * 2403 * The *ctx* should point to the lirc sample as passed into 2404 * the program. 2405 * 2406 * This helper is only available is the kernel was compiled with 2407 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2408 * "**y**". 2409 * Return 2410 * 0 2411 * 2412 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2413 * Description 2414 * Acquire a spinlock represented by the pointer *lock*, which is 2415 * stored as part of a value of a map. Taking the lock allows to 2416 * safely update the rest of the fields in that value. The 2417 * spinlock can (and must) later be released with a call to 2418 * **bpf_spin_unlock**\ (\ *lock*\ ). 2419 * 2420 * Spinlocks in BPF programs come with a number of restrictions 2421 * and constraints: 2422 * 2423 * * **bpf_spin_lock** objects are only allowed inside maps of 2424 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2425 * list could be extended in the future). 2426 * * BTF description of the map is mandatory. 2427 * * The BPF program can take ONE lock at a time, since taking two 2428 * or more could cause dead locks. 2429 * * Only one **struct bpf_spin_lock** is allowed per map element. 2430 * * When the lock is taken, calls (either BPF to BPF or helpers) 2431 * are not allowed. 2432 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2433 * allowed inside a spinlock-ed region. 2434 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2435 * the lock, on all execution paths, before it returns. 2436 * * The BPF program can access **struct bpf_spin_lock** only via 2437 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2438 * helpers. Loading or storing data into the **struct 2439 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2440 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2441 * of the map value must be a struct and have **struct 2442 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2443 * Nested lock inside another struct is not allowed. 2444 * * The **struct bpf_spin_lock** *lock* field in a map value must 2445 * be aligned on a multiple of 4 bytes in that value. 2446 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2447 * the **bpf_spin_lock** field to user space. 2448 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2449 * a BPF program, do not update the **bpf_spin_lock** field. 2450 * * **bpf_spin_lock** cannot be on the stack or inside a 2451 * networking packet (it can only be inside of a map values). 2452 * * **bpf_spin_lock** is available to root only. 2453 * * Tracing programs and socket filter programs cannot use 2454 * **bpf_spin_lock**\ () due to insufficient preemption checks 2455 * (but this may change in the future). 2456 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2457 * Return 2458 * 0 2459 * 2460 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2461 * Description 2462 * Release the *lock* previously locked by a call to 2463 * **bpf_spin_lock**\ (\ *lock*\ ). 2464 * Return 2465 * 0 2466 * 2467 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2468 * Description 2469 * This helper gets a **struct bpf_sock** pointer such 2470 * that all the fields in this **bpf_sock** can be accessed. 2471 * Return 2472 * A **struct bpf_sock** pointer on success, or **NULL** in 2473 * case of failure. 2474 * 2475 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2476 * Description 2477 * This helper gets a **struct bpf_tcp_sock** pointer from a 2478 * **struct bpf_sock** pointer. 2479 * Return 2480 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2481 * case of failure. 2482 * 2483 * int bpf_skb_ecn_set_ce(struct sk_buff *skb) 2484 * Description 2485 * Set ECN (Explicit Congestion Notification) field of IP header 2486 * to **CE** (Congestion Encountered) if current value is **ECT** 2487 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2488 * and IPv4. 2489 * Return 2490 * 1 if the **CE** flag is set (either by the current helper call 2491 * or because it was already present), 0 if it is not set. 2492 * 2493 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2494 * Description 2495 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2496 * **bpf_sk_release**\ () is unnecessary and not allowed. 2497 * Return 2498 * A **struct bpf_sock** pointer on success, or **NULL** in 2499 * case of failure. 2500 * 2501 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2502 * Description 2503 * Look for TCP socket matching *tuple*, optionally in a child 2504 * network namespace *netns*. The return value must be checked, 2505 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2506 * 2507 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2508 * that it also returns timewait or request sockets. Use 2509 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2510 * full structure. 2511 * 2512 * This helper is available only if the kernel was compiled with 2513 * **CONFIG_NET** configuration option. 2514 * Return 2515 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2516 * For sockets with reuseport option, the **struct bpf_sock** 2517 * result is from *reuse*\ **->socks**\ [] using the hash of the 2518 * tuple. 2519 * 2520 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2521 * Description 2522 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2523 * the listening socket in *sk*. 2524 * 2525 * *iph* points to the start of the IPv4 or IPv6 header, while 2526 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2527 * **sizeof**\ (**struct ip6hdr**). 2528 * 2529 * *th* points to the start of the TCP header, while *th_len* 2530 * contains **sizeof**\ (**struct tcphdr**). 2531 * 2532 * Return 2533 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2534 * error otherwise. 2535 * 2536 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2537 * Description 2538 * Get name of sysctl in /proc/sys/ and copy it into provided by 2539 * program buffer *buf* of size *buf_len*. 2540 * 2541 * The buffer is always NUL terminated, unless it's zero-sized. 2542 * 2543 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2544 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2545 * only (e.g. "tcp_mem"). 2546 * Return 2547 * Number of character copied (not including the trailing NUL). 2548 * 2549 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2550 * truncated name in this case). 2551 * 2552 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2553 * Description 2554 * Get current value of sysctl as it is presented in /proc/sys 2555 * (incl. newline, etc), and copy it as a string into provided 2556 * by program buffer *buf* of size *buf_len*. 2557 * 2558 * The whole value is copied, no matter what file position user 2559 * space issued e.g. sys_read at. 2560 * 2561 * The buffer is always NUL terminated, unless it's zero-sized. 2562 * Return 2563 * Number of character copied (not including the trailing NUL). 2564 * 2565 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2566 * truncated name in this case). 2567 * 2568 * **-EINVAL** if current value was unavailable, e.g. because 2569 * sysctl is uninitialized and read returns -EIO for it. 2570 * 2571 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2572 * Description 2573 * Get new value being written by user space to sysctl (before 2574 * the actual write happens) and copy it as a string into 2575 * provided by program buffer *buf* of size *buf_len*. 2576 * 2577 * User space may write new value at file position > 0. 2578 * 2579 * The buffer is always NUL terminated, unless it's zero-sized. 2580 * Return 2581 * Number of character copied (not including the trailing NUL). 2582 * 2583 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2584 * truncated name in this case). 2585 * 2586 * **-EINVAL** if sysctl is being read. 2587 * 2588 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2589 * Description 2590 * Override new value being written by user space to sysctl with 2591 * value provided by program in buffer *buf* of size *buf_len*. 2592 * 2593 * *buf* should contain a string in same form as provided by user 2594 * space on sysctl write. 2595 * 2596 * User space may write new value at file position > 0. To override 2597 * the whole sysctl value file position should be set to zero. 2598 * Return 2599 * 0 on success. 2600 * 2601 * **-E2BIG** if the *buf_len* is too big. 2602 * 2603 * **-EINVAL** if sysctl is being read. 2604 * 2605 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2606 * Description 2607 * Convert the initial part of the string from buffer *buf* of 2608 * size *buf_len* to a long integer according to the given base 2609 * and save the result in *res*. 2610 * 2611 * The string may begin with an arbitrary amount of white space 2612 * (as determined by **isspace**\ (3)) followed by a single 2613 * optional '**-**' sign. 2614 * 2615 * Five least significant bits of *flags* encode base, other bits 2616 * are currently unused. 2617 * 2618 * Base must be either 8, 10, 16 or 0 to detect it automatically 2619 * similar to user space **strtol**\ (3). 2620 * Return 2621 * Number of characters consumed on success. Must be positive but 2622 * no more than *buf_len*. 2623 * 2624 * **-EINVAL** if no valid digits were found or unsupported base 2625 * was provided. 2626 * 2627 * **-ERANGE** if resulting value was out of range. 2628 * 2629 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2630 * Description 2631 * Convert the initial part of the string from buffer *buf* of 2632 * size *buf_len* to an unsigned long integer according to the 2633 * given base and save the result in *res*. 2634 * 2635 * The string may begin with an arbitrary amount of white space 2636 * (as determined by **isspace**\ (3)). 2637 * 2638 * Five least significant bits of *flags* encode base, other bits 2639 * are currently unused. 2640 * 2641 * Base must be either 8, 10, 16 or 0 to detect it automatically 2642 * similar to user space **strtoul**\ (3). 2643 * Return 2644 * Number of characters consumed on success. Must be positive but 2645 * no more than *buf_len*. 2646 * 2647 * **-EINVAL** if no valid digits were found or unsupported base 2648 * was provided. 2649 * 2650 * **-ERANGE** if resulting value was out of range. 2651 * 2652 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2653 * Description 2654 * Get a bpf-local-storage from a *sk*. 2655 * 2656 * Logically, it could be thought of getting the value from 2657 * a *map* with *sk* as the **key**. From this 2658 * perspective, the usage is not much different from 2659 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2660 * helper enforces the key must be a full socket and the map must 2661 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2662 * 2663 * Underneath, the value is stored locally at *sk* instead of 2664 * the *map*. The *map* is used as the bpf-local-storage 2665 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2666 * searched against all bpf-local-storages residing at *sk*. 2667 * 2668 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2669 * used such that a new bpf-local-storage will be 2670 * created if one does not exist. *value* can be used 2671 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2672 * the initial value of a bpf-local-storage. If *value* is 2673 * **NULL**, the new bpf-local-storage will be zero initialized. 2674 * Return 2675 * A bpf-local-storage pointer is returned on success. 2676 * 2677 * **NULL** if not found or there was an error in adding 2678 * a new bpf-local-storage. 2679 * 2680 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2681 * Description 2682 * Delete a bpf-local-storage from a *sk*. 2683 * Return 2684 * 0 on success. 2685 * 2686 * **-ENOENT** if the bpf-local-storage cannot be found. 2687 * 2688 * int bpf_send_signal(u32 sig) 2689 * Description 2690 * Send signal *sig* to the current task. 2691 * Return 2692 * 0 on success or successfully queued. 2693 * 2694 * **-EBUSY** if work queue under nmi is full. 2695 * 2696 * **-EINVAL** if *sig* is invalid. 2697 * 2698 * **-EPERM** if no permission to send the *sig*. 2699 * 2700 * **-EAGAIN** if bpf program can try again. 2701 * 2702 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2703 * Description 2704 * Try to issue a SYN cookie for the packet with corresponding 2705 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2706 * 2707 * *iph* points to the start of the IPv4 or IPv6 header, while 2708 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2709 * **sizeof**\ (**struct ip6hdr**). 2710 * 2711 * *th* points to the start of the TCP header, while *th_len* 2712 * contains the length of the TCP header. 2713 * 2714 * Return 2715 * On success, lower 32 bits hold the generated SYN cookie in 2716 * followed by 16 bits which hold the MSS value for that cookie, 2717 * and the top 16 bits are unused. 2718 * 2719 * On failure, the returned value is one of the following: 2720 * 2721 * **-EINVAL** SYN cookie cannot be issued due to error 2722 * 2723 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2724 * 2725 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2726 * 2727 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2728 * 2729 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2730 * Description 2731 * Write raw *data* blob into a special BPF perf event held by 2732 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2733 * event must have the following attributes: **PERF_SAMPLE_RAW** 2734 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2735 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2736 * 2737 * The *flags* are used to indicate the index in *map* for which 2738 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2739 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2740 * to indicate that the index of the current CPU core should be 2741 * used. 2742 * 2743 * The value to write, of *size*, is passed through eBPF stack and 2744 * pointed by *data*. 2745 * 2746 * *ctx* is a pointer to in-kernel struct sk_buff. 2747 * 2748 * This helper is similar to **bpf_perf_event_output**\ () but 2749 * restricted to raw_tracepoint bpf programs. 2750 * Return 2751 * 0 on success, or a negative error in case of failure. 2752 * 2753 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2754 * Description 2755 * Safely attempt to read *size* bytes from user space address 2756 * *unsafe_ptr* and store the data in *dst*. 2757 * Return 2758 * 0 on success, or a negative error in case of failure. 2759 * 2760 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2761 * Description 2762 * Safely attempt to read *size* bytes from kernel space address 2763 * *unsafe_ptr* and store the data in *dst*. 2764 * Return 2765 * 0 on success, or a negative error in case of failure. 2766 * 2767 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2768 * Description 2769 * Copy a NUL terminated string from an unsafe user address 2770 * *unsafe_ptr* to *dst*. The *size* should include the 2771 * terminating NUL byte. In case the string length is smaller than 2772 * *size*, the target is not padded with further NUL bytes. If the 2773 * string length is larger than *size*, just *size*-1 bytes are 2774 * copied and the last byte is set to NUL. 2775 * 2776 * On success, the length of the copied string is returned. This 2777 * makes this helper useful in tracing programs for reading 2778 * strings, and more importantly to get its length at runtime. See 2779 * the following snippet: 2780 * 2781 * :: 2782 * 2783 * SEC("kprobe/sys_open") 2784 * void bpf_sys_open(struct pt_regs *ctx) 2785 * { 2786 * char buf[PATHLEN]; // PATHLEN is defined to 256 2787 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2788 * ctx->di); 2789 * 2790 * // Consume buf, for example push it to 2791 * // userspace via bpf_perf_event_output(); we 2792 * // can use res (the string length) as event 2793 * // size, after checking its boundaries. 2794 * } 2795 * 2796 * In comparison, using **bpf_probe_read_user()** helper here 2797 * instead to read the string would require to estimate the length 2798 * at compile time, and would often result in copying more memory 2799 * than necessary. 2800 * 2801 * Another useful use case is when parsing individual process 2802 * arguments or individual environment variables navigating 2803 * *current*\ **->mm->arg_start** and *current*\ 2804 * **->mm->env_start**: using this helper and the return value, 2805 * one can quickly iterate at the right offset of the memory area. 2806 * Return 2807 * On success, the strictly positive length of the string, 2808 * including the trailing NUL character. On error, a negative 2809 * value. 2810 * 2811 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 2812 * Description 2813 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2814 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply. 2815 * Return 2816 * On success, the strictly positive length of the string, including 2817 * the trailing NUL character. On error, a negative value. 2818 */ 2819 #define __BPF_FUNC_MAPPER(FN) \ 2820 FN(unspec), \ 2821 FN(map_lookup_elem), \ 2822 FN(map_update_elem), \ 2823 FN(map_delete_elem), \ 2824 FN(probe_read), \ 2825 FN(ktime_get_ns), \ 2826 FN(trace_printk), \ 2827 FN(get_prandom_u32), \ 2828 FN(get_smp_processor_id), \ 2829 FN(skb_store_bytes), \ 2830 FN(l3_csum_replace), \ 2831 FN(l4_csum_replace), \ 2832 FN(tail_call), \ 2833 FN(clone_redirect), \ 2834 FN(get_current_pid_tgid), \ 2835 FN(get_current_uid_gid), \ 2836 FN(get_current_comm), \ 2837 FN(get_cgroup_classid), \ 2838 FN(skb_vlan_push), \ 2839 FN(skb_vlan_pop), \ 2840 FN(skb_get_tunnel_key), \ 2841 FN(skb_set_tunnel_key), \ 2842 FN(perf_event_read), \ 2843 FN(redirect), \ 2844 FN(get_route_realm), \ 2845 FN(perf_event_output), \ 2846 FN(skb_load_bytes), \ 2847 FN(get_stackid), \ 2848 FN(csum_diff), \ 2849 FN(skb_get_tunnel_opt), \ 2850 FN(skb_set_tunnel_opt), \ 2851 FN(skb_change_proto), \ 2852 FN(skb_change_type), \ 2853 FN(skb_under_cgroup), \ 2854 FN(get_hash_recalc), \ 2855 FN(get_current_task), \ 2856 FN(probe_write_user), \ 2857 FN(current_task_under_cgroup), \ 2858 FN(skb_change_tail), \ 2859 FN(skb_pull_data), \ 2860 FN(csum_update), \ 2861 FN(set_hash_invalid), \ 2862 FN(get_numa_node_id), \ 2863 FN(skb_change_head), \ 2864 FN(xdp_adjust_head), \ 2865 FN(probe_read_str), \ 2866 FN(get_socket_cookie), \ 2867 FN(get_socket_uid), \ 2868 FN(set_hash), \ 2869 FN(setsockopt), \ 2870 FN(skb_adjust_room), \ 2871 FN(redirect_map), \ 2872 FN(sk_redirect_map), \ 2873 FN(sock_map_update), \ 2874 FN(xdp_adjust_meta), \ 2875 FN(perf_event_read_value), \ 2876 FN(perf_prog_read_value), \ 2877 FN(getsockopt), \ 2878 FN(override_return), \ 2879 FN(sock_ops_cb_flags_set), \ 2880 FN(msg_redirect_map), \ 2881 FN(msg_apply_bytes), \ 2882 FN(msg_cork_bytes), \ 2883 FN(msg_pull_data), \ 2884 FN(bind), \ 2885 FN(xdp_adjust_tail), \ 2886 FN(skb_get_xfrm_state), \ 2887 FN(get_stack), \ 2888 FN(skb_load_bytes_relative), \ 2889 FN(fib_lookup), \ 2890 FN(sock_hash_update), \ 2891 FN(msg_redirect_hash), \ 2892 FN(sk_redirect_hash), \ 2893 FN(lwt_push_encap), \ 2894 FN(lwt_seg6_store_bytes), \ 2895 FN(lwt_seg6_adjust_srh), \ 2896 FN(lwt_seg6_action), \ 2897 FN(rc_repeat), \ 2898 FN(rc_keydown), \ 2899 FN(skb_cgroup_id), \ 2900 FN(get_current_cgroup_id), \ 2901 FN(get_local_storage), \ 2902 FN(sk_select_reuseport), \ 2903 FN(skb_ancestor_cgroup_id), \ 2904 FN(sk_lookup_tcp), \ 2905 FN(sk_lookup_udp), \ 2906 FN(sk_release), \ 2907 FN(map_push_elem), \ 2908 FN(map_pop_elem), \ 2909 FN(map_peek_elem), \ 2910 FN(msg_push_data), \ 2911 FN(msg_pop_data), \ 2912 FN(rc_pointer_rel), \ 2913 FN(spin_lock), \ 2914 FN(spin_unlock), \ 2915 FN(sk_fullsock), \ 2916 FN(tcp_sock), \ 2917 FN(skb_ecn_set_ce), \ 2918 FN(get_listener_sock), \ 2919 FN(skc_lookup_tcp), \ 2920 FN(tcp_check_syncookie), \ 2921 FN(sysctl_get_name), \ 2922 FN(sysctl_get_current_value), \ 2923 FN(sysctl_get_new_value), \ 2924 FN(sysctl_set_new_value), \ 2925 FN(strtol), \ 2926 FN(strtoul), \ 2927 FN(sk_storage_get), \ 2928 FN(sk_storage_delete), \ 2929 FN(send_signal), \ 2930 FN(tcp_gen_syncookie), \ 2931 FN(skb_output), \ 2932 FN(probe_read_user), \ 2933 FN(probe_read_kernel), \ 2934 FN(probe_read_user_str), \ 2935 FN(probe_read_kernel_str), 2936 2937 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2938 * function eBPF program intends to call 2939 */ 2940 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2941 enum bpf_func_id { 2942 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2943 __BPF_FUNC_MAX_ID, 2944 }; 2945 #undef __BPF_ENUM_FN 2946 2947 /* All flags used by eBPF helper functions, placed here. */ 2948 2949 /* BPF_FUNC_skb_store_bytes flags. */ 2950 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2951 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2952 2953 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2954 * First 4 bits are for passing the header field size. 2955 */ 2956 #define BPF_F_HDR_FIELD_MASK 0xfULL 2957 2958 /* BPF_FUNC_l4_csum_replace flags. */ 2959 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2960 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2961 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2962 2963 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2964 #define BPF_F_INGRESS (1ULL << 0) 2965 2966 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2967 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2968 2969 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2970 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2971 #define BPF_F_USER_STACK (1ULL << 8) 2972 /* flags used by BPF_FUNC_get_stackid only. */ 2973 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2974 #define BPF_F_REUSE_STACKID (1ULL << 10) 2975 /* flags used by BPF_FUNC_get_stack only. */ 2976 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2977 2978 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2979 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2980 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2981 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2982 2983 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2984 * BPF_FUNC_perf_event_read_value flags. 2985 */ 2986 #define BPF_F_INDEX_MASK 0xffffffffULL 2987 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2988 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2989 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2990 2991 /* Current network namespace */ 2992 #define BPF_F_CURRENT_NETNS (-1L) 2993 2994 /* BPF_FUNC_skb_adjust_room flags. */ 2995 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 2996 2997 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 2998 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 2999 3000 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 3001 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 3002 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 3003 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 3004 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3005 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3006 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3007 3008 /* BPF_FUNC_sysctl_get_name flags. */ 3009 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0) 3010 3011 /* BPF_FUNC_sk_storage_get flags */ 3012 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0) 3013 3014 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 3015 enum bpf_adj_room_mode { 3016 BPF_ADJ_ROOM_NET, 3017 BPF_ADJ_ROOM_MAC, 3018 }; 3019 3020 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3021 enum bpf_hdr_start_off { 3022 BPF_HDR_START_MAC, 3023 BPF_HDR_START_NET, 3024 }; 3025 3026 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3027 enum bpf_lwt_encap_mode { 3028 BPF_LWT_ENCAP_SEG6, 3029 BPF_LWT_ENCAP_SEG6_INLINE, 3030 BPF_LWT_ENCAP_IP, 3031 }; 3032 3033 #define __bpf_md_ptr(type, name) \ 3034 union { \ 3035 type name; \ 3036 __u64 :64; \ 3037 } __attribute__((aligned(8))) 3038 3039 /* user accessible mirror of in-kernel sk_buff. 3040 * new fields can only be added to the end of this structure 3041 */ 3042 struct __sk_buff { 3043 __u32 len; 3044 __u32 pkt_type; 3045 __u32 mark; 3046 __u32 queue_mapping; 3047 __u32 protocol; 3048 __u32 vlan_present; 3049 __u32 vlan_tci; 3050 __u32 vlan_proto; 3051 __u32 priority; 3052 __u32 ingress_ifindex; 3053 __u32 ifindex; 3054 __u32 tc_index; 3055 __u32 cb[5]; 3056 __u32 hash; 3057 __u32 tc_classid; 3058 __u32 data; 3059 __u32 data_end; 3060 __u32 napi_id; 3061 3062 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3063 __u32 family; 3064 __u32 remote_ip4; /* Stored in network byte order */ 3065 __u32 local_ip4; /* Stored in network byte order */ 3066 __u32 remote_ip6[4]; /* Stored in network byte order */ 3067 __u32 local_ip6[4]; /* Stored in network byte order */ 3068 __u32 remote_port; /* Stored in network byte order */ 3069 __u32 local_port; /* stored in host byte order */ 3070 /* ... here. */ 3071 3072 __u32 data_meta; 3073 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3074 __u64 tstamp; 3075 __u32 wire_len; 3076 __u32 gso_segs; 3077 __bpf_md_ptr(struct bpf_sock *, sk); 3078 }; 3079 3080 struct bpf_tunnel_key { 3081 __u32 tunnel_id; 3082 union { 3083 __u32 remote_ipv4; 3084 __u32 remote_ipv6[4]; 3085 }; 3086 __u8 tunnel_tos; 3087 __u8 tunnel_ttl; 3088 __u16 tunnel_ext; /* Padding, future use. */ 3089 __u32 tunnel_label; 3090 }; 3091 3092 /* user accessible mirror of in-kernel xfrm_state. 3093 * new fields can only be added to the end of this structure 3094 */ 3095 struct bpf_xfrm_state { 3096 __u32 reqid; 3097 __u32 spi; /* Stored in network byte order */ 3098 __u16 family; 3099 __u16 ext; /* Padding, future use. */ 3100 union { 3101 __u32 remote_ipv4; /* Stored in network byte order */ 3102 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3103 }; 3104 }; 3105 3106 /* Generic BPF return codes which all BPF program types may support. 3107 * The values are binary compatible with their TC_ACT_* counter-part to 3108 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3109 * programs. 3110 * 3111 * XDP is handled seprately, see XDP_*. 3112 */ 3113 enum bpf_ret_code { 3114 BPF_OK = 0, 3115 /* 1 reserved */ 3116 BPF_DROP = 2, 3117 /* 3-6 reserved */ 3118 BPF_REDIRECT = 7, 3119 /* >127 are reserved for prog type specific return codes. 3120 * 3121 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3122 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3123 * changed and should be routed based on its new L3 header. 3124 * (This is an L3 redirect, as opposed to L2 redirect 3125 * represented by BPF_REDIRECT above). 3126 */ 3127 BPF_LWT_REROUTE = 128, 3128 }; 3129 3130 struct bpf_sock { 3131 __u32 bound_dev_if; 3132 __u32 family; 3133 __u32 type; 3134 __u32 protocol; 3135 __u32 mark; 3136 __u32 priority; 3137 /* IP address also allows 1 and 2 bytes access */ 3138 __u32 src_ip4; 3139 __u32 src_ip6[4]; 3140 __u32 src_port; /* host byte order */ 3141 __u32 dst_port; /* network byte order */ 3142 __u32 dst_ip4; 3143 __u32 dst_ip6[4]; 3144 __u32 state; 3145 }; 3146 3147 struct bpf_tcp_sock { 3148 __u32 snd_cwnd; /* Sending congestion window */ 3149 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3150 __u32 rtt_min; 3151 __u32 snd_ssthresh; /* Slow start size threshold */ 3152 __u32 rcv_nxt; /* What we want to receive next */ 3153 __u32 snd_nxt; /* Next sequence we send */ 3154 __u32 snd_una; /* First byte we want an ack for */ 3155 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3156 __u32 ecn_flags; /* ECN status bits. */ 3157 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3158 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3159 __u32 packets_out; /* Packets which are "in flight" */ 3160 __u32 retrans_out; /* Retransmitted packets out */ 3161 __u32 total_retrans; /* Total retransmits for entire connection */ 3162 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3163 * total number of segments in. 3164 */ 3165 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3166 * total number of data segments in. 3167 */ 3168 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3169 * The total number of segments sent. 3170 */ 3171 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3172 * total number of data segments sent. 3173 */ 3174 __u32 lost_out; /* Lost packets */ 3175 __u32 sacked_out; /* SACK'd packets */ 3176 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3177 * sum(delta(rcv_nxt)), or how many bytes 3178 * were acked. 3179 */ 3180 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3181 * sum(delta(snd_una)), or how many bytes 3182 * were acked. 3183 */ 3184 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3185 * total number of DSACK blocks received 3186 */ 3187 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3188 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3189 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3190 }; 3191 3192 struct bpf_sock_tuple { 3193 union { 3194 struct { 3195 __be32 saddr; 3196 __be32 daddr; 3197 __be16 sport; 3198 __be16 dport; 3199 } ipv4; 3200 struct { 3201 __be32 saddr[4]; 3202 __be32 daddr[4]; 3203 __be16 sport; 3204 __be16 dport; 3205 } ipv6; 3206 }; 3207 }; 3208 3209 struct bpf_xdp_sock { 3210 __u32 queue_id; 3211 }; 3212 3213 #define XDP_PACKET_HEADROOM 256 3214 3215 /* User return codes for XDP prog type. 3216 * A valid XDP program must return one of these defined values. All other 3217 * return codes are reserved for future use. Unknown return codes will 3218 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3219 */ 3220 enum xdp_action { 3221 XDP_ABORTED = 0, 3222 XDP_DROP, 3223 XDP_PASS, 3224 XDP_TX, 3225 XDP_REDIRECT, 3226 }; 3227 3228 /* user accessible metadata for XDP packet hook 3229 * new fields must be added to the end of this structure 3230 */ 3231 struct xdp_md { 3232 __u32 data; 3233 __u32 data_end; 3234 __u32 data_meta; 3235 /* Below access go through struct xdp_rxq_info */ 3236 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3237 __u32 rx_queue_index; /* rxq->queue_index */ 3238 }; 3239 3240 enum sk_action { 3241 SK_DROP = 0, 3242 SK_PASS, 3243 }; 3244 3245 /* user accessible metadata for SK_MSG packet hook, new fields must 3246 * be added to the end of this structure 3247 */ 3248 struct sk_msg_md { 3249 __bpf_md_ptr(void *, data); 3250 __bpf_md_ptr(void *, data_end); 3251 3252 __u32 family; 3253 __u32 remote_ip4; /* Stored in network byte order */ 3254 __u32 local_ip4; /* Stored in network byte order */ 3255 __u32 remote_ip6[4]; /* Stored in network byte order */ 3256 __u32 local_ip6[4]; /* Stored in network byte order */ 3257 __u32 remote_port; /* Stored in network byte order */ 3258 __u32 local_port; /* stored in host byte order */ 3259 __u32 size; /* Total size of sk_msg */ 3260 }; 3261 3262 struct sk_reuseport_md { 3263 /* 3264 * Start of directly accessible data. It begins from 3265 * the tcp/udp header. 3266 */ 3267 __bpf_md_ptr(void *, data); 3268 /* End of directly accessible data */ 3269 __bpf_md_ptr(void *, data_end); 3270 /* 3271 * Total length of packet (starting from the tcp/udp header). 3272 * Note that the directly accessible bytes (data_end - data) 3273 * could be less than this "len". Those bytes could be 3274 * indirectly read by a helper "bpf_skb_load_bytes()". 3275 */ 3276 __u32 len; 3277 /* 3278 * Eth protocol in the mac header (network byte order). e.g. 3279 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3280 */ 3281 __u32 eth_protocol; 3282 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3283 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3284 __u32 hash; /* A hash of the packet 4 tuples */ 3285 }; 3286 3287 #define BPF_TAG_SIZE 8 3288 3289 struct bpf_prog_info { 3290 __u32 type; 3291 __u32 id; 3292 __u8 tag[BPF_TAG_SIZE]; 3293 __u32 jited_prog_len; 3294 __u32 xlated_prog_len; 3295 __aligned_u64 jited_prog_insns; 3296 __aligned_u64 xlated_prog_insns; 3297 __u64 load_time; /* ns since boottime */ 3298 __u32 created_by_uid; 3299 __u32 nr_map_ids; 3300 __aligned_u64 map_ids; 3301 char name[BPF_OBJ_NAME_LEN]; 3302 __u32 ifindex; 3303 __u32 gpl_compatible:1; 3304 __u32 :31; /* alignment pad */ 3305 __u64 netns_dev; 3306 __u64 netns_ino; 3307 __u32 nr_jited_ksyms; 3308 __u32 nr_jited_func_lens; 3309 __aligned_u64 jited_ksyms; 3310 __aligned_u64 jited_func_lens; 3311 __u32 btf_id; 3312 __u32 func_info_rec_size; 3313 __aligned_u64 func_info; 3314 __u32 nr_func_info; 3315 __u32 nr_line_info; 3316 __aligned_u64 line_info; 3317 __aligned_u64 jited_line_info; 3318 __u32 nr_jited_line_info; 3319 __u32 line_info_rec_size; 3320 __u32 jited_line_info_rec_size; 3321 __u32 nr_prog_tags; 3322 __aligned_u64 prog_tags; 3323 __u64 run_time_ns; 3324 __u64 run_cnt; 3325 } __attribute__((aligned(8))); 3326 3327 struct bpf_map_info { 3328 __u32 type; 3329 __u32 id; 3330 __u32 key_size; 3331 __u32 value_size; 3332 __u32 max_entries; 3333 __u32 map_flags; 3334 char name[BPF_OBJ_NAME_LEN]; 3335 __u32 ifindex; 3336 __u32 :32; 3337 __u64 netns_dev; 3338 __u64 netns_ino; 3339 __u32 btf_id; 3340 __u32 btf_key_type_id; 3341 __u32 btf_value_type_id; 3342 } __attribute__((aligned(8))); 3343 3344 struct bpf_btf_info { 3345 __aligned_u64 btf; 3346 __u32 btf_size; 3347 __u32 id; 3348 } __attribute__((aligned(8))); 3349 3350 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3351 * by user and intended to be used by socket (e.g. to bind to, depends on 3352 * attach attach type). 3353 */ 3354 struct bpf_sock_addr { 3355 __u32 user_family; /* Allows 4-byte read, but no write. */ 3356 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3357 * Stored in network byte order. 3358 */ 3359 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3360 * Stored in network byte order. 3361 */ 3362 __u32 user_port; /* Allows 4-byte read and write. 3363 * Stored in network byte order 3364 */ 3365 __u32 family; /* Allows 4-byte read, but no write */ 3366 __u32 type; /* Allows 4-byte read, but no write */ 3367 __u32 protocol; /* Allows 4-byte read, but no write */ 3368 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3369 * Stored in network byte order. 3370 */ 3371 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3372 * Stored in network byte order. 3373 */ 3374 __bpf_md_ptr(struct bpf_sock *, sk); 3375 }; 3376 3377 /* User bpf_sock_ops struct to access socket values and specify request ops 3378 * and their replies. 3379 * Some of this fields are in network (bigendian) byte order and may need 3380 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3381 * New fields can only be added at the end of this structure 3382 */ 3383 struct bpf_sock_ops { 3384 __u32 op; 3385 union { 3386 __u32 args[4]; /* Optionally passed to bpf program */ 3387 __u32 reply; /* Returned by bpf program */ 3388 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3389 }; 3390 __u32 family; 3391 __u32 remote_ip4; /* Stored in network byte order */ 3392 __u32 local_ip4; /* Stored in network byte order */ 3393 __u32 remote_ip6[4]; /* Stored in network byte order */ 3394 __u32 local_ip6[4]; /* Stored in network byte order */ 3395 __u32 remote_port; /* Stored in network byte order */ 3396 __u32 local_port; /* stored in host byte order */ 3397 __u32 is_fullsock; /* Some TCP fields are only valid if 3398 * there is a full socket. If not, the 3399 * fields read as zero. 3400 */ 3401 __u32 snd_cwnd; 3402 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3403 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3404 __u32 state; 3405 __u32 rtt_min; 3406 __u32 snd_ssthresh; 3407 __u32 rcv_nxt; 3408 __u32 snd_nxt; 3409 __u32 snd_una; 3410 __u32 mss_cache; 3411 __u32 ecn_flags; 3412 __u32 rate_delivered; 3413 __u32 rate_interval_us; 3414 __u32 packets_out; 3415 __u32 retrans_out; 3416 __u32 total_retrans; 3417 __u32 segs_in; 3418 __u32 data_segs_in; 3419 __u32 segs_out; 3420 __u32 data_segs_out; 3421 __u32 lost_out; 3422 __u32 sacked_out; 3423 __u32 sk_txhash; 3424 __u64 bytes_received; 3425 __u64 bytes_acked; 3426 __bpf_md_ptr(struct bpf_sock *, sk); 3427 }; 3428 3429 /* Definitions for bpf_sock_ops_cb_flags */ 3430 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3431 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3432 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3433 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3) 3434 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently 3435 * supported cb flags 3436 */ 3437 3438 /* List of known BPF sock_ops operators. 3439 * New entries can only be added at the end 3440 */ 3441 enum { 3442 BPF_SOCK_OPS_VOID, 3443 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3444 * -1 if default value should be used 3445 */ 3446 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3447 * window (in packets) or -1 if default 3448 * value should be used 3449 */ 3450 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3451 * active connection is initialized 3452 */ 3453 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3454 * active connection is 3455 * established 3456 */ 3457 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3458 * passive connection is 3459 * established 3460 */ 3461 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3462 * needs ECN 3463 */ 3464 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3465 * based on the path and may be 3466 * dependent on the congestion control 3467 * algorithm. In general it indicates 3468 * a congestion threshold. RTTs above 3469 * this indicate congestion 3470 */ 3471 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3472 * Arg1: value of icsk_retransmits 3473 * Arg2: value of icsk_rto 3474 * Arg3: whether RTO has expired 3475 */ 3476 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3477 * Arg1: sequence number of 1st byte 3478 * Arg2: # segments 3479 * Arg3: return value of 3480 * tcp_transmit_skb (0 => success) 3481 */ 3482 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3483 * Arg1: old_state 3484 * Arg2: new_state 3485 */ 3486 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3487 * socket transition to LISTEN state. 3488 */ 3489 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 3490 */ 3491 }; 3492 3493 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3494 * changes between the TCP and BPF versions. Ideally this should never happen. 3495 * If it does, we need to add code to convert them before calling 3496 * the BPF sock_ops function. 3497 */ 3498 enum { 3499 BPF_TCP_ESTABLISHED = 1, 3500 BPF_TCP_SYN_SENT, 3501 BPF_TCP_SYN_RECV, 3502 BPF_TCP_FIN_WAIT1, 3503 BPF_TCP_FIN_WAIT2, 3504 BPF_TCP_TIME_WAIT, 3505 BPF_TCP_CLOSE, 3506 BPF_TCP_CLOSE_WAIT, 3507 BPF_TCP_LAST_ACK, 3508 BPF_TCP_LISTEN, 3509 BPF_TCP_CLOSING, /* Now a valid state */ 3510 BPF_TCP_NEW_SYN_RECV, 3511 3512 BPF_TCP_MAX_STATES /* Leave at the end! */ 3513 }; 3514 3515 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3516 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3517 3518 struct bpf_perf_event_value { 3519 __u64 counter; 3520 __u64 enabled; 3521 __u64 running; 3522 }; 3523 3524 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3525 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3526 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3527 3528 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3529 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3530 3531 struct bpf_cgroup_dev_ctx { 3532 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3533 __u32 access_type; 3534 __u32 major; 3535 __u32 minor; 3536 }; 3537 3538 struct bpf_raw_tracepoint_args { 3539 __u64 args[0]; 3540 }; 3541 3542 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3543 * OUTPUT: Do lookup from egress perspective; default is ingress 3544 */ 3545 #define BPF_FIB_LOOKUP_DIRECT (1U << 0) 3546 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1) 3547 3548 enum { 3549 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3550 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3551 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3552 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3553 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3554 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3555 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3556 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3557 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3558 }; 3559 3560 struct bpf_fib_lookup { 3561 /* input: network family for lookup (AF_INET, AF_INET6) 3562 * output: network family of egress nexthop 3563 */ 3564 __u8 family; 3565 3566 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3567 __u8 l4_protocol; 3568 __be16 sport; 3569 __be16 dport; 3570 3571 /* total length of packet from network header - used for MTU check */ 3572 __u16 tot_len; 3573 3574 /* input: L3 device index for lookup 3575 * output: device index from FIB lookup 3576 */ 3577 __u32 ifindex; 3578 3579 union { 3580 /* inputs to lookup */ 3581 __u8 tos; /* AF_INET */ 3582 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3583 3584 /* output: metric of fib result (IPv4/IPv6 only) */ 3585 __u32 rt_metric; 3586 }; 3587 3588 union { 3589 __be32 ipv4_src; 3590 __u32 ipv6_src[4]; /* in6_addr; network order */ 3591 }; 3592 3593 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3594 * network header. output: bpf_fib_lookup sets to gateway address 3595 * if FIB lookup returns gateway route 3596 */ 3597 union { 3598 __be32 ipv4_dst; 3599 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3600 }; 3601 3602 /* output */ 3603 __be16 h_vlan_proto; 3604 __be16 h_vlan_TCI; 3605 __u8 smac[6]; /* ETH_ALEN */ 3606 __u8 dmac[6]; /* ETH_ALEN */ 3607 }; 3608 3609 enum bpf_task_fd_type { 3610 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3611 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3612 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3613 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3614 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3615 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3616 }; 3617 3618 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0) 3619 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1) 3620 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2) 3621 3622 struct bpf_flow_keys { 3623 __u16 nhoff; 3624 __u16 thoff; 3625 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3626 __u8 is_frag; 3627 __u8 is_first_frag; 3628 __u8 is_encap; 3629 __u8 ip_proto; 3630 __be16 n_proto; 3631 __be16 sport; 3632 __be16 dport; 3633 union { 3634 struct { 3635 __be32 ipv4_src; 3636 __be32 ipv4_dst; 3637 }; 3638 struct { 3639 __u32 ipv6_src[4]; /* in6_addr; network order */ 3640 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3641 }; 3642 }; 3643 __u32 flags; 3644 __be32 flow_label; 3645 }; 3646 3647 struct bpf_func_info { 3648 __u32 insn_off; 3649 __u32 type_id; 3650 }; 3651 3652 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3653 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3654 3655 struct bpf_line_info { 3656 __u32 insn_off; 3657 __u32 file_name_off; 3658 __u32 line_off; 3659 __u32 line_col; 3660 }; 3661 3662 struct bpf_spin_lock { 3663 __u32 val; 3664 }; 3665 3666 struct bpf_sysctl { 3667 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3668 * Allows 1,2,4-byte read, but no write. 3669 */ 3670 __u32 file_pos; /* Sysctl file position to read from, write to. 3671 * Allows 1,2,4-byte read an 4-byte write. 3672 */ 3673 }; 3674 3675 struct bpf_sockopt { 3676 __bpf_md_ptr(struct bpf_sock *, sk); 3677 __bpf_md_ptr(void *, optval); 3678 __bpf_md_ptr(void *, optval_end); 3679 3680 __s32 level; 3681 __s32 optname; 3682 __s32 optlen; 3683 __s32 retval; 3684 }; 3685 3686 #endif /* _UAPI__LINUX_BPF_H__ */ 3687