1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 /* BPF syscall commands, see bpf(2) man-page for details. */ 85 enum bpf_cmd { 86 BPF_MAP_CREATE, 87 BPF_MAP_LOOKUP_ELEM, 88 BPF_MAP_UPDATE_ELEM, 89 BPF_MAP_DELETE_ELEM, 90 BPF_MAP_GET_NEXT_KEY, 91 BPF_PROG_LOAD, 92 BPF_OBJ_PIN, 93 BPF_OBJ_GET, 94 BPF_PROG_ATTACH, 95 BPF_PROG_DETACH, 96 BPF_PROG_TEST_RUN, 97 BPF_PROG_GET_NEXT_ID, 98 BPF_MAP_GET_NEXT_ID, 99 BPF_PROG_GET_FD_BY_ID, 100 BPF_MAP_GET_FD_BY_ID, 101 BPF_OBJ_GET_INFO_BY_FD, 102 BPF_PROG_QUERY, 103 BPF_RAW_TRACEPOINT_OPEN, 104 BPF_BTF_LOAD, 105 BPF_BTF_GET_FD_BY_ID, 106 BPF_TASK_FD_QUERY, 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 108 BPF_MAP_FREEZE, 109 }; 110 111 enum bpf_map_type { 112 BPF_MAP_TYPE_UNSPEC, 113 BPF_MAP_TYPE_HASH, 114 BPF_MAP_TYPE_ARRAY, 115 BPF_MAP_TYPE_PROG_ARRAY, 116 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 117 BPF_MAP_TYPE_PERCPU_HASH, 118 BPF_MAP_TYPE_PERCPU_ARRAY, 119 BPF_MAP_TYPE_STACK_TRACE, 120 BPF_MAP_TYPE_CGROUP_ARRAY, 121 BPF_MAP_TYPE_LRU_HASH, 122 BPF_MAP_TYPE_LRU_PERCPU_HASH, 123 BPF_MAP_TYPE_LPM_TRIE, 124 BPF_MAP_TYPE_ARRAY_OF_MAPS, 125 BPF_MAP_TYPE_HASH_OF_MAPS, 126 BPF_MAP_TYPE_DEVMAP, 127 BPF_MAP_TYPE_SOCKMAP, 128 BPF_MAP_TYPE_CPUMAP, 129 BPF_MAP_TYPE_XSKMAP, 130 BPF_MAP_TYPE_SOCKHASH, 131 BPF_MAP_TYPE_CGROUP_STORAGE, 132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 134 BPF_MAP_TYPE_QUEUE, 135 BPF_MAP_TYPE_STACK, 136 }; 137 138 /* Note that tracing related programs such as 139 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 140 * are not subject to a stable API since kernel internal data 141 * structures can change from release to release and may 142 * therefore break existing tracing BPF programs. Tracing BPF 143 * programs correspond to /a/ specific kernel which is to be 144 * analyzed, and not /a/ specific kernel /and/ all future ones. 145 */ 146 enum bpf_prog_type { 147 BPF_PROG_TYPE_UNSPEC, 148 BPF_PROG_TYPE_SOCKET_FILTER, 149 BPF_PROG_TYPE_KPROBE, 150 BPF_PROG_TYPE_SCHED_CLS, 151 BPF_PROG_TYPE_SCHED_ACT, 152 BPF_PROG_TYPE_TRACEPOINT, 153 BPF_PROG_TYPE_XDP, 154 BPF_PROG_TYPE_PERF_EVENT, 155 BPF_PROG_TYPE_CGROUP_SKB, 156 BPF_PROG_TYPE_CGROUP_SOCK, 157 BPF_PROG_TYPE_LWT_IN, 158 BPF_PROG_TYPE_LWT_OUT, 159 BPF_PROG_TYPE_LWT_XMIT, 160 BPF_PROG_TYPE_SOCK_OPS, 161 BPF_PROG_TYPE_SK_SKB, 162 BPF_PROG_TYPE_CGROUP_DEVICE, 163 BPF_PROG_TYPE_SK_MSG, 164 BPF_PROG_TYPE_RAW_TRACEPOINT, 165 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 166 BPF_PROG_TYPE_LWT_SEG6LOCAL, 167 BPF_PROG_TYPE_LIRC_MODE2, 168 BPF_PROG_TYPE_SK_REUSEPORT, 169 BPF_PROG_TYPE_FLOW_DISSECTOR, 170 }; 171 172 enum bpf_attach_type { 173 BPF_CGROUP_INET_INGRESS, 174 BPF_CGROUP_INET_EGRESS, 175 BPF_CGROUP_INET_SOCK_CREATE, 176 BPF_CGROUP_SOCK_OPS, 177 BPF_SK_SKB_STREAM_PARSER, 178 BPF_SK_SKB_STREAM_VERDICT, 179 BPF_CGROUP_DEVICE, 180 BPF_SK_MSG_VERDICT, 181 BPF_CGROUP_INET4_BIND, 182 BPF_CGROUP_INET6_BIND, 183 BPF_CGROUP_INET4_CONNECT, 184 BPF_CGROUP_INET6_CONNECT, 185 BPF_CGROUP_INET4_POST_BIND, 186 BPF_CGROUP_INET6_POST_BIND, 187 BPF_CGROUP_UDP4_SENDMSG, 188 BPF_CGROUP_UDP6_SENDMSG, 189 BPF_LIRC_MODE2, 190 BPF_FLOW_DISSECTOR, 191 __MAX_BPF_ATTACH_TYPE 192 }; 193 194 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 195 196 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 197 * 198 * NONE(default): No further bpf programs allowed in the subtree. 199 * 200 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 201 * the program in this cgroup yields to sub-cgroup program. 202 * 203 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 204 * that cgroup program gets run in addition to the program in this cgroup. 205 * 206 * Only one program is allowed to be attached to a cgroup with 207 * NONE or BPF_F_ALLOW_OVERRIDE flag. 208 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 209 * release old program and attach the new one. Attach flags has to match. 210 * 211 * Multiple programs are allowed to be attached to a cgroup with 212 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 213 * (those that were attached first, run first) 214 * The programs of sub-cgroup are executed first, then programs of 215 * this cgroup and then programs of parent cgroup. 216 * When children program makes decision (like picking TCP CA or sock bind) 217 * parent program has a chance to override it. 218 * 219 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 220 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 221 * Ex1: 222 * cgrp1 (MULTI progs A, B) -> 223 * cgrp2 (OVERRIDE prog C) -> 224 * cgrp3 (MULTI prog D) -> 225 * cgrp4 (OVERRIDE prog E) -> 226 * cgrp5 (NONE prog F) 227 * the event in cgrp5 triggers execution of F,D,A,B in that order. 228 * if prog F is detached, the execution is E,D,A,B 229 * if prog F and D are detached, the execution is E,A,B 230 * if prog F, E and D are detached, the execution is C,A,B 231 * 232 * All eligible programs are executed regardless of return code from 233 * earlier programs. 234 */ 235 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 236 #define BPF_F_ALLOW_MULTI (1U << 1) 237 238 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 239 * verifier will perform strict alignment checking as if the kernel 240 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 241 * and NET_IP_ALIGN defined to 2. 242 */ 243 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 244 245 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 246 * verifier will allow any alignment whatsoever. On platforms 247 * with strict alignment requirements for loads ands stores (such 248 * as sparc and mips) the verifier validates that all loads and 249 * stores provably follow this requirement. This flag turns that 250 * checking and enforcement off. 251 * 252 * It is mostly used for testing when we want to validate the 253 * context and memory access aspects of the verifier, but because 254 * of an unaligned access the alignment check would trigger before 255 * the one we are interested in. 256 */ 257 #define BPF_F_ANY_ALIGNMENT (1U << 1) 258 259 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 260 * two extensions: 261 * 262 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 263 * insn[0].imm: map fd map fd 264 * insn[1].imm: 0 offset into value 265 * insn[0].off: 0 0 266 * insn[1].off: 0 0 267 * ldimm64 rewrite: address of map address of map[0]+offset 268 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 269 */ 270 #define BPF_PSEUDO_MAP_FD 1 271 #define BPF_PSEUDO_MAP_VALUE 2 272 273 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 274 * offset to another bpf function 275 */ 276 #define BPF_PSEUDO_CALL 1 277 278 /* flags for BPF_MAP_UPDATE_ELEM command */ 279 #define BPF_ANY 0 /* create new element or update existing */ 280 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 281 #define BPF_EXIST 2 /* update existing element */ 282 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 283 284 /* flags for BPF_MAP_CREATE command */ 285 #define BPF_F_NO_PREALLOC (1U << 0) 286 /* Instead of having one common LRU list in the 287 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 288 * which can scale and perform better. 289 * Note, the LRU nodes (including free nodes) cannot be moved 290 * across different LRU lists. 291 */ 292 #define BPF_F_NO_COMMON_LRU (1U << 1) 293 /* Specify numa node during map creation */ 294 #define BPF_F_NUMA_NODE (1U << 2) 295 296 #define BPF_OBJ_NAME_LEN 16U 297 298 /* Flags for accessing BPF object from syscall side. */ 299 #define BPF_F_RDONLY (1U << 3) 300 #define BPF_F_WRONLY (1U << 4) 301 302 /* Flag for stack_map, store build_id+offset instead of pointer */ 303 #define BPF_F_STACK_BUILD_ID (1U << 5) 304 305 /* Zero-initialize hash function seed. This should only be used for testing. */ 306 #define BPF_F_ZERO_SEED (1U << 6) 307 308 /* Flags for accessing BPF object from program side. */ 309 #define BPF_F_RDONLY_PROG (1U << 7) 310 #define BPF_F_WRONLY_PROG (1U << 8) 311 312 /* flags for BPF_PROG_QUERY */ 313 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 314 315 enum bpf_stack_build_id_status { 316 /* user space need an empty entry to identify end of a trace */ 317 BPF_STACK_BUILD_ID_EMPTY = 0, 318 /* with valid build_id and offset */ 319 BPF_STACK_BUILD_ID_VALID = 1, 320 /* couldn't get build_id, fallback to ip */ 321 BPF_STACK_BUILD_ID_IP = 2, 322 }; 323 324 #define BPF_BUILD_ID_SIZE 20 325 struct bpf_stack_build_id { 326 __s32 status; 327 unsigned char build_id[BPF_BUILD_ID_SIZE]; 328 union { 329 __u64 offset; 330 __u64 ip; 331 }; 332 }; 333 334 union bpf_attr { 335 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 336 __u32 map_type; /* one of enum bpf_map_type */ 337 __u32 key_size; /* size of key in bytes */ 338 __u32 value_size; /* size of value in bytes */ 339 __u32 max_entries; /* max number of entries in a map */ 340 __u32 map_flags; /* BPF_MAP_CREATE related 341 * flags defined above. 342 */ 343 __u32 inner_map_fd; /* fd pointing to the inner map */ 344 __u32 numa_node; /* numa node (effective only if 345 * BPF_F_NUMA_NODE is set). 346 */ 347 char map_name[BPF_OBJ_NAME_LEN]; 348 __u32 map_ifindex; /* ifindex of netdev to create on */ 349 __u32 btf_fd; /* fd pointing to a BTF type data */ 350 __u32 btf_key_type_id; /* BTF type_id of the key */ 351 __u32 btf_value_type_id; /* BTF type_id of the value */ 352 }; 353 354 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 355 __u32 map_fd; 356 __aligned_u64 key; 357 union { 358 __aligned_u64 value; 359 __aligned_u64 next_key; 360 }; 361 __u64 flags; 362 }; 363 364 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 365 __u32 prog_type; /* one of enum bpf_prog_type */ 366 __u32 insn_cnt; 367 __aligned_u64 insns; 368 __aligned_u64 license; 369 __u32 log_level; /* verbosity level of verifier */ 370 __u32 log_size; /* size of user buffer */ 371 __aligned_u64 log_buf; /* user supplied buffer */ 372 __u32 kern_version; /* not used */ 373 __u32 prog_flags; 374 char prog_name[BPF_OBJ_NAME_LEN]; 375 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 376 /* For some prog types expected attach type must be known at 377 * load time to verify attach type specific parts of prog 378 * (context accesses, allowed helpers, etc). 379 */ 380 __u32 expected_attach_type; 381 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 382 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 383 __aligned_u64 func_info; /* func info */ 384 __u32 func_info_cnt; /* number of bpf_func_info records */ 385 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 386 __aligned_u64 line_info; /* line info */ 387 __u32 line_info_cnt; /* number of bpf_line_info records */ 388 }; 389 390 struct { /* anonymous struct used by BPF_OBJ_* commands */ 391 __aligned_u64 pathname; 392 __u32 bpf_fd; 393 __u32 file_flags; 394 }; 395 396 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 397 __u32 target_fd; /* container object to attach to */ 398 __u32 attach_bpf_fd; /* eBPF program to attach */ 399 __u32 attach_type; 400 __u32 attach_flags; 401 }; 402 403 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 404 __u32 prog_fd; 405 __u32 retval; 406 __u32 data_size_in; /* input: len of data_in */ 407 __u32 data_size_out; /* input/output: len of data_out 408 * returns ENOSPC if data_out 409 * is too small. 410 */ 411 __aligned_u64 data_in; 412 __aligned_u64 data_out; 413 __u32 repeat; 414 __u32 duration; 415 __u32 ctx_size_in; /* input: len of ctx_in */ 416 __u32 ctx_size_out; /* input/output: len of ctx_out 417 * returns ENOSPC if ctx_out 418 * is too small. 419 */ 420 __aligned_u64 ctx_in; 421 __aligned_u64 ctx_out; 422 } test; 423 424 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 425 union { 426 __u32 start_id; 427 __u32 prog_id; 428 __u32 map_id; 429 __u32 btf_id; 430 }; 431 __u32 next_id; 432 __u32 open_flags; 433 }; 434 435 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 436 __u32 bpf_fd; 437 __u32 info_len; 438 __aligned_u64 info; 439 } info; 440 441 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 442 __u32 target_fd; /* container object to query */ 443 __u32 attach_type; 444 __u32 query_flags; 445 __u32 attach_flags; 446 __aligned_u64 prog_ids; 447 __u32 prog_cnt; 448 } query; 449 450 struct { 451 __u64 name; 452 __u32 prog_fd; 453 } raw_tracepoint; 454 455 struct { /* anonymous struct for BPF_BTF_LOAD */ 456 __aligned_u64 btf; 457 __aligned_u64 btf_log_buf; 458 __u32 btf_size; 459 __u32 btf_log_size; 460 __u32 btf_log_level; 461 }; 462 463 struct { 464 __u32 pid; /* input: pid */ 465 __u32 fd; /* input: fd */ 466 __u32 flags; /* input: flags */ 467 __u32 buf_len; /* input/output: buf len */ 468 __aligned_u64 buf; /* input/output: 469 * tp_name for tracepoint 470 * symbol for kprobe 471 * filename for uprobe 472 */ 473 __u32 prog_id; /* output: prod_id */ 474 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 475 __u64 probe_offset; /* output: probe_offset */ 476 __u64 probe_addr; /* output: probe_addr */ 477 } task_fd_query; 478 } __attribute__((aligned(8))); 479 480 /* The description below is an attempt at providing documentation to eBPF 481 * developers about the multiple available eBPF helper functions. It can be 482 * parsed and used to produce a manual page. The workflow is the following, 483 * and requires the rst2man utility: 484 * 485 * $ ./scripts/bpf_helpers_doc.py \ 486 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 487 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 488 * $ man /tmp/bpf-helpers.7 489 * 490 * Note that in order to produce this external documentation, some RST 491 * formatting is used in the descriptions to get "bold" and "italics" in 492 * manual pages. Also note that the few trailing white spaces are 493 * intentional, removing them would break paragraphs for rst2man. 494 * 495 * Start of BPF helper function descriptions: 496 * 497 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 498 * Description 499 * Perform a lookup in *map* for an entry associated to *key*. 500 * Return 501 * Map value associated to *key*, or **NULL** if no entry was 502 * found. 503 * 504 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 505 * Description 506 * Add or update the value of the entry associated to *key* in 507 * *map* with *value*. *flags* is one of: 508 * 509 * **BPF_NOEXIST** 510 * The entry for *key* must not exist in the map. 511 * **BPF_EXIST** 512 * The entry for *key* must already exist in the map. 513 * **BPF_ANY** 514 * No condition on the existence of the entry for *key*. 515 * 516 * Flag value **BPF_NOEXIST** cannot be used for maps of types 517 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 518 * elements always exist), the helper would return an error. 519 * Return 520 * 0 on success, or a negative error in case of failure. 521 * 522 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 523 * Description 524 * Delete entry with *key* from *map*. 525 * Return 526 * 0 on success, or a negative error in case of failure. 527 * 528 * int bpf_probe_read(void *dst, u32 size, const void *src) 529 * Description 530 * For tracing programs, safely attempt to read *size* bytes from 531 * address *src* and store the data in *dst*. 532 * Return 533 * 0 on success, or a negative error in case of failure. 534 * 535 * u64 bpf_ktime_get_ns(void) 536 * Description 537 * Return the time elapsed since system boot, in nanoseconds. 538 * Return 539 * Current *ktime*. 540 * 541 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 542 * Description 543 * This helper is a "printk()-like" facility for debugging. It 544 * prints a message defined by format *fmt* (of size *fmt_size*) 545 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 546 * available. It can take up to three additional **u64** 547 * arguments (as an eBPF helpers, the total number of arguments is 548 * limited to five). 549 * 550 * Each time the helper is called, it appends a line to the trace. 551 * The format of the trace is customizable, and the exact output 552 * one will get depends on the options set in 553 * *\/sys/kernel/debug/tracing/trace_options* (see also the 554 * *README* file under the same directory). However, it usually 555 * defaults to something like: 556 * 557 * :: 558 * 559 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 560 * 561 * In the above: 562 * 563 * * ``telnet`` is the name of the current task. 564 * * ``470`` is the PID of the current task. 565 * * ``001`` is the CPU number on which the task is 566 * running. 567 * * In ``.N..``, each character refers to a set of 568 * options (whether irqs are enabled, scheduling 569 * options, whether hard/softirqs are running, level of 570 * preempt_disabled respectively). **N** means that 571 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 572 * are set. 573 * * ``419421.045894`` is a timestamp. 574 * * ``0x00000001`` is a fake value used by BPF for the 575 * instruction pointer register. 576 * * ``<formatted msg>`` is the message formatted with 577 * *fmt*. 578 * 579 * The conversion specifiers supported by *fmt* are similar, but 580 * more limited than for printk(). They are **%d**, **%i**, 581 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 582 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 583 * of field, padding with zeroes, etc.) is available, and the 584 * helper will return **-EINVAL** (but print nothing) if it 585 * encounters an unknown specifier. 586 * 587 * Also, note that **bpf_trace_printk**\ () is slow, and should 588 * only be used for debugging purposes. For this reason, a notice 589 * bloc (spanning several lines) is printed to kernel logs and 590 * states that the helper should not be used "for production use" 591 * the first time this helper is used (or more precisely, when 592 * **trace_printk**\ () buffers are allocated). For passing values 593 * to user space, perf events should be preferred. 594 * Return 595 * The number of bytes written to the buffer, or a negative error 596 * in case of failure. 597 * 598 * u32 bpf_get_prandom_u32(void) 599 * Description 600 * Get a pseudo-random number. 601 * 602 * From a security point of view, this helper uses its own 603 * pseudo-random internal state, and cannot be used to infer the 604 * seed of other random functions in the kernel. However, it is 605 * essential to note that the generator used by the helper is not 606 * cryptographically secure. 607 * Return 608 * A random 32-bit unsigned value. 609 * 610 * u32 bpf_get_smp_processor_id(void) 611 * Description 612 * Get the SMP (symmetric multiprocessing) processor id. Note that 613 * all programs run with preemption disabled, which means that the 614 * SMP processor id is stable during all the execution of the 615 * program. 616 * Return 617 * The SMP id of the processor running the program. 618 * 619 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 620 * Description 621 * Store *len* bytes from address *from* into the packet 622 * associated to *skb*, at *offset*. *flags* are a combination of 623 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 624 * checksum for the packet after storing the bytes) and 625 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 626 * **->swhash** and *skb*\ **->l4hash** to 0). 627 * 628 * A call to this helper is susceptible to change the underlaying 629 * packet buffer. Therefore, at load time, all checks on pointers 630 * previously done by the verifier are invalidated and must be 631 * performed again, if the helper is used in combination with 632 * direct packet access. 633 * Return 634 * 0 on success, or a negative error in case of failure. 635 * 636 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 637 * Description 638 * Recompute the layer 3 (e.g. IP) checksum for the packet 639 * associated to *skb*. Computation is incremental, so the helper 640 * must know the former value of the header field that was 641 * modified (*from*), the new value of this field (*to*), and the 642 * number of bytes (2 or 4) for this field, stored in *size*. 643 * Alternatively, it is possible to store the difference between 644 * the previous and the new values of the header field in *to*, by 645 * setting *from* and *size* to 0. For both methods, *offset* 646 * indicates the location of the IP checksum within the packet. 647 * 648 * This helper works in combination with **bpf_csum_diff**\ (), 649 * which does not update the checksum in-place, but offers more 650 * flexibility and can handle sizes larger than 2 or 4 for the 651 * checksum to update. 652 * 653 * A call to this helper is susceptible to change the underlaying 654 * packet buffer. Therefore, at load time, all checks on pointers 655 * previously done by the verifier are invalidated and must be 656 * performed again, if the helper is used in combination with 657 * direct packet access. 658 * Return 659 * 0 on success, or a negative error in case of failure. 660 * 661 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 662 * Description 663 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 664 * packet associated to *skb*. Computation is incremental, so the 665 * helper must know the former value of the header field that was 666 * modified (*from*), the new value of this field (*to*), and the 667 * number of bytes (2 or 4) for this field, stored on the lowest 668 * four bits of *flags*. Alternatively, it is possible to store 669 * the difference between the previous and the new values of the 670 * header field in *to*, by setting *from* and the four lowest 671 * bits of *flags* to 0. For both methods, *offset* indicates the 672 * location of the IP checksum within the packet. In addition to 673 * the size of the field, *flags* can be added (bitwise OR) actual 674 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 675 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 676 * for updates resulting in a null checksum the value is set to 677 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 678 * the checksum is to be computed against a pseudo-header. 679 * 680 * This helper works in combination with **bpf_csum_diff**\ (), 681 * which does not update the checksum in-place, but offers more 682 * flexibility and can handle sizes larger than 2 or 4 for the 683 * checksum to update. 684 * 685 * A call to this helper is susceptible to change the underlaying 686 * packet buffer. Therefore, at load time, all checks on pointers 687 * previously done by the verifier are invalidated and must be 688 * performed again, if the helper is used in combination with 689 * direct packet access. 690 * Return 691 * 0 on success, or a negative error in case of failure. 692 * 693 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 694 * Description 695 * This special helper is used to trigger a "tail call", or in 696 * other words, to jump into another eBPF program. The same stack 697 * frame is used (but values on stack and in registers for the 698 * caller are not accessible to the callee). This mechanism allows 699 * for program chaining, either for raising the maximum number of 700 * available eBPF instructions, or to execute given programs in 701 * conditional blocks. For security reasons, there is an upper 702 * limit to the number of successive tail calls that can be 703 * performed. 704 * 705 * Upon call of this helper, the program attempts to jump into a 706 * program referenced at index *index* in *prog_array_map*, a 707 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 708 * *ctx*, a pointer to the context. 709 * 710 * If the call succeeds, the kernel immediately runs the first 711 * instruction of the new program. This is not a function call, 712 * and it never returns to the previous program. If the call 713 * fails, then the helper has no effect, and the caller continues 714 * to run its subsequent instructions. A call can fail if the 715 * destination program for the jump does not exist (i.e. *index* 716 * is superior to the number of entries in *prog_array_map*), or 717 * if the maximum number of tail calls has been reached for this 718 * chain of programs. This limit is defined in the kernel by the 719 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 720 * which is currently set to 32. 721 * Return 722 * 0 on success, or a negative error in case of failure. 723 * 724 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 725 * Description 726 * Clone and redirect the packet associated to *skb* to another 727 * net device of index *ifindex*. Both ingress and egress 728 * interfaces can be used for redirection. The **BPF_F_INGRESS** 729 * value in *flags* is used to make the distinction (ingress path 730 * is selected if the flag is present, egress path otherwise). 731 * This is the only flag supported for now. 732 * 733 * In comparison with **bpf_redirect**\ () helper, 734 * **bpf_clone_redirect**\ () has the associated cost of 735 * duplicating the packet buffer, but this can be executed out of 736 * the eBPF program. Conversely, **bpf_redirect**\ () is more 737 * efficient, but it is handled through an action code where the 738 * redirection happens only after the eBPF program has returned. 739 * 740 * A call to this helper is susceptible to change the underlaying 741 * packet buffer. Therefore, at load time, all checks on pointers 742 * previously done by the verifier are invalidated and must be 743 * performed again, if the helper is used in combination with 744 * direct packet access. 745 * Return 746 * 0 on success, or a negative error in case of failure. 747 * 748 * u64 bpf_get_current_pid_tgid(void) 749 * Return 750 * A 64-bit integer containing the current tgid and pid, and 751 * created as such: 752 * *current_task*\ **->tgid << 32 \|** 753 * *current_task*\ **->pid**. 754 * 755 * u64 bpf_get_current_uid_gid(void) 756 * Return 757 * A 64-bit integer containing the current GID and UID, and 758 * created as such: *current_gid* **<< 32 \|** *current_uid*. 759 * 760 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 761 * Description 762 * Copy the **comm** attribute of the current task into *buf* of 763 * *size_of_buf*. The **comm** attribute contains the name of 764 * the executable (excluding the path) for the current task. The 765 * *size_of_buf* must be strictly positive. On success, the 766 * helper makes sure that the *buf* is NUL-terminated. On failure, 767 * it is filled with zeroes. 768 * Return 769 * 0 on success, or a negative error in case of failure. 770 * 771 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 772 * Description 773 * Retrieve the classid for the current task, i.e. for the net_cls 774 * cgroup to which *skb* belongs. 775 * 776 * This helper can be used on TC egress path, but not on ingress. 777 * 778 * The net_cls cgroup provides an interface to tag network packets 779 * based on a user-provided identifier for all traffic coming from 780 * the tasks belonging to the related cgroup. See also the related 781 * kernel documentation, available from the Linux sources in file 782 * *Documentation/cgroup-v1/net_cls.txt*. 783 * 784 * The Linux kernel has two versions for cgroups: there are 785 * cgroups v1 and cgroups v2. Both are available to users, who can 786 * use a mixture of them, but note that the net_cls cgroup is for 787 * cgroup v1 only. This makes it incompatible with BPF programs 788 * run on cgroups, which is a cgroup-v2-only feature (a socket can 789 * only hold data for one version of cgroups at a time). 790 * 791 * This helper is only available is the kernel was compiled with 792 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 793 * "**y**" or to "**m**". 794 * Return 795 * The classid, or 0 for the default unconfigured classid. 796 * 797 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 798 * Description 799 * Push a *vlan_tci* (VLAN tag control information) of protocol 800 * *vlan_proto* to the packet associated to *skb*, then update 801 * the checksum. Note that if *vlan_proto* is different from 802 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 803 * be **ETH_P_8021Q**. 804 * 805 * A call to this helper is susceptible to change the underlaying 806 * packet buffer. Therefore, at load time, all checks on pointers 807 * previously done by the verifier are invalidated and must be 808 * performed again, if the helper is used in combination with 809 * direct packet access. 810 * Return 811 * 0 on success, or a negative error in case of failure. 812 * 813 * int bpf_skb_vlan_pop(struct sk_buff *skb) 814 * Description 815 * Pop a VLAN header from the packet associated to *skb*. 816 * 817 * A call to this helper is susceptible to change the underlaying 818 * packet buffer. Therefore, at load time, all checks on pointers 819 * previously done by the verifier are invalidated and must be 820 * performed again, if the helper is used in combination with 821 * direct packet access. 822 * Return 823 * 0 on success, or a negative error in case of failure. 824 * 825 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 826 * Description 827 * Get tunnel metadata. This helper takes a pointer *key* to an 828 * empty **struct bpf_tunnel_key** of **size**, that will be 829 * filled with tunnel metadata for the packet associated to *skb*. 830 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 831 * indicates that the tunnel is based on IPv6 protocol instead of 832 * IPv4. 833 * 834 * The **struct bpf_tunnel_key** is an object that generalizes the 835 * principal parameters used by various tunneling protocols into a 836 * single struct. This way, it can be used to easily make a 837 * decision based on the contents of the encapsulation header, 838 * "summarized" in this struct. In particular, it holds the IP 839 * address of the remote end (IPv4 or IPv6, depending on the case) 840 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 841 * this struct exposes the *key*\ **->tunnel_id**, which is 842 * generally mapped to a VNI (Virtual Network Identifier), making 843 * it programmable together with the **bpf_skb_set_tunnel_key**\ 844 * () helper. 845 * 846 * Let's imagine that the following code is part of a program 847 * attached to the TC ingress interface, on one end of a GRE 848 * tunnel, and is supposed to filter out all messages coming from 849 * remote ends with IPv4 address other than 10.0.0.1: 850 * 851 * :: 852 * 853 * int ret; 854 * struct bpf_tunnel_key key = {}; 855 * 856 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 857 * if (ret < 0) 858 * return TC_ACT_SHOT; // drop packet 859 * 860 * if (key.remote_ipv4 != 0x0a000001) 861 * return TC_ACT_SHOT; // drop packet 862 * 863 * return TC_ACT_OK; // accept packet 864 * 865 * This interface can also be used with all encapsulation devices 866 * that can operate in "collect metadata" mode: instead of having 867 * one network device per specific configuration, the "collect 868 * metadata" mode only requires a single device where the 869 * configuration can be extracted from this helper. 870 * 871 * This can be used together with various tunnels such as VXLan, 872 * Geneve, GRE or IP in IP (IPIP). 873 * Return 874 * 0 on success, or a negative error in case of failure. 875 * 876 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 877 * Description 878 * Populate tunnel metadata for packet associated to *skb.* The 879 * tunnel metadata is set to the contents of *key*, of *size*. The 880 * *flags* can be set to a combination of the following values: 881 * 882 * **BPF_F_TUNINFO_IPV6** 883 * Indicate that the tunnel is based on IPv6 protocol 884 * instead of IPv4. 885 * **BPF_F_ZERO_CSUM_TX** 886 * For IPv4 packets, add a flag to tunnel metadata 887 * indicating that checksum computation should be skipped 888 * and checksum set to zeroes. 889 * **BPF_F_DONT_FRAGMENT** 890 * Add a flag to tunnel metadata indicating that the 891 * packet should not be fragmented. 892 * **BPF_F_SEQ_NUMBER** 893 * Add a flag to tunnel metadata indicating that a 894 * sequence number should be added to tunnel header before 895 * sending the packet. This flag was added for GRE 896 * encapsulation, but might be used with other protocols 897 * as well in the future. 898 * 899 * Here is a typical usage on the transmit path: 900 * 901 * :: 902 * 903 * struct bpf_tunnel_key key; 904 * populate key ... 905 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 906 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 907 * 908 * See also the description of the **bpf_skb_get_tunnel_key**\ () 909 * helper for additional information. 910 * Return 911 * 0 on success, or a negative error in case of failure. 912 * 913 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 914 * Description 915 * Read the value of a perf event counter. This helper relies on a 916 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 917 * the perf event counter is selected when *map* is updated with 918 * perf event file descriptors. The *map* is an array whose size 919 * is the number of available CPUs, and each cell contains a value 920 * relative to one CPU. The value to retrieve is indicated by 921 * *flags*, that contains the index of the CPU to look up, masked 922 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 923 * **BPF_F_CURRENT_CPU** to indicate that the value for the 924 * current CPU should be retrieved. 925 * 926 * Note that before Linux 4.13, only hardware perf event can be 927 * retrieved. 928 * 929 * Also, be aware that the newer helper 930 * **bpf_perf_event_read_value**\ () is recommended over 931 * **bpf_perf_event_read**\ () in general. The latter has some ABI 932 * quirks where error and counter value are used as a return code 933 * (which is wrong to do since ranges may overlap). This issue is 934 * fixed with **bpf_perf_event_read_value**\ (), which at the same 935 * time provides more features over the **bpf_perf_event_read**\ 936 * () interface. Please refer to the description of 937 * **bpf_perf_event_read_value**\ () for details. 938 * Return 939 * The value of the perf event counter read from the map, or a 940 * negative error code in case of failure. 941 * 942 * int bpf_redirect(u32 ifindex, u64 flags) 943 * Description 944 * Redirect the packet to another net device of index *ifindex*. 945 * This helper is somewhat similar to **bpf_clone_redirect**\ 946 * (), except that the packet is not cloned, which provides 947 * increased performance. 948 * 949 * Except for XDP, both ingress and egress interfaces can be used 950 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 951 * to make the distinction (ingress path is selected if the flag 952 * is present, egress path otherwise). Currently, XDP only 953 * supports redirection to the egress interface, and accepts no 954 * flag at all. 955 * 956 * The same effect can be attained with the more generic 957 * **bpf_redirect_map**\ (), which requires specific maps to be 958 * used but offers better performance. 959 * Return 960 * For XDP, the helper returns **XDP_REDIRECT** on success or 961 * **XDP_ABORTED** on error. For other program types, the values 962 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 963 * error. 964 * 965 * u32 bpf_get_route_realm(struct sk_buff *skb) 966 * Description 967 * Retrieve the realm or the route, that is to say the 968 * **tclassid** field of the destination for the *skb*. The 969 * indentifier retrieved is a user-provided tag, similar to the 970 * one used with the net_cls cgroup (see description for 971 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 972 * held by a route (a destination entry), not by a task. 973 * 974 * Retrieving this identifier works with the clsact TC egress hook 975 * (see also **tc-bpf(8)**), or alternatively on conventional 976 * classful egress qdiscs, but not on TC ingress path. In case of 977 * clsact TC egress hook, this has the advantage that, internally, 978 * the destination entry has not been dropped yet in the transmit 979 * path. Therefore, the destination entry does not need to be 980 * artificially held via **netif_keep_dst**\ () for a classful 981 * qdisc until the *skb* is freed. 982 * 983 * This helper is available only if the kernel was compiled with 984 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 985 * Return 986 * The realm of the route for the packet associated to *skb*, or 0 987 * if none was found. 988 * 989 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 990 * Description 991 * Write raw *data* blob into a special BPF perf event held by 992 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 993 * event must have the following attributes: **PERF_SAMPLE_RAW** 994 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 995 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 996 * 997 * The *flags* are used to indicate the index in *map* for which 998 * the value must be put, masked with **BPF_F_INDEX_MASK**. 999 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1000 * to indicate that the index of the current CPU core should be 1001 * used. 1002 * 1003 * The value to write, of *size*, is passed through eBPF stack and 1004 * pointed by *data*. 1005 * 1006 * The context of the program *ctx* needs also be passed to the 1007 * helper. 1008 * 1009 * On user space, a program willing to read the values needs to 1010 * call **perf_event_open**\ () on the perf event (either for 1011 * one or for all CPUs) and to store the file descriptor into the 1012 * *map*. This must be done before the eBPF program can send data 1013 * into it. An example is available in file 1014 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1015 * tree (the eBPF program counterpart is in 1016 * *samples/bpf/trace_output_kern.c*). 1017 * 1018 * **bpf_perf_event_output**\ () achieves better performance 1019 * than **bpf_trace_printk**\ () for sharing data with user 1020 * space, and is much better suitable for streaming data from eBPF 1021 * programs. 1022 * 1023 * Note that this helper is not restricted to tracing use cases 1024 * and can be used with programs attached to TC or XDP as well, 1025 * where it allows for passing data to user space listeners. Data 1026 * can be: 1027 * 1028 * * Only custom structs, 1029 * * Only the packet payload, or 1030 * * A combination of both. 1031 * Return 1032 * 0 on success, or a negative error in case of failure. 1033 * 1034 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 1035 * Description 1036 * This helper was provided as an easy way to load data from a 1037 * packet. It can be used to load *len* bytes from *offset* from 1038 * the packet associated to *skb*, into the buffer pointed by 1039 * *to*. 1040 * 1041 * Since Linux 4.7, usage of this helper has mostly been replaced 1042 * by "direct packet access", enabling packet data to be 1043 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1044 * pointing respectively to the first byte of packet data and to 1045 * the byte after the last byte of packet data. However, it 1046 * remains useful if one wishes to read large quantities of data 1047 * at once from a packet into the eBPF stack. 1048 * Return 1049 * 0 on success, or a negative error in case of failure. 1050 * 1051 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 1052 * Description 1053 * Walk a user or a kernel stack and return its id. To achieve 1054 * this, the helper needs *ctx*, which is a pointer to the context 1055 * on which the tracing program is executed, and a pointer to a 1056 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1057 * 1058 * The last argument, *flags*, holds the number of stack frames to 1059 * skip (from 0 to 255), masked with 1060 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1061 * a combination of the following flags: 1062 * 1063 * **BPF_F_USER_STACK** 1064 * Collect a user space stack instead of a kernel stack. 1065 * **BPF_F_FAST_STACK_CMP** 1066 * Compare stacks by hash only. 1067 * **BPF_F_REUSE_STACKID** 1068 * If two different stacks hash into the same *stackid*, 1069 * discard the old one. 1070 * 1071 * The stack id retrieved is a 32 bit long integer handle which 1072 * can be further combined with other data (including other stack 1073 * ids) and used as a key into maps. This can be useful for 1074 * generating a variety of graphs (such as flame graphs or off-cpu 1075 * graphs). 1076 * 1077 * For walking a stack, this helper is an improvement over 1078 * **bpf_probe_read**\ (), which can be used with unrolled loops 1079 * but is not efficient and consumes a lot of eBPF instructions. 1080 * Instead, **bpf_get_stackid**\ () can collect up to 1081 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1082 * this limit can be controlled with the **sysctl** program, and 1083 * that it should be manually increased in order to profile long 1084 * user stacks (such as stacks for Java programs). To do so, use: 1085 * 1086 * :: 1087 * 1088 * # sysctl kernel.perf_event_max_stack=<new value> 1089 * Return 1090 * The positive or null stack id on success, or a negative error 1091 * in case of failure. 1092 * 1093 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1094 * Description 1095 * Compute a checksum difference, from the raw buffer pointed by 1096 * *from*, of length *from_size* (that must be a multiple of 4), 1097 * towards the raw buffer pointed by *to*, of size *to_size* 1098 * (same remark). An optional *seed* can be added to the value 1099 * (this can be cascaded, the seed may come from a previous call 1100 * to the helper). 1101 * 1102 * This is flexible enough to be used in several ways: 1103 * 1104 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1105 * checksum, it can be used when pushing new data. 1106 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1107 * checksum, it can be used when removing data from a packet. 1108 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1109 * can be used to compute a diff. Note that *from_size* and 1110 * *to_size* do not need to be equal. 1111 * 1112 * This helper can be used in combination with 1113 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1114 * which one can feed in the difference computed with 1115 * **bpf_csum_diff**\ (). 1116 * Return 1117 * The checksum result, or a negative error code in case of 1118 * failure. 1119 * 1120 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1121 * Description 1122 * Retrieve tunnel options metadata for the packet associated to 1123 * *skb*, and store the raw tunnel option data to the buffer *opt* 1124 * of *size*. 1125 * 1126 * This helper can be used with encapsulation devices that can 1127 * operate in "collect metadata" mode (please refer to the related 1128 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1129 * more details). A particular example where this can be used is 1130 * in combination with the Geneve encapsulation protocol, where it 1131 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1132 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1133 * the eBPF program. This allows for full customization of these 1134 * headers. 1135 * Return 1136 * The size of the option data retrieved. 1137 * 1138 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1139 * Description 1140 * Set tunnel options metadata for the packet associated to *skb* 1141 * to the option data contained in the raw buffer *opt* of *size*. 1142 * 1143 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1144 * helper for additional information. 1145 * Return 1146 * 0 on success, or a negative error in case of failure. 1147 * 1148 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1149 * Description 1150 * Change the protocol of the *skb* to *proto*. Currently 1151 * supported are transition from IPv4 to IPv6, and from IPv6 to 1152 * IPv4. The helper takes care of the groundwork for the 1153 * transition, including resizing the socket buffer. The eBPF 1154 * program is expected to fill the new headers, if any, via 1155 * **skb_store_bytes**\ () and to recompute the checksums with 1156 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1157 * (). The main case for this helper is to perform NAT64 1158 * operations out of an eBPF program. 1159 * 1160 * Internally, the GSO type is marked as dodgy so that headers are 1161 * checked and segments are recalculated by the GSO/GRO engine. 1162 * The size for GSO target is adapted as well. 1163 * 1164 * All values for *flags* are reserved for future usage, and must 1165 * be left at zero. 1166 * 1167 * A call to this helper is susceptible to change the underlaying 1168 * packet buffer. Therefore, at load time, all checks on pointers 1169 * previously done by the verifier are invalidated and must be 1170 * performed again, if the helper is used in combination with 1171 * direct packet access. 1172 * Return 1173 * 0 on success, or a negative error in case of failure. 1174 * 1175 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1176 * Description 1177 * Change the packet type for the packet associated to *skb*. This 1178 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1179 * the eBPF program does not have a write access to *skb*\ 1180 * **->pkt_type** beside this helper. Using a helper here allows 1181 * for graceful handling of errors. 1182 * 1183 * The major use case is to change incoming *skb*s to 1184 * **PACKET_HOST** in a programmatic way instead of having to 1185 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1186 * example. 1187 * 1188 * Note that *type* only allows certain values. At this time, they 1189 * are: 1190 * 1191 * **PACKET_HOST** 1192 * Packet is for us. 1193 * **PACKET_BROADCAST** 1194 * Send packet to all. 1195 * **PACKET_MULTICAST** 1196 * Send packet to group. 1197 * **PACKET_OTHERHOST** 1198 * Send packet to someone else. 1199 * Return 1200 * 0 on success, or a negative error in case of failure. 1201 * 1202 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1203 * Description 1204 * Check whether *skb* is a descendant of the cgroup2 held by 1205 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1206 * Return 1207 * The return value depends on the result of the test, and can be: 1208 * 1209 * * 0, if the *skb* failed the cgroup2 descendant test. 1210 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1211 * * A negative error code, if an error occurred. 1212 * 1213 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1214 * Description 1215 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1216 * not set, in particular if the hash was cleared due to mangling, 1217 * recompute this hash. Later accesses to the hash can be done 1218 * directly with *skb*\ **->hash**. 1219 * 1220 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1221 * prototype with **bpf_skb_change_proto**\ (), or calling 1222 * **bpf_skb_store_bytes**\ () with the 1223 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1224 * the hash and to trigger a new computation for the next call to 1225 * **bpf_get_hash_recalc**\ (). 1226 * Return 1227 * The 32-bit hash. 1228 * 1229 * u64 bpf_get_current_task(void) 1230 * Return 1231 * A pointer to the current task struct. 1232 * 1233 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1234 * Description 1235 * Attempt in a safe way to write *len* bytes from the buffer 1236 * *src* to *dst* in memory. It only works for threads that are in 1237 * user context, and *dst* must be a valid user space address. 1238 * 1239 * This helper should not be used to implement any kind of 1240 * security mechanism because of TOC-TOU attacks, but rather to 1241 * debug, divert, and manipulate execution of semi-cooperative 1242 * processes. 1243 * 1244 * Keep in mind that this feature is meant for experiments, and it 1245 * has a risk of crashing the system and running programs. 1246 * Therefore, when an eBPF program using this helper is attached, 1247 * a warning including PID and process name is printed to kernel 1248 * logs. 1249 * Return 1250 * 0 on success, or a negative error in case of failure. 1251 * 1252 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1253 * Description 1254 * Check whether the probe is being run is the context of a given 1255 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1256 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1257 * Return 1258 * The return value depends on the result of the test, and can be: 1259 * 1260 * * 0, if the *skb* task belongs to the cgroup2. 1261 * * 1, if the *skb* task does not belong to the cgroup2. 1262 * * A negative error code, if an error occurred. 1263 * 1264 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1265 * Description 1266 * Resize (trim or grow) the packet associated to *skb* to the 1267 * new *len*. The *flags* are reserved for future usage, and must 1268 * be left at zero. 1269 * 1270 * The basic idea is that the helper performs the needed work to 1271 * change the size of the packet, then the eBPF program rewrites 1272 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1273 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1274 * and others. This helper is a slow path utility intended for 1275 * replies with control messages. And because it is targeted for 1276 * slow path, the helper itself can afford to be slow: it 1277 * implicitly linearizes, unclones and drops offloads from the 1278 * *skb*. 1279 * 1280 * A call to this helper is susceptible to change the underlaying 1281 * packet buffer. Therefore, at load time, all checks on pointers 1282 * previously done by the verifier are invalidated and must be 1283 * performed again, if the helper is used in combination with 1284 * direct packet access. 1285 * Return 1286 * 0 on success, or a negative error in case of failure. 1287 * 1288 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1289 * Description 1290 * Pull in non-linear data in case the *skb* is non-linear and not 1291 * all of *len* are part of the linear section. Make *len* bytes 1292 * from *skb* readable and writable. If a zero value is passed for 1293 * *len*, then the whole length of the *skb* is pulled. 1294 * 1295 * This helper is only needed for reading and writing with direct 1296 * packet access. 1297 * 1298 * For direct packet access, testing that offsets to access 1299 * are within packet boundaries (test on *skb*\ **->data_end**) is 1300 * susceptible to fail if offsets are invalid, or if the requested 1301 * data is in non-linear parts of the *skb*. On failure the 1302 * program can just bail out, or in the case of a non-linear 1303 * buffer, use a helper to make the data available. The 1304 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1305 * the data. Another one consists in using **bpf_skb_pull_data** 1306 * to pull in once the non-linear parts, then retesting and 1307 * eventually access the data. 1308 * 1309 * At the same time, this also makes sure the *skb* is uncloned, 1310 * which is a necessary condition for direct write. As this needs 1311 * to be an invariant for the write part only, the verifier 1312 * detects writes and adds a prologue that is calling 1313 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1314 * the very beginning in case it is indeed cloned. 1315 * 1316 * A call to this helper is susceptible to change the underlaying 1317 * packet buffer. Therefore, at load time, all checks on pointers 1318 * previously done by the verifier are invalidated and must be 1319 * performed again, if the helper is used in combination with 1320 * direct packet access. 1321 * Return 1322 * 0 on success, or a negative error in case of failure. 1323 * 1324 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1325 * Description 1326 * Add the checksum *csum* into *skb*\ **->csum** in case the 1327 * driver has supplied a checksum for the entire packet into that 1328 * field. Return an error otherwise. This helper is intended to be 1329 * used in combination with **bpf_csum_diff**\ (), in particular 1330 * when the checksum needs to be updated after data has been 1331 * written into the packet through direct packet access. 1332 * Return 1333 * The checksum on success, or a negative error code in case of 1334 * failure. 1335 * 1336 * void bpf_set_hash_invalid(struct sk_buff *skb) 1337 * Description 1338 * Invalidate the current *skb*\ **->hash**. It can be used after 1339 * mangling on headers through direct packet access, in order to 1340 * indicate that the hash is outdated and to trigger a 1341 * recalculation the next time the kernel tries to access this 1342 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1343 * 1344 * int bpf_get_numa_node_id(void) 1345 * Description 1346 * Return the id of the current NUMA node. The primary use case 1347 * for this helper is the selection of sockets for the local NUMA 1348 * node, when the program is attached to sockets using the 1349 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1350 * but the helper is also available to other eBPF program types, 1351 * similarly to **bpf_get_smp_processor_id**\ (). 1352 * Return 1353 * The id of current NUMA node. 1354 * 1355 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1356 * Description 1357 * Grows headroom of packet associated to *skb* and adjusts the 1358 * offset of the MAC header accordingly, adding *len* bytes of 1359 * space. It automatically extends and reallocates memory as 1360 * required. 1361 * 1362 * This helper can be used on a layer 3 *skb* to push a MAC header 1363 * for redirection into a layer 2 device. 1364 * 1365 * All values for *flags* are reserved for future usage, and must 1366 * be left at zero. 1367 * 1368 * A call to this helper is susceptible to change the underlaying 1369 * packet buffer. Therefore, at load time, all checks on pointers 1370 * previously done by the verifier are invalidated and must be 1371 * performed again, if the helper is used in combination with 1372 * direct packet access. 1373 * Return 1374 * 0 on success, or a negative error in case of failure. 1375 * 1376 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1377 * Description 1378 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1379 * it is possible to use a negative value for *delta*. This helper 1380 * can be used to prepare the packet for pushing or popping 1381 * headers. 1382 * 1383 * A call to this helper is susceptible to change the underlaying 1384 * packet buffer. Therefore, at load time, all checks on pointers 1385 * previously done by the verifier are invalidated and must be 1386 * performed again, if the helper is used in combination with 1387 * direct packet access. 1388 * Return 1389 * 0 on success, or a negative error in case of failure. 1390 * 1391 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1392 * Description 1393 * Copy a NUL terminated string from an unsafe address 1394 * *unsafe_ptr* to *dst*. The *size* should include the 1395 * terminating NUL byte. In case the string length is smaller than 1396 * *size*, the target is not padded with further NUL bytes. If the 1397 * string length is larger than *size*, just *size*-1 bytes are 1398 * copied and the last byte is set to NUL. 1399 * 1400 * On success, the length of the copied string is returned. This 1401 * makes this helper useful in tracing programs for reading 1402 * strings, and more importantly to get its length at runtime. See 1403 * the following snippet: 1404 * 1405 * :: 1406 * 1407 * SEC("kprobe/sys_open") 1408 * void bpf_sys_open(struct pt_regs *ctx) 1409 * { 1410 * char buf[PATHLEN]; // PATHLEN is defined to 256 1411 * int res = bpf_probe_read_str(buf, sizeof(buf), 1412 * ctx->di); 1413 * 1414 * // Consume buf, for example push it to 1415 * // userspace via bpf_perf_event_output(); we 1416 * // can use res (the string length) as event 1417 * // size, after checking its boundaries. 1418 * } 1419 * 1420 * In comparison, using **bpf_probe_read()** helper here instead 1421 * to read the string would require to estimate the length at 1422 * compile time, and would often result in copying more memory 1423 * than necessary. 1424 * 1425 * Another useful use case is when parsing individual process 1426 * arguments or individual environment variables navigating 1427 * *current*\ **->mm->arg_start** and *current*\ 1428 * **->mm->env_start**: using this helper and the return value, 1429 * one can quickly iterate at the right offset of the memory area. 1430 * Return 1431 * On success, the strictly positive length of the string, 1432 * including the trailing NUL character. On error, a negative 1433 * value. 1434 * 1435 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1436 * Description 1437 * If the **struct sk_buff** pointed by *skb* has a known socket, 1438 * retrieve the cookie (generated by the kernel) of this socket. 1439 * If no cookie has been set yet, generate a new cookie. Once 1440 * generated, the socket cookie remains stable for the life of the 1441 * socket. This helper can be useful for monitoring per socket 1442 * networking traffic statistics as it provides a unique socket 1443 * identifier per namespace. 1444 * Return 1445 * A 8-byte long non-decreasing number on success, or 0 if the 1446 * socket field is missing inside *skb*. 1447 * 1448 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1449 * Description 1450 * Equivalent to bpf_get_socket_cookie() helper that accepts 1451 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1452 * Return 1453 * A 8-byte long non-decreasing number. 1454 * 1455 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1456 * Description 1457 * Equivalent to bpf_get_socket_cookie() helper that accepts 1458 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1459 * Return 1460 * A 8-byte long non-decreasing number. 1461 * 1462 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1463 * Return 1464 * The owner UID of the socket associated to *skb*. If the socket 1465 * is **NULL**, or if it is not a full socket (i.e. if it is a 1466 * time-wait or a request socket instead), **overflowuid** value 1467 * is returned (note that **overflowuid** might also be the actual 1468 * UID value for the socket). 1469 * 1470 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1471 * Description 1472 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1473 * to value *hash*. 1474 * Return 1475 * 0 1476 * 1477 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1478 * Description 1479 * Emulate a call to **setsockopt()** on the socket associated to 1480 * *bpf_socket*, which must be a full socket. The *level* at 1481 * which the option resides and the name *optname* of the option 1482 * must be specified, see **setsockopt(2)** for more information. 1483 * The option value of length *optlen* is pointed by *optval*. 1484 * 1485 * This helper actually implements a subset of **setsockopt()**. 1486 * It supports the following *level*\ s: 1487 * 1488 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1489 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1490 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1491 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1492 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1493 * **TCP_BPF_SNDCWND_CLAMP**. 1494 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1495 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1496 * Return 1497 * 0 on success, or a negative error in case of failure. 1498 * 1499 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1500 * Description 1501 * Grow or shrink the room for data in the packet associated to 1502 * *skb* by *len_diff*, and according to the selected *mode*. 1503 * 1504 * There are two supported modes at this time: 1505 * 1506 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1507 * (room space is added or removed below the layer 2 header). 1508 * 1509 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1510 * (room space is added or removed below the layer 3 header). 1511 * 1512 * The following flags are supported at this time: 1513 * 1514 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1515 * Adjusting mss in this way is not allowed for datagrams. 1516 * 1517 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 **: 1518 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 **: 1519 * Any new space is reserved to hold a tunnel header. 1520 * Configure skb offsets and other fields accordingly. 1521 * 1522 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE **: 1523 * * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **: 1524 * Use with ENCAP_L3 flags to further specify the tunnel type. 1525 * 1526 * * **BPF_F_ADJ_ROOM_ENCAP_L2(len) **: 1527 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1528 * type; **len** is the length of the inner MAC header. 1529 * 1530 * A call to this helper is susceptible to change the underlaying 1531 * packet buffer. Therefore, at load time, all checks on pointers 1532 * previously done by the verifier are invalidated and must be 1533 * performed again, if the helper is used in combination with 1534 * direct packet access. 1535 * Return 1536 * 0 on success, or a negative error in case of failure. 1537 * 1538 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1539 * Description 1540 * Redirect the packet to the endpoint referenced by *map* at 1541 * index *key*. Depending on its type, this *map* can contain 1542 * references to net devices (for forwarding packets through other 1543 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1544 * but this is only implemented for native XDP (with driver 1545 * support) as of this writing). 1546 * 1547 * All values for *flags* are reserved for future usage, and must 1548 * be left at zero. 1549 * 1550 * When used to redirect packets to net devices, this helper 1551 * provides a high performance increase over **bpf_redirect**\ (). 1552 * This is due to various implementation details of the underlying 1553 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1554 * () tries to send packet as a "bulk" to the device. 1555 * Return 1556 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1557 * 1558 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1559 * Description 1560 * Redirect the packet to the socket referenced by *map* (of type 1561 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1562 * egress interfaces can be used for redirection. The 1563 * **BPF_F_INGRESS** value in *flags* is used to make the 1564 * distinction (ingress path is selected if the flag is present, 1565 * egress path otherwise). This is the only flag supported for now. 1566 * Return 1567 * **SK_PASS** on success, or **SK_DROP** on error. 1568 * 1569 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1570 * Description 1571 * Add an entry to, or update a *map* referencing sockets. The 1572 * *skops* is used as a new value for the entry associated to 1573 * *key*. *flags* is one of: 1574 * 1575 * **BPF_NOEXIST** 1576 * The entry for *key* must not exist in the map. 1577 * **BPF_EXIST** 1578 * The entry for *key* must already exist in the map. 1579 * **BPF_ANY** 1580 * No condition on the existence of the entry for *key*. 1581 * 1582 * If the *map* has eBPF programs (parser and verdict), those will 1583 * be inherited by the socket being added. If the socket is 1584 * already attached to eBPF programs, this results in an error. 1585 * Return 1586 * 0 on success, or a negative error in case of failure. 1587 * 1588 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1589 * Description 1590 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1591 * *delta* (which can be positive or negative). Note that this 1592 * operation modifies the address stored in *xdp_md*\ **->data**, 1593 * so the latter must be loaded only after the helper has been 1594 * called. 1595 * 1596 * The use of *xdp_md*\ **->data_meta** is optional and programs 1597 * are not required to use it. The rationale is that when the 1598 * packet is processed with XDP (e.g. as DoS filter), it is 1599 * possible to push further meta data along with it before passing 1600 * to the stack, and to give the guarantee that an ingress eBPF 1601 * program attached as a TC classifier on the same device can pick 1602 * this up for further post-processing. Since TC works with socket 1603 * buffers, it remains possible to set from XDP the **mark** or 1604 * **priority** pointers, or other pointers for the socket buffer. 1605 * Having this scratch space generic and programmable allows for 1606 * more flexibility as the user is free to store whatever meta 1607 * data they need. 1608 * 1609 * A call to this helper is susceptible to change the underlaying 1610 * packet buffer. Therefore, at load time, all checks on pointers 1611 * previously done by the verifier are invalidated and must be 1612 * performed again, if the helper is used in combination with 1613 * direct packet access. 1614 * Return 1615 * 0 on success, or a negative error in case of failure. 1616 * 1617 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1618 * Description 1619 * Read the value of a perf event counter, and store it into *buf* 1620 * of size *buf_size*. This helper relies on a *map* of type 1621 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1622 * counter is selected when *map* is updated with perf event file 1623 * descriptors. The *map* is an array whose size is the number of 1624 * available CPUs, and each cell contains a value relative to one 1625 * CPU. The value to retrieve is indicated by *flags*, that 1626 * contains the index of the CPU to look up, masked with 1627 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1628 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1629 * current CPU should be retrieved. 1630 * 1631 * This helper behaves in a way close to 1632 * **bpf_perf_event_read**\ () helper, save that instead of 1633 * just returning the value observed, it fills the *buf* 1634 * structure. This allows for additional data to be retrieved: in 1635 * particular, the enabled and running times (in *buf*\ 1636 * **->enabled** and *buf*\ **->running**, respectively) are 1637 * copied. In general, **bpf_perf_event_read_value**\ () is 1638 * recommended over **bpf_perf_event_read**\ (), which has some 1639 * ABI issues and provides fewer functionalities. 1640 * 1641 * These values are interesting, because hardware PMU (Performance 1642 * Monitoring Unit) counters are limited resources. When there are 1643 * more PMU based perf events opened than available counters, 1644 * kernel will multiplex these events so each event gets certain 1645 * percentage (but not all) of the PMU time. In case that 1646 * multiplexing happens, the number of samples or counter value 1647 * will not reflect the case compared to when no multiplexing 1648 * occurs. This makes comparison between different runs difficult. 1649 * Typically, the counter value should be normalized before 1650 * comparing to other experiments. The usual normalization is done 1651 * as follows. 1652 * 1653 * :: 1654 * 1655 * normalized_counter = counter * t_enabled / t_running 1656 * 1657 * Where t_enabled is the time enabled for event and t_running is 1658 * the time running for event since last normalization. The 1659 * enabled and running times are accumulated since the perf event 1660 * open. To achieve scaling factor between two invocations of an 1661 * eBPF program, users can can use CPU id as the key (which is 1662 * typical for perf array usage model) to remember the previous 1663 * value and do the calculation inside the eBPF program. 1664 * Return 1665 * 0 on success, or a negative error in case of failure. 1666 * 1667 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1668 * Description 1669 * For en eBPF program attached to a perf event, retrieve the 1670 * value of the event counter associated to *ctx* and store it in 1671 * the structure pointed by *buf* and of size *buf_size*. Enabled 1672 * and running times are also stored in the structure (see 1673 * description of helper **bpf_perf_event_read_value**\ () for 1674 * more details). 1675 * Return 1676 * 0 on success, or a negative error in case of failure. 1677 * 1678 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1679 * Description 1680 * Emulate a call to **getsockopt()** on the socket associated to 1681 * *bpf_socket*, which must be a full socket. The *level* at 1682 * which the option resides and the name *optname* of the option 1683 * must be specified, see **getsockopt(2)** for more information. 1684 * The retrieved value is stored in the structure pointed by 1685 * *opval* and of length *optlen*. 1686 * 1687 * This helper actually implements a subset of **getsockopt()**. 1688 * It supports the following *level*\ s: 1689 * 1690 * * **IPPROTO_TCP**, which supports *optname* 1691 * **TCP_CONGESTION**. 1692 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1693 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1694 * Return 1695 * 0 on success, or a negative error in case of failure. 1696 * 1697 * int bpf_override_return(struct pt_reg *regs, u64 rc) 1698 * Description 1699 * Used for error injection, this helper uses kprobes to override 1700 * the return value of the probed function, and to set it to *rc*. 1701 * The first argument is the context *regs* on which the kprobe 1702 * works. 1703 * 1704 * This helper works by setting setting the PC (program counter) 1705 * to an override function which is run in place of the original 1706 * probed function. This means the probed function is not run at 1707 * all. The replacement function just returns with the required 1708 * value. 1709 * 1710 * This helper has security implications, and thus is subject to 1711 * restrictions. It is only available if the kernel was compiled 1712 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1713 * option, and in this case it only works on functions tagged with 1714 * **ALLOW_ERROR_INJECTION** in the kernel code. 1715 * 1716 * Also, the helper is only available for the architectures having 1717 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1718 * x86 architecture is the only one to support this feature. 1719 * Return 1720 * 0 1721 * 1722 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1723 * Description 1724 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1725 * for the full TCP socket associated to *bpf_sock_ops* to 1726 * *argval*. 1727 * 1728 * The primary use of this field is to determine if there should 1729 * be calls to eBPF programs of type 1730 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1731 * code. A program of the same type can change its value, per 1732 * connection and as necessary, when the connection is 1733 * established. This field is directly accessible for reading, but 1734 * this helper must be used for updates in order to return an 1735 * error if an eBPF program tries to set a callback that is not 1736 * supported in the current kernel. 1737 * 1738 * The supported callback values that *argval* can combine are: 1739 * 1740 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1741 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1742 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1743 * 1744 * Here are some examples of where one could call such eBPF 1745 * program: 1746 * 1747 * * When RTO fires. 1748 * * When a packet is retransmitted. 1749 * * When the connection terminates. 1750 * * When a packet is sent. 1751 * * When a packet is received. 1752 * Return 1753 * Code **-EINVAL** if the socket is not a full TCP socket; 1754 * otherwise, a positive number containing the bits that could not 1755 * be set is returned (which comes down to 0 if all bits were set 1756 * as required). 1757 * 1758 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1759 * Description 1760 * This helper is used in programs implementing policies at the 1761 * socket level. If the message *msg* is allowed to pass (i.e. if 1762 * the verdict eBPF program returns **SK_PASS**), redirect it to 1763 * the socket referenced by *map* (of type 1764 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1765 * egress interfaces can be used for redirection. The 1766 * **BPF_F_INGRESS** value in *flags* is used to make the 1767 * distinction (ingress path is selected if the flag is present, 1768 * egress path otherwise). This is the only flag supported for now. 1769 * Return 1770 * **SK_PASS** on success, or **SK_DROP** on error. 1771 * 1772 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1773 * Description 1774 * For socket policies, apply the verdict of the eBPF program to 1775 * the next *bytes* (number of bytes) of message *msg*. 1776 * 1777 * For example, this helper can be used in the following cases: 1778 * 1779 * * A single **sendmsg**\ () or **sendfile**\ () system call 1780 * contains multiple logical messages that the eBPF program is 1781 * supposed to read and for which it should apply a verdict. 1782 * * An eBPF program only cares to read the first *bytes* of a 1783 * *msg*. If the message has a large payload, then setting up 1784 * and calling the eBPF program repeatedly for all bytes, even 1785 * though the verdict is already known, would create unnecessary 1786 * overhead. 1787 * 1788 * When called from within an eBPF program, the helper sets a 1789 * counter internal to the BPF infrastructure, that is used to 1790 * apply the last verdict to the next *bytes*. If *bytes* is 1791 * smaller than the current data being processed from a 1792 * **sendmsg**\ () or **sendfile**\ () system call, the first 1793 * *bytes* will be sent and the eBPF program will be re-run with 1794 * the pointer for start of data pointing to byte number *bytes* 1795 * **+ 1**. If *bytes* is larger than the current data being 1796 * processed, then the eBPF verdict will be applied to multiple 1797 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1798 * consumed. 1799 * 1800 * Note that if a socket closes with the internal counter holding 1801 * a non-zero value, this is not a problem because data is not 1802 * being buffered for *bytes* and is sent as it is received. 1803 * Return 1804 * 0 1805 * 1806 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1807 * Description 1808 * For socket policies, prevent the execution of the verdict eBPF 1809 * program for message *msg* until *bytes* (byte number) have been 1810 * accumulated. 1811 * 1812 * This can be used when one needs a specific number of bytes 1813 * before a verdict can be assigned, even if the data spans 1814 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1815 * case would be a user calling **sendmsg**\ () repeatedly with 1816 * 1-byte long message segments. Obviously, this is bad for 1817 * performance, but it is still valid. If the eBPF program needs 1818 * *bytes* bytes to validate a header, this helper can be used to 1819 * prevent the eBPF program to be called again until *bytes* have 1820 * been accumulated. 1821 * Return 1822 * 0 1823 * 1824 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1825 * Description 1826 * For socket policies, pull in non-linear data from user space 1827 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1828 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1829 * respectively. 1830 * 1831 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1832 * *msg* it can only parse data that the (**data**, **data_end**) 1833 * pointers have already consumed. For **sendmsg**\ () hooks this 1834 * is likely the first scatterlist element. But for calls relying 1835 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1836 * be the range (**0**, **0**) because the data is shared with 1837 * user space and by default the objective is to avoid allowing 1838 * user space to modify data while (or after) eBPF verdict is 1839 * being decided. This helper can be used to pull in data and to 1840 * set the start and end pointer to given values. Data will be 1841 * copied if necessary (i.e. if data was not linear and if start 1842 * and end pointers do not point to the same chunk). 1843 * 1844 * A call to this helper is susceptible to change the underlaying 1845 * packet buffer. Therefore, at load time, all checks on pointers 1846 * previously done by the verifier are invalidated and must be 1847 * performed again, if the helper is used in combination with 1848 * direct packet access. 1849 * 1850 * All values for *flags* are reserved for future usage, and must 1851 * be left at zero. 1852 * Return 1853 * 0 on success, or a negative error in case of failure. 1854 * 1855 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1856 * Description 1857 * Bind the socket associated to *ctx* to the address pointed by 1858 * *addr*, of length *addr_len*. This allows for making outgoing 1859 * connection from the desired IP address, which can be useful for 1860 * example when all processes inside a cgroup should use one 1861 * single IP address on a host that has multiple IP configured. 1862 * 1863 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1864 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1865 * **AF_INET6**). Looking for a free port to bind to can be 1866 * expensive, therefore binding to port is not permitted by the 1867 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1868 * must be set to zero. 1869 * Return 1870 * 0 on success, or a negative error in case of failure. 1871 * 1872 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1873 * Description 1874 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1875 * only possible to shrink the packet as of this writing, 1876 * therefore *delta* must be a negative integer. 1877 * 1878 * A call to this helper is susceptible to change the underlaying 1879 * packet buffer. Therefore, at load time, all checks on pointers 1880 * previously done by the verifier are invalidated and must be 1881 * performed again, if the helper is used in combination with 1882 * direct packet access. 1883 * Return 1884 * 0 on success, or a negative error in case of failure. 1885 * 1886 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1887 * Description 1888 * Retrieve the XFRM state (IP transform framework, see also 1889 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1890 * 1891 * The retrieved value is stored in the **struct bpf_xfrm_state** 1892 * pointed by *xfrm_state* and of length *size*. 1893 * 1894 * All values for *flags* are reserved for future usage, and must 1895 * be left at zero. 1896 * 1897 * This helper is available only if the kernel was compiled with 1898 * **CONFIG_XFRM** configuration option. 1899 * Return 1900 * 0 on success, or a negative error in case of failure. 1901 * 1902 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1903 * Description 1904 * Return a user or a kernel stack in bpf program provided buffer. 1905 * To achieve this, the helper needs *ctx*, which is a pointer 1906 * to the context on which the tracing program is executed. 1907 * To store the stacktrace, the bpf program provides *buf* with 1908 * a nonnegative *size*. 1909 * 1910 * The last argument, *flags*, holds the number of stack frames to 1911 * skip (from 0 to 255), masked with 1912 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1913 * the following flags: 1914 * 1915 * **BPF_F_USER_STACK** 1916 * Collect a user space stack instead of a kernel stack. 1917 * **BPF_F_USER_BUILD_ID** 1918 * Collect buildid+offset instead of ips for user stack, 1919 * only valid if **BPF_F_USER_STACK** is also specified. 1920 * 1921 * **bpf_get_stack**\ () can collect up to 1922 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1923 * to sufficient large buffer size. Note that 1924 * this limit can be controlled with the **sysctl** program, and 1925 * that it should be manually increased in order to profile long 1926 * user stacks (such as stacks for Java programs). To do so, use: 1927 * 1928 * :: 1929 * 1930 * # sysctl kernel.perf_event_max_stack=<new value> 1931 * Return 1932 * A non-negative value equal to or less than *size* on success, 1933 * or a negative error in case of failure. 1934 * 1935 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1936 * Description 1937 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1938 * it provides an easy way to load *len* bytes from *offset* 1939 * from the packet associated to *skb*, into the buffer pointed 1940 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1941 * a fifth argument *start_header* exists in order to select a 1942 * base offset to start from. *start_header* can be one of: 1943 * 1944 * **BPF_HDR_START_MAC** 1945 * Base offset to load data from is *skb*'s mac header. 1946 * **BPF_HDR_START_NET** 1947 * Base offset to load data from is *skb*'s network header. 1948 * 1949 * In general, "direct packet access" is the preferred method to 1950 * access packet data, however, this helper is in particular useful 1951 * in socket filters where *skb*\ **->data** does not always point 1952 * to the start of the mac header and where "direct packet access" 1953 * is not available. 1954 * Return 1955 * 0 on success, or a negative error in case of failure. 1956 * 1957 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 1958 * Description 1959 * Do FIB lookup in kernel tables using parameters in *params*. 1960 * If lookup is successful and result shows packet is to be 1961 * forwarded, the neighbor tables are searched for the nexthop. 1962 * If successful (ie., FIB lookup shows forwarding and nexthop 1963 * is resolved), the nexthop address is returned in ipv4_dst 1964 * or ipv6_dst based on family, smac is set to mac address of 1965 * egress device, dmac is set to nexthop mac address, rt_metric 1966 * is set to metric from route (IPv4/IPv6 only), and ifindex 1967 * is set to the device index of the nexthop from the FIB lookup. 1968 * 1969 * *plen* argument is the size of the passed in struct. 1970 * *flags* argument can be a combination of one or more of the 1971 * following values: 1972 * 1973 * **BPF_FIB_LOOKUP_DIRECT** 1974 * Do a direct table lookup vs full lookup using FIB 1975 * rules. 1976 * **BPF_FIB_LOOKUP_OUTPUT** 1977 * Perform lookup from an egress perspective (default is 1978 * ingress). 1979 * 1980 * *ctx* is either **struct xdp_md** for XDP programs or 1981 * **struct sk_buff** tc cls_act programs. 1982 * Return 1983 * * < 0 if any input argument is invalid 1984 * * 0 on success (packet is forwarded, nexthop neighbor exists) 1985 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 1986 * packet is not forwarded or needs assist from full stack 1987 * 1988 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 1989 * Description 1990 * Add an entry to, or update a sockhash *map* referencing sockets. 1991 * The *skops* is used as a new value for the entry associated to 1992 * *key*. *flags* is one of: 1993 * 1994 * **BPF_NOEXIST** 1995 * The entry for *key* must not exist in the map. 1996 * **BPF_EXIST** 1997 * The entry for *key* must already exist in the map. 1998 * **BPF_ANY** 1999 * No condition on the existence of the entry for *key*. 2000 * 2001 * If the *map* has eBPF programs (parser and verdict), those will 2002 * be inherited by the socket being added. If the socket is 2003 * already attached to eBPF programs, this results in an error. 2004 * Return 2005 * 0 on success, or a negative error in case of failure. 2006 * 2007 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2008 * Description 2009 * This helper is used in programs implementing policies at the 2010 * socket level. If the message *msg* is allowed to pass (i.e. if 2011 * the verdict eBPF program returns **SK_PASS**), redirect it to 2012 * the socket referenced by *map* (of type 2013 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2014 * egress interfaces can be used for redirection. The 2015 * **BPF_F_INGRESS** value in *flags* is used to make the 2016 * distinction (ingress path is selected if the flag is present, 2017 * egress path otherwise). This is the only flag supported for now. 2018 * Return 2019 * **SK_PASS** on success, or **SK_DROP** on error. 2020 * 2021 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2022 * Description 2023 * This helper is used in programs implementing policies at the 2024 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2025 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2026 * to the socket referenced by *map* (of type 2027 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2028 * egress interfaces can be used for redirection. The 2029 * **BPF_F_INGRESS** value in *flags* is used to make the 2030 * distinction (ingress path is selected if the flag is present, 2031 * egress otherwise). This is the only flag supported for now. 2032 * Return 2033 * **SK_PASS** on success, or **SK_DROP** on error. 2034 * 2035 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2036 * Description 2037 * Encapsulate the packet associated to *skb* within a Layer 3 2038 * protocol header. This header is provided in the buffer at 2039 * address *hdr*, with *len* its size in bytes. *type* indicates 2040 * the protocol of the header and can be one of: 2041 * 2042 * **BPF_LWT_ENCAP_SEG6** 2043 * IPv6 encapsulation with Segment Routing Header 2044 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2045 * the IPv6 header is computed by the kernel. 2046 * **BPF_LWT_ENCAP_SEG6_INLINE** 2047 * Only works if *skb* contains an IPv6 packet. Insert a 2048 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2049 * the IPv6 header. 2050 * **BPF_LWT_ENCAP_IP** 2051 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2052 * must be IPv4 or IPv6, followed by zero or more 2053 * additional headers, up to LWT_BPF_MAX_HEADROOM total 2054 * bytes in all prepended headers. Please note that 2055 * if skb_is_gso(skb) is true, no more than two headers 2056 * can be prepended, and the inner header, if present, 2057 * should be either GRE or UDP/GUE. 2058 * 2059 * BPF_LWT_ENCAP_SEG6*** types can be called by bpf programs of 2060 * type BPF_PROG_TYPE_LWT_IN; BPF_LWT_ENCAP_IP type can be called 2061 * by bpf programs of types BPF_PROG_TYPE_LWT_IN and 2062 * BPF_PROG_TYPE_LWT_XMIT. 2063 * 2064 * A call to this helper is susceptible to change the underlaying 2065 * packet buffer. Therefore, at load time, all checks on pointers 2066 * previously done by the verifier are invalidated and must be 2067 * performed again, if the helper is used in combination with 2068 * direct packet access. 2069 * Return 2070 * 0 on success, or a negative error in case of failure. 2071 * 2072 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2073 * Description 2074 * Store *len* bytes from address *from* into the packet 2075 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2076 * inside the outermost IPv6 Segment Routing Header can be 2077 * modified through this helper. 2078 * 2079 * A call to this helper is susceptible to change the underlaying 2080 * packet buffer. Therefore, at load time, all checks on pointers 2081 * previously done by the verifier are invalidated and must be 2082 * performed again, if the helper is used in combination with 2083 * direct packet access. 2084 * Return 2085 * 0 on success, or a negative error in case of failure. 2086 * 2087 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2088 * Description 2089 * Adjust the size allocated to TLVs in the outermost IPv6 2090 * Segment Routing Header contained in the packet associated to 2091 * *skb*, at position *offset* by *delta* bytes. Only offsets 2092 * after the segments are accepted. *delta* can be as well 2093 * positive (growing) as negative (shrinking). 2094 * 2095 * A call to this helper is susceptible to change the underlaying 2096 * packet buffer. Therefore, at load time, all checks on pointers 2097 * previously done by the verifier are invalidated and must be 2098 * performed again, if the helper is used in combination with 2099 * direct packet access. 2100 * Return 2101 * 0 on success, or a negative error in case of failure. 2102 * 2103 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2104 * Description 2105 * Apply an IPv6 Segment Routing action of type *action* to the 2106 * packet associated to *skb*. Each action takes a parameter 2107 * contained at address *param*, and of length *param_len* bytes. 2108 * *action* can be one of: 2109 * 2110 * **SEG6_LOCAL_ACTION_END_X** 2111 * End.X action: Endpoint with Layer-3 cross-connect. 2112 * Type of *param*: **struct in6_addr**. 2113 * **SEG6_LOCAL_ACTION_END_T** 2114 * End.T action: Endpoint with specific IPv6 table lookup. 2115 * Type of *param*: **int**. 2116 * **SEG6_LOCAL_ACTION_END_B6** 2117 * End.B6 action: Endpoint bound to an SRv6 policy. 2118 * Type of param: **struct ipv6_sr_hdr**. 2119 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2120 * End.B6.Encap action: Endpoint bound to an SRv6 2121 * encapsulation policy. 2122 * Type of param: **struct ipv6_sr_hdr**. 2123 * 2124 * A call to this helper is susceptible to change the underlaying 2125 * packet buffer. Therefore, at load time, all checks on pointers 2126 * previously done by the verifier are invalidated and must be 2127 * performed again, if the helper is used in combination with 2128 * direct packet access. 2129 * Return 2130 * 0 on success, or a negative error in case of failure. 2131 * 2132 * int bpf_rc_repeat(void *ctx) 2133 * Description 2134 * This helper is used in programs implementing IR decoding, to 2135 * report a successfully decoded repeat key message. This delays 2136 * the generation of a key up event for previously generated 2137 * key down event. 2138 * 2139 * Some IR protocols like NEC have a special IR message for 2140 * repeating last button, for when a button is held down. 2141 * 2142 * The *ctx* should point to the lirc sample as passed into 2143 * the program. 2144 * 2145 * This helper is only available is the kernel was compiled with 2146 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2147 * "**y**". 2148 * Return 2149 * 0 2150 * 2151 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2152 * Description 2153 * This helper is used in programs implementing IR decoding, to 2154 * report a successfully decoded key press with *scancode*, 2155 * *toggle* value in the given *protocol*. The scancode will be 2156 * translated to a keycode using the rc keymap, and reported as 2157 * an input key down event. After a period a key up event is 2158 * generated. This period can be extended by calling either 2159 * **bpf_rc_keydown**\ () again with the same values, or calling 2160 * **bpf_rc_repeat**\ (). 2161 * 2162 * Some protocols include a toggle bit, in case the button was 2163 * released and pressed again between consecutive scancodes. 2164 * 2165 * The *ctx* should point to the lirc sample as passed into 2166 * the program. 2167 * 2168 * The *protocol* is the decoded protocol number (see 2169 * **enum rc_proto** for some predefined values). 2170 * 2171 * This helper is only available is the kernel was compiled with 2172 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2173 * "**y**". 2174 * Return 2175 * 0 2176 * 2177 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2178 * Description 2179 * Return the cgroup v2 id of the socket associated with the *skb*. 2180 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2181 * helper for cgroup v1 by providing a tag resp. identifier that 2182 * can be matched on or used for map lookups e.g. to implement 2183 * policy. The cgroup v2 id of a given path in the hierarchy is 2184 * exposed in user space through the f_handle API in order to get 2185 * to the same 64-bit id. 2186 * 2187 * This helper can be used on TC egress path, but not on ingress, 2188 * and is available only if the kernel was compiled with the 2189 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2190 * Return 2191 * The id is returned or 0 in case the id could not be retrieved. 2192 * 2193 * u64 bpf_get_current_cgroup_id(void) 2194 * Return 2195 * A 64-bit integer containing the current cgroup id based 2196 * on the cgroup within which the current task is running. 2197 * 2198 * void *bpf_get_local_storage(void *map, u64 flags) 2199 * Description 2200 * Get the pointer to the local storage area. 2201 * The type and the size of the local storage is defined 2202 * by the *map* argument. 2203 * The *flags* meaning is specific for each map type, 2204 * and has to be 0 for cgroup local storage. 2205 * 2206 * Depending on the BPF program type, a local storage area 2207 * can be shared between multiple instances of the BPF program, 2208 * running simultaneously. 2209 * 2210 * A user should care about the synchronization by himself. 2211 * For example, by using the **BPF_STX_XADD** instruction to alter 2212 * the shared data. 2213 * Return 2214 * A pointer to the local storage area. 2215 * 2216 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2217 * Description 2218 * Select a **SO_REUSEPORT** socket from a 2219 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2220 * It checks the selected socket is matching the incoming 2221 * request in the socket buffer. 2222 * Return 2223 * 0 on success, or a negative error in case of failure. 2224 * 2225 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2226 * Description 2227 * Return id of cgroup v2 that is ancestor of cgroup associated 2228 * with the *skb* at the *ancestor_level*. The root cgroup is at 2229 * *ancestor_level* zero and each step down the hierarchy 2230 * increments the level. If *ancestor_level* == level of cgroup 2231 * associated with *skb*, then return value will be same as that 2232 * of **bpf_skb_cgroup_id**\ (). 2233 * 2234 * The helper is useful to implement policies based on cgroups 2235 * that are upper in hierarchy than immediate cgroup associated 2236 * with *skb*. 2237 * 2238 * The format of returned id and helper limitations are same as in 2239 * **bpf_skb_cgroup_id**\ (). 2240 * Return 2241 * The id is returned or 0 in case the id could not be retrieved. 2242 * 2243 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2244 * Description 2245 * Look for TCP socket matching *tuple*, optionally in a child 2246 * network namespace *netns*. The return value must be checked, 2247 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2248 * 2249 * The *ctx* should point to the context of the program, such as 2250 * the skb or socket (depending on the hook in use). This is used 2251 * to determine the base network namespace for the lookup. 2252 * 2253 * *tuple_size* must be one of: 2254 * 2255 * **sizeof**\ (*tuple*\ **->ipv4**) 2256 * Look for an IPv4 socket. 2257 * **sizeof**\ (*tuple*\ **->ipv6**) 2258 * Look for an IPv6 socket. 2259 * 2260 * If the *netns* is a negative signed 32-bit integer, then the 2261 * socket lookup table in the netns associated with the *ctx* will 2262 * will be used. For the TC hooks, this is the netns of the device 2263 * in the skb. For socket hooks, this is the netns of the socket. 2264 * If *netns* is any other signed 32-bit value greater than or 2265 * equal to zero then it specifies the ID of the netns relative to 2266 * the netns associated with the *ctx*. *netns* values beyond the 2267 * range of 32-bit integers are reserved for future use. 2268 * 2269 * All values for *flags* are reserved for future usage, and must 2270 * be left at zero. 2271 * 2272 * This helper is available only if the kernel was compiled with 2273 * **CONFIG_NET** configuration option. 2274 * Return 2275 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2276 * For sockets with reuseport option, the **struct bpf_sock** 2277 * result is from **reuse->socks**\ [] using the hash of the tuple. 2278 * 2279 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2280 * Description 2281 * Look for UDP socket matching *tuple*, optionally in a child 2282 * network namespace *netns*. The return value must be checked, 2283 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2284 * 2285 * The *ctx* should point to the context of the program, such as 2286 * the skb or socket (depending on the hook in use). This is used 2287 * to determine the base network namespace for the lookup. 2288 * 2289 * *tuple_size* must be one of: 2290 * 2291 * **sizeof**\ (*tuple*\ **->ipv4**) 2292 * Look for an IPv4 socket. 2293 * **sizeof**\ (*tuple*\ **->ipv6**) 2294 * Look for an IPv6 socket. 2295 * 2296 * If the *netns* is a negative signed 32-bit integer, then the 2297 * socket lookup table in the netns associated with the *ctx* will 2298 * will be used. For the TC hooks, this is the netns of the device 2299 * in the skb. For socket hooks, this is the netns of the socket. 2300 * If *netns* is any other signed 32-bit value greater than or 2301 * equal to zero then it specifies the ID of the netns relative to 2302 * the netns associated with the *ctx*. *netns* values beyond the 2303 * range of 32-bit integers are reserved for future use. 2304 * 2305 * All values for *flags* are reserved for future usage, and must 2306 * be left at zero. 2307 * 2308 * This helper is available only if the kernel was compiled with 2309 * **CONFIG_NET** configuration option. 2310 * Return 2311 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2312 * For sockets with reuseport option, the **struct bpf_sock** 2313 * result is from **reuse->socks**\ [] using the hash of the tuple. 2314 * 2315 * int bpf_sk_release(struct bpf_sock *sock) 2316 * Description 2317 * Release the reference held by *sock*. *sock* must be a 2318 * non-**NULL** pointer that was returned from 2319 * **bpf_sk_lookup_xxx**\ (). 2320 * Return 2321 * 0 on success, or a negative error in case of failure. 2322 * 2323 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2324 * Description 2325 * Push an element *value* in *map*. *flags* is one of: 2326 * 2327 * **BPF_EXIST** 2328 * If the queue/stack is full, the oldest element is 2329 * removed to make room for this. 2330 * Return 2331 * 0 on success, or a negative error in case of failure. 2332 * 2333 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2334 * Description 2335 * Pop an element from *map*. 2336 * Return 2337 * 0 on success, or a negative error in case of failure. 2338 * 2339 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2340 * Description 2341 * Get an element from *map* without removing it. 2342 * Return 2343 * 0 on success, or a negative error in case of failure. 2344 * 2345 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 2346 * Description 2347 * For socket policies, insert *len* bytes into *msg* at offset 2348 * *start*. 2349 * 2350 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2351 * *msg* it may want to insert metadata or options into the *msg*. 2352 * This can later be read and used by any of the lower layer BPF 2353 * hooks. 2354 * 2355 * This helper may fail if under memory pressure (a malloc 2356 * fails) in these cases BPF programs will get an appropriate 2357 * error and BPF programs will need to handle them. 2358 * Return 2359 * 0 on success, or a negative error in case of failure. 2360 * 2361 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags) 2362 * Description 2363 * Will remove *pop* bytes from a *msg* starting at byte *start*. 2364 * This may result in **ENOMEM** errors under certain situations if 2365 * an allocation and copy are required due to a full ring buffer. 2366 * However, the helper will try to avoid doing the allocation 2367 * if possible. Other errors can occur if input parameters are 2368 * invalid either due to *start* byte not being valid part of *msg* 2369 * payload and/or *pop* value being to large. 2370 * Return 2371 * 0 on success, or a negative error in case of failure. 2372 * 2373 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2374 * Description 2375 * This helper is used in programs implementing IR decoding, to 2376 * report a successfully decoded pointer movement. 2377 * 2378 * The *ctx* should point to the lirc sample as passed into 2379 * the program. 2380 * 2381 * This helper is only available is the kernel was compiled with 2382 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2383 * "**y**". 2384 * Return 2385 * 0 2386 * 2387 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2388 * Description 2389 * Acquire a spinlock represented by the pointer *lock*, which is 2390 * stored as part of a value of a map. Taking the lock allows to 2391 * safely update the rest of the fields in that value. The 2392 * spinlock can (and must) later be released with a call to 2393 * **bpf_spin_unlock**\ (\ *lock*\ ). 2394 * 2395 * Spinlocks in BPF programs come with a number of restrictions 2396 * and constraints: 2397 * 2398 * * **bpf_spin_lock** objects are only allowed inside maps of 2399 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2400 * list could be extended in the future). 2401 * * BTF description of the map is mandatory. 2402 * * The BPF program can take ONE lock at a time, since taking two 2403 * or more could cause dead locks. 2404 * * Only one **struct bpf_spin_lock** is allowed per map element. 2405 * * When the lock is taken, calls (either BPF to BPF or helpers) 2406 * are not allowed. 2407 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2408 * allowed inside a spinlock-ed region. 2409 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2410 * the lock, on all execution paths, before it returns. 2411 * * The BPF program can access **struct bpf_spin_lock** only via 2412 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2413 * helpers. Loading or storing data into the **struct 2414 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2415 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2416 * of the map value must be a struct and have **struct 2417 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2418 * Nested lock inside another struct is not allowed. 2419 * * The **struct bpf_spin_lock** *lock* field in a map value must 2420 * be aligned on a multiple of 4 bytes in that value. 2421 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2422 * the **bpf_spin_lock** field to user space. 2423 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2424 * a BPF program, do not update the **bpf_spin_lock** field. 2425 * * **bpf_spin_lock** cannot be on the stack or inside a 2426 * networking packet (it can only be inside of a map values). 2427 * * **bpf_spin_lock** is available to root only. 2428 * * Tracing programs and socket filter programs cannot use 2429 * **bpf_spin_lock**\ () due to insufficient preemption checks 2430 * (but this may change in the future). 2431 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2432 * Return 2433 * 0 2434 * 2435 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2436 * Description 2437 * Release the *lock* previously locked by a call to 2438 * **bpf_spin_lock**\ (\ *lock*\ ). 2439 * Return 2440 * 0 2441 * 2442 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2443 * Description 2444 * This helper gets a **struct bpf_sock** pointer such 2445 * that all the fields in this **bpf_sock** can be accessed. 2446 * Return 2447 * A **struct bpf_sock** pointer on success, or **NULL** in 2448 * case of failure. 2449 * 2450 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2451 * Description 2452 * This helper gets a **struct bpf_tcp_sock** pointer from a 2453 * **struct bpf_sock** pointer. 2454 * Return 2455 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2456 * case of failure. 2457 * 2458 * int bpf_skb_ecn_set_ce(struct sk_buf *skb) 2459 * Description 2460 * Set ECN (Explicit Congestion Notification) field of IP header 2461 * to **CE** (Congestion Encountered) if current value is **ECT** 2462 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2463 * and IPv4. 2464 * Return 2465 * 1 if the **CE** flag is set (either by the current helper call 2466 * or because it was already present), 0 if it is not set. 2467 * 2468 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2469 * Description 2470 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2471 * **bpf_sk_release**\ () is unnecessary and not allowed. 2472 * Return 2473 * A **struct bpf_sock** pointer on success, or **NULL** in 2474 * case of failure. 2475 * 2476 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2477 * Description 2478 * Look for TCP socket matching *tuple*, optionally in a child 2479 * network namespace *netns*. The return value must be checked, 2480 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2481 * 2482 * This function is identical to bpf_sk_lookup_tcp, except that it 2483 * also returns timewait or request sockets. Use bpf_sk_fullsock 2484 * or bpf_tcp_socket to access the full structure. 2485 * 2486 * This helper is available only if the kernel was compiled with 2487 * **CONFIG_NET** configuration option. 2488 * Return 2489 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2490 * For sockets with reuseport option, the **struct bpf_sock** 2491 * result is from **reuse->socks**\ [] using the hash of the tuple. 2492 * 2493 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2494 * Description 2495 * Check whether iph and th contain a valid SYN cookie ACK for 2496 * the listening socket in sk. 2497 * 2498 * iph points to the start of the IPv4 or IPv6 header, while 2499 * iph_len contains sizeof(struct iphdr) or sizeof(struct ip6hdr). 2500 * 2501 * th points to the start of the TCP header, while th_len contains 2502 * sizeof(struct tcphdr). 2503 * 2504 * Return 2505 * 0 if iph and th are a valid SYN cookie ACK, or a negative error 2506 * otherwise. 2507 */ 2508 #define __BPF_FUNC_MAPPER(FN) \ 2509 FN(unspec), \ 2510 FN(map_lookup_elem), \ 2511 FN(map_update_elem), \ 2512 FN(map_delete_elem), \ 2513 FN(probe_read), \ 2514 FN(ktime_get_ns), \ 2515 FN(trace_printk), \ 2516 FN(get_prandom_u32), \ 2517 FN(get_smp_processor_id), \ 2518 FN(skb_store_bytes), \ 2519 FN(l3_csum_replace), \ 2520 FN(l4_csum_replace), \ 2521 FN(tail_call), \ 2522 FN(clone_redirect), \ 2523 FN(get_current_pid_tgid), \ 2524 FN(get_current_uid_gid), \ 2525 FN(get_current_comm), \ 2526 FN(get_cgroup_classid), \ 2527 FN(skb_vlan_push), \ 2528 FN(skb_vlan_pop), \ 2529 FN(skb_get_tunnel_key), \ 2530 FN(skb_set_tunnel_key), \ 2531 FN(perf_event_read), \ 2532 FN(redirect), \ 2533 FN(get_route_realm), \ 2534 FN(perf_event_output), \ 2535 FN(skb_load_bytes), \ 2536 FN(get_stackid), \ 2537 FN(csum_diff), \ 2538 FN(skb_get_tunnel_opt), \ 2539 FN(skb_set_tunnel_opt), \ 2540 FN(skb_change_proto), \ 2541 FN(skb_change_type), \ 2542 FN(skb_under_cgroup), \ 2543 FN(get_hash_recalc), \ 2544 FN(get_current_task), \ 2545 FN(probe_write_user), \ 2546 FN(current_task_under_cgroup), \ 2547 FN(skb_change_tail), \ 2548 FN(skb_pull_data), \ 2549 FN(csum_update), \ 2550 FN(set_hash_invalid), \ 2551 FN(get_numa_node_id), \ 2552 FN(skb_change_head), \ 2553 FN(xdp_adjust_head), \ 2554 FN(probe_read_str), \ 2555 FN(get_socket_cookie), \ 2556 FN(get_socket_uid), \ 2557 FN(set_hash), \ 2558 FN(setsockopt), \ 2559 FN(skb_adjust_room), \ 2560 FN(redirect_map), \ 2561 FN(sk_redirect_map), \ 2562 FN(sock_map_update), \ 2563 FN(xdp_adjust_meta), \ 2564 FN(perf_event_read_value), \ 2565 FN(perf_prog_read_value), \ 2566 FN(getsockopt), \ 2567 FN(override_return), \ 2568 FN(sock_ops_cb_flags_set), \ 2569 FN(msg_redirect_map), \ 2570 FN(msg_apply_bytes), \ 2571 FN(msg_cork_bytes), \ 2572 FN(msg_pull_data), \ 2573 FN(bind), \ 2574 FN(xdp_adjust_tail), \ 2575 FN(skb_get_xfrm_state), \ 2576 FN(get_stack), \ 2577 FN(skb_load_bytes_relative), \ 2578 FN(fib_lookup), \ 2579 FN(sock_hash_update), \ 2580 FN(msg_redirect_hash), \ 2581 FN(sk_redirect_hash), \ 2582 FN(lwt_push_encap), \ 2583 FN(lwt_seg6_store_bytes), \ 2584 FN(lwt_seg6_adjust_srh), \ 2585 FN(lwt_seg6_action), \ 2586 FN(rc_repeat), \ 2587 FN(rc_keydown), \ 2588 FN(skb_cgroup_id), \ 2589 FN(get_current_cgroup_id), \ 2590 FN(get_local_storage), \ 2591 FN(sk_select_reuseport), \ 2592 FN(skb_ancestor_cgroup_id), \ 2593 FN(sk_lookup_tcp), \ 2594 FN(sk_lookup_udp), \ 2595 FN(sk_release), \ 2596 FN(map_push_elem), \ 2597 FN(map_pop_elem), \ 2598 FN(map_peek_elem), \ 2599 FN(msg_push_data), \ 2600 FN(msg_pop_data), \ 2601 FN(rc_pointer_rel), \ 2602 FN(spin_lock), \ 2603 FN(spin_unlock), \ 2604 FN(sk_fullsock), \ 2605 FN(tcp_sock), \ 2606 FN(skb_ecn_set_ce), \ 2607 FN(get_listener_sock), \ 2608 FN(skc_lookup_tcp), \ 2609 FN(tcp_check_syncookie), 2610 2611 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2612 * function eBPF program intends to call 2613 */ 2614 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2615 enum bpf_func_id { 2616 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2617 __BPF_FUNC_MAX_ID, 2618 }; 2619 #undef __BPF_ENUM_FN 2620 2621 /* All flags used by eBPF helper functions, placed here. */ 2622 2623 /* BPF_FUNC_skb_store_bytes flags. */ 2624 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2625 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2626 2627 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2628 * First 4 bits are for passing the header field size. 2629 */ 2630 #define BPF_F_HDR_FIELD_MASK 0xfULL 2631 2632 /* BPF_FUNC_l4_csum_replace flags. */ 2633 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2634 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2635 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2636 2637 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2638 #define BPF_F_INGRESS (1ULL << 0) 2639 2640 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2641 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2642 2643 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2644 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2645 #define BPF_F_USER_STACK (1ULL << 8) 2646 /* flags used by BPF_FUNC_get_stackid only. */ 2647 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2648 #define BPF_F_REUSE_STACKID (1ULL << 10) 2649 /* flags used by BPF_FUNC_get_stack only. */ 2650 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2651 2652 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2653 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2654 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2655 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2656 2657 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2658 * BPF_FUNC_perf_event_read_value flags. 2659 */ 2660 #define BPF_F_INDEX_MASK 0xffffffffULL 2661 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2662 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2663 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2664 2665 /* Current network namespace */ 2666 #define BPF_F_CURRENT_NETNS (-1L) 2667 2668 /* BPF_FUNC_skb_adjust_room flags. */ 2669 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 2670 2671 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 2672 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 2673 2674 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 2675 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 2676 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 2677 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 2678 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 2679 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 2680 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 2681 2682 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 2683 enum bpf_adj_room_mode { 2684 BPF_ADJ_ROOM_NET, 2685 BPF_ADJ_ROOM_MAC, 2686 }; 2687 2688 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 2689 enum bpf_hdr_start_off { 2690 BPF_HDR_START_MAC, 2691 BPF_HDR_START_NET, 2692 }; 2693 2694 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2695 enum bpf_lwt_encap_mode { 2696 BPF_LWT_ENCAP_SEG6, 2697 BPF_LWT_ENCAP_SEG6_INLINE, 2698 BPF_LWT_ENCAP_IP, 2699 }; 2700 2701 #define __bpf_md_ptr(type, name) \ 2702 union { \ 2703 type name; \ 2704 __u64 :64; \ 2705 } __attribute__((aligned(8))) 2706 2707 /* user accessible mirror of in-kernel sk_buff. 2708 * new fields can only be added to the end of this structure 2709 */ 2710 struct __sk_buff { 2711 __u32 len; 2712 __u32 pkt_type; 2713 __u32 mark; 2714 __u32 queue_mapping; 2715 __u32 protocol; 2716 __u32 vlan_present; 2717 __u32 vlan_tci; 2718 __u32 vlan_proto; 2719 __u32 priority; 2720 __u32 ingress_ifindex; 2721 __u32 ifindex; 2722 __u32 tc_index; 2723 __u32 cb[5]; 2724 __u32 hash; 2725 __u32 tc_classid; 2726 __u32 data; 2727 __u32 data_end; 2728 __u32 napi_id; 2729 2730 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 2731 __u32 family; 2732 __u32 remote_ip4; /* Stored in network byte order */ 2733 __u32 local_ip4; /* Stored in network byte order */ 2734 __u32 remote_ip6[4]; /* Stored in network byte order */ 2735 __u32 local_ip6[4]; /* Stored in network byte order */ 2736 __u32 remote_port; /* Stored in network byte order */ 2737 __u32 local_port; /* stored in host byte order */ 2738 /* ... here. */ 2739 2740 __u32 data_meta; 2741 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 2742 __u64 tstamp; 2743 __u32 wire_len; 2744 __u32 gso_segs; 2745 __bpf_md_ptr(struct bpf_sock *, sk); 2746 }; 2747 2748 struct bpf_tunnel_key { 2749 __u32 tunnel_id; 2750 union { 2751 __u32 remote_ipv4; 2752 __u32 remote_ipv6[4]; 2753 }; 2754 __u8 tunnel_tos; 2755 __u8 tunnel_ttl; 2756 __u16 tunnel_ext; /* Padding, future use. */ 2757 __u32 tunnel_label; 2758 }; 2759 2760 /* user accessible mirror of in-kernel xfrm_state. 2761 * new fields can only be added to the end of this structure 2762 */ 2763 struct bpf_xfrm_state { 2764 __u32 reqid; 2765 __u32 spi; /* Stored in network byte order */ 2766 __u16 family; 2767 __u16 ext; /* Padding, future use. */ 2768 union { 2769 __u32 remote_ipv4; /* Stored in network byte order */ 2770 __u32 remote_ipv6[4]; /* Stored in network byte order */ 2771 }; 2772 }; 2773 2774 /* Generic BPF return codes which all BPF program types may support. 2775 * The values are binary compatible with their TC_ACT_* counter-part to 2776 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 2777 * programs. 2778 * 2779 * XDP is handled seprately, see XDP_*. 2780 */ 2781 enum bpf_ret_code { 2782 BPF_OK = 0, 2783 /* 1 reserved */ 2784 BPF_DROP = 2, 2785 /* 3-6 reserved */ 2786 BPF_REDIRECT = 7, 2787 /* >127 are reserved for prog type specific return codes. 2788 * 2789 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 2790 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 2791 * changed and should be routed based on its new L3 header. 2792 * (This is an L3 redirect, as opposed to L2 redirect 2793 * represented by BPF_REDIRECT above). 2794 */ 2795 BPF_LWT_REROUTE = 128, 2796 }; 2797 2798 struct bpf_sock { 2799 __u32 bound_dev_if; 2800 __u32 family; 2801 __u32 type; 2802 __u32 protocol; 2803 __u32 mark; 2804 __u32 priority; 2805 /* IP address also allows 1 and 2 bytes access */ 2806 __u32 src_ip4; 2807 __u32 src_ip6[4]; 2808 __u32 src_port; /* host byte order */ 2809 __u32 dst_port; /* network byte order */ 2810 __u32 dst_ip4; 2811 __u32 dst_ip6[4]; 2812 __u32 state; 2813 }; 2814 2815 struct bpf_tcp_sock { 2816 __u32 snd_cwnd; /* Sending congestion window */ 2817 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 2818 __u32 rtt_min; 2819 __u32 snd_ssthresh; /* Slow start size threshold */ 2820 __u32 rcv_nxt; /* What we want to receive next */ 2821 __u32 snd_nxt; /* Next sequence we send */ 2822 __u32 snd_una; /* First byte we want an ack for */ 2823 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 2824 __u32 ecn_flags; /* ECN status bits. */ 2825 __u32 rate_delivered; /* saved rate sample: packets delivered */ 2826 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 2827 __u32 packets_out; /* Packets which are "in flight" */ 2828 __u32 retrans_out; /* Retransmitted packets out */ 2829 __u32 total_retrans; /* Total retransmits for entire connection */ 2830 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 2831 * total number of segments in. 2832 */ 2833 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 2834 * total number of data segments in. 2835 */ 2836 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 2837 * The total number of segments sent. 2838 */ 2839 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 2840 * total number of data segments sent. 2841 */ 2842 __u32 lost_out; /* Lost packets */ 2843 __u32 sacked_out; /* SACK'd packets */ 2844 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 2845 * sum(delta(rcv_nxt)), or how many bytes 2846 * were acked. 2847 */ 2848 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 2849 * sum(delta(snd_una)), or how many bytes 2850 * were acked. 2851 */ 2852 }; 2853 2854 struct bpf_sock_tuple { 2855 union { 2856 struct { 2857 __be32 saddr; 2858 __be32 daddr; 2859 __be16 sport; 2860 __be16 dport; 2861 } ipv4; 2862 struct { 2863 __be32 saddr[4]; 2864 __be32 daddr[4]; 2865 __be16 sport; 2866 __be16 dport; 2867 } ipv6; 2868 }; 2869 }; 2870 2871 #define XDP_PACKET_HEADROOM 256 2872 2873 /* User return codes for XDP prog type. 2874 * A valid XDP program must return one of these defined values. All other 2875 * return codes are reserved for future use. Unknown return codes will 2876 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 2877 */ 2878 enum xdp_action { 2879 XDP_ABORTED = 0, 2880 XDP_DROP, 2881 XDP_PASS, 2882 XDP_TX, 2883 XDP_REDIRECT, 2884 }; 2885 2886 /* user accessible metadata for XDP packet hook 2887 * new fields must be added to the end of this structure 2888 */ 2889 struct xdp_md { 2890 __u32 data; 2891 __u32 data_end; 2892 __u32 data_meta; 2893 /* Below access go through struct xdp_rxq_info */ 2894 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 2895 __u32 rx_queue_index; /* rxq->queue_index */ 2896 }; 2897 2898 enum sk_action { 2899 SK_DROP = 0, 2900 SK_PASS, 2901 }; 2902 2903 /* user accessible metadata for SK_MSG packet hook, new fields must 2904 * be added to the end of this structure 2905 */ 2906 struct sk_msg_md { 2907 __bpf_md_ptr(void *, data); 2908 __bpf_md_ptr(void *, data_end); 2909 2910 __u32 family; 2911 __u32 remote_ip4; /* Stored in network byte order */ 2912 __u32 local_ip4; /* Stored in network byte order */ 2913 __u32 remote_ip6[4]; /* Stored in network byte order */ 2914 __u32 local_ip6[4]; /* Stored in network byte order */ 2915 __u32 remote_port; /* Stored in network byte order */ 2916 __u32 local_port; /* stored in host byte order */ 2917 __u32 size; /* Total size of sk_msg */ 2918 }; 2919 2920 struct sk_reuseport_md { 2921 /* 2922 * Start of directly accessible data. It begins from 2923 * the tcp/udp header. 2924 */ 2925 __bpf_md_ptr(void *, data); 2926 /* End of directly accessible data */ 2927 __bpf_md_ptr(void *, data_end); 2928 /* 2929 * Total length of packet (starting from the tcp/udp header). 2930 * Note that the directly accessible bytes (data_end - data) 2931 * could be less than this "len". Those bytes could be 2932 * indirectly read by a helper "bpf_skb_load_bytes()". 2933 */ 2934 __u32 len; 2935 /* 2936 * Eth protocol in the mac header (network byte order). e.g. 2937 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 2938 */ 2939 __u32 eth_protocol; 2940 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 2941 __u32 bind_inany; /* Is sock bound to an INANY address? */ 2942 __u32 hash; /* A hash of the packet 4 tuples */ 2943 }; 2944 2945 #define BPF_TAG_SIZE 8 2946 2947 struct bpf_prog_info { 2948 __u32 type; 2949 __u32 id; 2950 __u8 tag[BPF_TAG_SIZE]; 2951 __u32 jited_prog_len; 2952 __u32 xlated_prog_len; 2953 __aligned_u64 jited_prog_insns; 2954 __aligned_u64 xlated_prog_insns; 2955 __u64 load_time; /* ns since boottime */ 2956 __u32 created_by_uid; 2957 __u32 nr_map_ids; 2958 __aligned_u64 map_ids; 2959 char name[BPF_OBJ_NAME_LEN]; 2960 __u32 ifindex; 2961 __u32 gpl_compatible:1; 2962 __u64 netns_dev; 2963 __u64 netns_ino; 2964 __u32 nr_jited_ksyms; 2965 __u32 nr_jited_func_lens; 2966 __aligned_u64 jited_ksyms; 2967 __aligned_u64 jited_func_lens; 2968 __u32 btf_id; 2969 __u32 func_info_rec_size; 2970 __aligned_u64 func_info; 2971 __u32 nr_func_info; 2972 __u32 nr_line_info; 2973 __aligned_u64 line_info; 2974 __aligned_u64 jited_line_info; 2975 __u32 nr_jited_line_info; 2976 __u32 line_info_rec_size; 2977 __u32 jited_line_info_rec_size; 2978 __u32 nr_prog_tags; 2979 __aligned_u64 prog_tags; 2980 __u64 run_time_ns; 2981 __u64 run_cnt; 2982 } __attribute__((aligned(8))); 2983 2984 struct bpf_map_info { 2985 __u32 type; 2986 __u32 id; 2987 __u32 key_size; 2988 __u32 value_size; 2989 __u32 max_entries; 2990 __u32 map_flags; 2991 char name[BPF_OBJ_NAME_LEN]; 2992 __u32 ifindex; 2993 __u32 :32; 2994 __u64 netns_dev; 2995 __u64 netns_ino; 2996 __u32 btf_id; 2997 __u32 btf_key_type_id; 2998 __u32 btf_value_type_id; 2999 } __attribute__((aligned(8))); 3000 3001 struct bpf_btf_info { 3002 __aligned_u64 btf; 3003 __u32 btf_size; 3004 __u32 id; 3005 } __attribute__((aligned(8))); 3006 3007 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3008 * by user and intended to be used by socket (e.g. to bind to, depends on 3009 * attach attach type). 3010 */ 3011 struct bpf_sock_addr { 3012 __u32 user_family; /* Allows 4-byte read, but no write. */ 3013 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3014 * Stored in network byte order. 3015 */ 3016 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 3017 * Stored in network byte order. 3018 */ 3019 __u32 user_port; /* Allows 4-byte read and write. 3020 * Stored in network byte order 3021 */ 3022 __u32 family; /* Allows 4-byte read, but no write */ 3023 __u32 type; /* Allows 4-byte read, but no write */ 3024 __u32 protocol; /* Allows 4-byte read, but no write */ 3025 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write. 3026 * Stored in network byte order. 3027 */ 3028 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 3029 * Stored in network byte order. 3030 */ 3031 }; 3032 3033 /* User bpf_sock_ops struct to access socket values and specify request ops 3034 * and their replies. 3035 * Some of this fields are in network (bigendian) byte order and may need 3036 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3037 * New fields can only be added at the end of this structure 3038 */ 3039 struct bpf_sock_ops { 3040 __u32 op; 3041 union { 3042 __u32 args[4]; /* Optionally passed to bpf program */ 3043 __u32 reply; /* Returned by bpf program */ 3044 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3045 }; 3046 __u32 family; 3047 __u32 remote_ip4; /* Stored in network byte order */ 3048 __u32 local_ip4; /* Stored in network byte order */ 3049 __u32 remote_ip6[4]; /* Stored in network byte order */ 3050 __u32 local_ip6[4]; /* Stored in network byte order */ 3051 __u32 remote_port; /* Stored in network byte order */ 3052 __u32 local_port; /* stored in host byte order */ 3053 __u32 is_fullsock; /* Some TCP fields are only valid if 3054 * there is a full socket. If not, the 3055 * fields read as zero. 3056 */ 3057 __u32 snd_cwnd; 3058 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3059 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3060 __u32 state; 3061 __u32 rtt_min; 3062 __u32 snd_ssthresh; 3063 __u32 rcv_nxt; 3064 __u32 snd_nxt; 3065 __u32 snd_una; 3066 __u32 mss_cache; 3067 __u32 ecn_flags; 3068 __u32 rate_delivered; 3069 __u32 rate_interval_us; 3070 __u32 packets_out; 3071 __u32 retrans_out; 3072 __u32 total_retrans; 3073 __u32 segs_in; 3074 __u32 data_segs_in; 3075 __u32 segs_out; 3076 __u32 data_segs_out; 3077 __u32 lost_out; 3078 __u32 sacked_out; 3079 __u32 sk_txhash; 3080 __u64 bytes_received; 3081 __u64 bytes_acked; 3082 }; 3083 3084 /* Definitions for bpf_sock_ops_cb_flags */ 3085 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3086 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3087 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3088 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 3089 * supported cb flags 3090 */ 3091 3092 /* List of known BPF sock_ops operators. 3093 * New entries can only be added at the end 3094 */ 3095 enum { 3096 BPF_SOCK_OPS_VOID, 3097 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3098 * -1 if default value should be used 3099 */ 3100 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3101 * window (in packets) or -1 if default 3102 * value should be used 3103 */ 3104 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3105 * active connection is initialized 3106 */ 3107 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3108 * active connection is 3109 * established 3110 */ 3111 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3112 * passive connection is 3113 * established 3114 */ 3115 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3116 * needs ECN 3117 */ 3118 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3119 * based on the path and may be 3120 * dependent on the congestion control 3121 * algorithm. In general it indicates 3122 * a congestion threshold. RTTs above 3123 * this indicate congestion 3124 */ 3125 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3126 * Arg1: value of icsk_retransmits 3127 * Arg2: value of icsk_rto 3128 * Arg3: whether RTO has expired 3129 */ 3130 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3131 * Arg1: sequence number of 1st byte 3132 * Arg2: # segments 3133 * Arg3: return value of 3134 * tcp_transmit_skb (0 => success) 3135 */ 3136 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3137 * Arg1: old_state 3138 * Arg2: new_state 3139 */ 3140 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3141 * socket transition to LISTEN state. 3142 */ 3143 }; 3144 3145 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3146 * changes between the TCP and BPF versions. Ideally this should never happen. 3147 * If it does, we need to add code to convert them before calling 3148 * the BPF sock_ops function. 3149 */ 3150 enum { 3151 BPF_TCP_ESTABLISHED = 1, 3152 BPF_TCP_SYN_SENT, 3153 BPF_TCP_SYN_RECV, 3154 BPF_TCP_FIN_WAIT1, 3155 BPF_TCP_FIN_WAIT2, 3156 BPF_TCP_TIME_WAIT, 3157 BPF_TCP_CLOSE, 3158 BPF_TCP_CLOSE_WAIT, 3159 BPF_TCP_LAST_ACK, 3160 BPF_TCP_LISTEN, 3161 BPF_TCP_CLOSING, /* Now a valid state */ 3162 BPF_TCP_NEW_SYN_RECV, 3163 3164 BPF_TCP_MAX_STATES /* Leave at the end! */ 3165 }; 3166 3167 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3168 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3169 3170 struct bpf_perf_event_value { 3171 __u64 counter; 3172 __u64 enabled; 3173 __u64 running; 3174 }; 3175 3176 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3177 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3178 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3179 3180 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3181 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3182 3183 struct bpf_cgroup_dev_ctx { 3184 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3185 __u32 access_type; 3186 __u32 major; 3187 __u32 minor; 3188 }; 3189 3190 struct bpf_raw_tracepoint_args { 3191 __u64 args[0]; 3192 }; 3193 3194 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3195 * OUTPUT: Do lookup from egress perspective; default is ingress 3196 */ 3197 #define BPF_FIB_LOOKUP_DIRECT BIT(0) 3198 #define BPF_FIB_LOOKUP_OUTPUT BIT(1) 3199 3200 enum { 3201 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3202 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3203 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3204 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3205 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3206 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3207 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3208 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3209 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3210 }; 3211 3212 struct bpf_fib_lookup { 3213 /* input: network family for lookup (AF_INET, AF_INET6) 3214 * output: network family of egress nexthop 3215 */ 3216 __u8 family; 3217 3218 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3219 __u8 l4_protocol; 3220 __be16 sport; 3221 __be16 dport; 3222 3223 /* total length of packet from network header - used for MTU check */ 3224 __u16 tot_len; 3225 3226 /* input: L3 device index for lookup 3227 * output: device index from FIB lookup 3228 */ 3229 __u32 ifindex; 3230 3231 union { 3232 /* inputs to lookup */ 3233 __u8 tos; /* AF_INET */ 3234 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3235 3236 /* output: metric of fib result (IPv4/IPv6 only) */ 3237 __u32 rt_metric; 3238 }; 3239 3240 union { 3241 __be32 ipv4_src; 3242 __u32 ipv6_src[4]; /* in6_addr; network order */ 3243 }; 3244 3245 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3246 * network header. output: bpf_fib_lookup sets to gateway address 3247 * if FIB lookup returns gateway route 3248 */ 3249 union { 3250 __be32 ipv4_dst; 3251 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3252 }; 3253 3254 /* output */ 3255 __be16 h_vlan_proto; 3256 __be16 h_vlan_TCI; 3257 __u8 smac[6]; /* ETH_ALEN */ 3258 __u8 dmac[6]; /* ETH_ALEN */ 3259 }; 3260 3261 enum bpf_task_fd_type { 3262 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3263 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3264 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3265 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3266 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3267 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3268 }; 3269 3270 struct bpf_flow_keys { 3271 __u16 nhoff; 3272 __u16 thoff; 3273 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3274 __u8 is_frag; 3275 __u8 is_first_frag; 3276 __u8 is_encap; 3277 __u8 ip_proto; 3278 __be16 n_proto; 3279 __be16 sport; 3280 __be16 dport; 3281 union { 3282 struct { 3283 __be32 ipv4_src; 3284 __be32 ipv4_dst; 3285 }; 3286 struct { 3287 __u32 ipv6_src[4]; /* in6_addr; network order */ 3288 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3289 }; 3290 }; 3291 }; 3292 3293 struct bpf_func_info { 3294 __u32 insn_off; 3295 __u32 type_id; 3296 }; 3297 3298 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3299 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3300 3301 struct bpf_line_info { 3302 __u32 insn_off; 3303 __u32 file_name_off; 3304 __u32 line_off; 3305 __u32 line_col; 3306 }; 3307 3308 struct bpf_spin_lock { 3309 __u32 val; 3310 }; 3311 #endif /* _UAPI__LINUX_BPF_H__ */ 3312