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 }; 109 110 enum bpf_map_type { 111 BPF_MAP_TYPE_UNSPEC, 112 BPF_MAP_TYPE_HASH, 113 BPF_MAP_TYPE_ARRAY, 114 BPF_MAP_TYPE_PROG_ARRAY, 115 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 116 BPF_MAP_TYPE_PERCPU_HASH, 117 BPF_MAP_TYPE_PERCPU_ARRAY, 118 BPF_MAP_TYPE_STACK_TRACE, 119 BPF_MAP_TYPE_CGROUP_ARRAY, 120 BPF_MAP_TYPE_LRU_HASH, 121 BPF_MAP_TYPE_LRU_PERCPU_HASH, 122 BPF_MAP_TYPE_LPM_TRIE, 123 BPF_MAP_TYPE_ARRAY_OF_MAPS, 124 BPF_MAP_TYPE_HASH_OF_MAPS, 125 BPF_MAP_TYPE_DEVMAP, 126 BPF_MAP_TYPE_SOCKMAP, 127 BPF_MAP_TYPE_CPUMAP, 128 BPF_MAP_TYPE_XSKMAP, 129 BPF_MAP_TYPE_SOCKHASH, 130 BPF_MAP_TYPE_CGROUP_STORAGE, 131 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 132 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 133 BPF_MAP_TYPE_QUEUE, 134 BPF_MAP_TYPE_STACK, 135 }; 136 137 /* Note that tracing related programs such as 138 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 139 * are not subject to a stable API since kernel internal data 140 * structures can change from release to release and may 141 * therefore break existing tracing BPF programs. Tracing BPF 142 * programs correspond to /a/ specific kernel which is to be 143 * analyzed, and not /a/ specific kernel /and/ all future ones. 144 */ 145 enum bpf_prog_type { 146 BPF_PROG_TYPE_UNSPEC, 147 BPF_PROG_TYPE_SOCKET_FILTER, 148 BPF_PROG_TYPE_KPROBE, 149 BPF_PROG_TYPE_SCHED_CLS, 150 BPF_PROG_TYPE_SCHED_ACT, 151 BPF_PROG_TYPE_TRACEPOINT, 152 BPF_PROG_TYPE_XDP, 153 BPF_PROG_TYPE_PERF_EVENT, 154 BPF_PROG_TYPE_CGROUP_SKB, 155 BPF_PROG_TYPE_CGROUP_SOCK, 156 BPF_PROG_TYPE_LWT_IN, 157 BPF_PROG_TYPE_LWT_OUT, 158 BPF_PROG_TYPE_LWT_XMIT, 159 BPF_PROG_TYPE_SOCK_OPS, 160 BPF_PROG_TYPE_SK_SKB, 161 BPF_PROG_TYPE_CGROUP_DEVICE, 162 BPF_PROG_TYPE_SK_MSG, 163 BPF_PROG_TYPE_RAW_TRACEPOINT, 164 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 165 BPF_PROG_TYPE_LWT_SEG6LOCAL, 166 BPF_PROG_TYPE_LIRC_MODE2, 167 BPF_PROG_TYPE_SK_REUSEPORT, 168 BPF_PROG_TYPE_FLOW_DISSECTOR, 169 }; 170 171 enum bpf_attach_type { 172 BPF_CGROUP_INET_INGRESS, 173 BPF_CGROUP_INET_EGRESS, 174 BPF_CGROUP_INET_SOCK_CREATE, 175 BPF_CGROUP_SOCK_OPS, 176 BPF_SK_SKB_STREAM_PARSER, 177 BPF_SK_SKB_STREAM_VERDICT, 178 BPF_CGROUP_DEVICE, 179 BPF_SK_MSG_VERDICT, 180 BPF_CGROUP_INET4_BIND, 181 BPF_CGROUP_INET6_BIND, 182 BPF_CGROUP_INET4_CONNECT, 183 BPF_CGROUP_INET6_CONNECT, 184 BPF_CGROUP_INET4_POST_BIND, 185 BPF_CGROUP_INET6_POST_BIND, 186 BPF_CGROUP_UDP4_SENDMSG, 187 BPF_CGROUP_UDP6_SENDMSG, 188 BPF_LIRC_MODE2, 189 BPF_FLOW_DISSECTOR, 190 __MAX_BPF_ATTACH_TYPE 191 }; 192 193 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 194 195 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 196 * 197 * NONE(default): No further bpf programs allowed in the subtree. 198 * 199 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 200 * the program in this cgroup yields to sub-cgroup program. 201 * 202 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 203 * that cgroup program gets run in addition to the program in this cgroup. 204 * 205 * Only one program is allowed to be attached to a cgroup with 206 * NONE or BPF_F_ALLOW_OVERRIDE flag. 207 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 208 * release old program and attach the new one. Attach flags has to match. 209 * 210 * Multiple programs are allowed to be attached to a cgroup with 211 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 212 * (those that were attached first, run first) 213 * The programs of sub-cgroup are executed first, then programs of 214 * this cgroup and then programs of parent cgroup. 215 * When children program makes decision (like picking TCP CA or sock bind) 216 * parent program has a chance to override it. 217 * 218 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 219 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 220 * Ex1: 221 * cgrp1 (MULTI progs A, B) -> 222 * cgrp2 (OVERRIDE prog C) -> 223 * cgrp3 (MULTI prog D) -> 224 * cgrp4 (OVERRIDE prog E) -> 225 * cgrp5 (NONE prog F) 226 * the event in cgrp5 triggers execution of F,D,A,B in that order. 227 * if prog F is detached, the execution is E,D,A,B 228 * if prog F and D are detached, the execution is E,A,B 229 * if prog F, E and D are detached, the execution is C,A,B 230 * 231 * All eligible programs are executed regardless of return code from 232 * earlier programs. 233 */ 234 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 235 #define BPF_F_ALLOW_MULTI (1U << 1) 236 237 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 238 * verifier will perform strict alignment checking as if the kernel 239 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 240 * and NET_IP_ALIGN defined to 2. 241 */ 242 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 243 244 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 245 * verifier will allow any alignment whatsoever. On platforms 246 * with strict alignment requirements for loads ands stores (such 247 * as sparc and mips) the verifier validates that all loads and 248 * stores provably follow this requirement. This flag turns that 249 * checking and enforcement off. 250 * 251 * It is mostly used for testing when we want to validate the 252 * context and memory access aspects of the verifier, but because 253 * of an unaligned access the alignment check would trigger before 254 * the one we are interested in. 255 */ 256 #define BPF_F_ANY_ALIGNMENT (1U << 1) 257 258 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */ 259 #define BPF_PSEUDO_MAP_FD 1 260 261 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 262 * offset to another bpf function 263 */ 264 #define BPF_PSEUDO_CALL 1 265 266 /* flags for BPF_MAP_UPDATE_ELEM command */ 267 #define BPF_ANY 0 /* create new element or update existing */ 268 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 269 #define BPF_EXIST 2 /* update existing element */ 270 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 271 272 /* flags for BPF_MAP_CREATE command */ 273 #define BPF_F_NO_PREALLOC (1U << 0) 274 /* Instead of having one common LRU list in the 275 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 276 * which can scale and perform better. 277 * Note, the LRU nodes (including free nodes) cannot be moved 278 * across different LRU lists. 279 */ 280 #define BPF_F_NO_COMMON_LRU (1U << 1) 281 /* Specify numa node during map creation */ 282 #define BPF_F_NUMA_NODE (1U << 2) 283 284 #define BPF_OBJ_NAME_LEN 16U 285 286 /* Flags for accessing BPF object */ 287 #define BPF_F_RDONLY (1U << 3) 288 #define BPF_F_WRONLY (1U << 4) 289 290 /* Flag for stack_map, store build_id+offset instead of pointer */ 291 #define BPF_F_STACK_BUILD_ID (1U << 5) 292 293 /* Zero-initialize hash function seed. This should only be used for testing. */ 294 #define BPF_F_ZERO_SEED (1U << 6) 295 296 /* flags for BPF_PROG_QUERY */ 297 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 298 299 enum bpf_stack_build_id_status { 300 /* user space need an empty entry to identify end of a trace */ 301 BPF_STACK_BUILD_ID_EMPTY = 0, 302 /* with valid build_id and offset */ 303 BPF_STACK_BUILD_ID_VALID = 1, 304 /* couldn't get build_id, fallback to ip */ 305 BPF_STACK_BUILD_ID_IP = 2, 306 }; 307 308 #define BPF_BUILD_ID_SIZE 20 309 struct bpf_stack_build_id { 310 __s32 status; 311 unsigned char build_id[BPF_BUILD_ID_SIZE]; 312 union { 313 __u64 offset; 314 __u64 ip; 315 }; 316 }; 317 318 union bpf_attr { 319 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 320 __u32 map_type; /* one of enum bpf_map_type */ 321 __u32 key_size; /* size of key in bytes */ 322 __u32 value_size; /* size of value in bytes */ 323 __u32 max_entries; /* max number of entries in a map */ 324 __u32 map_flags; /* BPF_MAP_CREATE related 325 * flags defined above. 326 */ 327 __u32 inner_map_fd; /* fd pointing to the inner map */ 328 __u32 numa_node; /* numa node (effective only if 329 * BPF_F_NUMA_NODE is set). 330 */ 331 char map_name[BPF_OBJ_NAME_LEN]; 332 __u32 map_ifindex; /* ifindex of netdev to create on */ 333 __u32 btf_fd; /* fd pointing to a BTF type data */ 334 __u32 btf_key_type_id; /* BTF type_id of the key */ 335 __u32 btf_value_type_id; /* BTF type_id of the value */ 336 }; 337 338 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 339 __u32 map_fd; 340 __aligned_u64 key; 341 union { 342 __aligned_u64 value; 343 __aligned_u64 next_key; 344 }; 345 __u64 flags; 346 }; 347 348 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 349 __u32 prog_type; /* one of enum bpf_prog_type */ 350 __u32 insn_cnt; 351 __aligned_u64 insns; 352 __aligned_u64 license; 353 __u32 log_level; /* verbosity level of verifier */ 354 __u32 log_size; /* size of user buffer */ 355 __aligned_u64 log_buf; /* user supplied buffer */ 356 __u32 kern_version; /* not used */ 357 __u32 prog_flags; 358 char prog_name[BPF_OBJ_NAME_LEN]; 359 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 360 /* For some prog types expected attach type must be known at 361 * load time to verify attach type specific parts of prog 362 * (context accesses, allowed helpers, etc). 363 */ 364 __u32 expected_attach_type; 365 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 366 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 367 __aligned_u64 func_info; /* func info */ 368 __u32 func_info_cnt; /* number of bpf_func_info records */ 369 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 370 __aligned_u64 line_info; /* line info */ 371 __u32 line_info_cnt; /* number of bpf_line_info records */ 372 }; 373 374 struct { /* anonymous struct used by BPF_OBJ_* commands */ 375 __aligned_u64 pathname; 376 __u32 bpf_fd; 377 __u32 file_flags; 378 }; 379 380 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 381 __u32 target_fd; /* container object to attach to */ 382 __u32 attach_bpf_fd; /* eBPF program to attach */ 383 __u32 attach_type; 384 __u32 attach_flags; 385 }; 386 387 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 388 __u32 prog_fd; 389 __u32 retval; 390 __u32 data_size_in; /* input: len of data_in */ 391 __u32 data_size_out; /* input/output: len of data_out 392 * returns ENOSPC if data_out 393 * is too small. 394 */ 395 __aligned_u64 data_in; 396 __aligned_u64 data_out; 397 __u32 repeat; 398 __u32 duration; 399 } test; 400 401 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 402 union { 403 __u32 start_id; 404 __u32 prog_id; 405 __u32 map_id; 406 __u32 btf_id; 407 }; 408 __u32 next_id; 409 __u32 open_flags; 410 }; 411 412 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 413 __u32 bpf_fd; 414 __u32 info_len; 415 __aligned_u64 info; 416 } info; 417 418 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 419 __u32 target_fd; /* container object to query */ 420 __u32 attach_type; 421 __u32 query_flags; 422 __u32 attach_flags; 423 __aligned_u64 prog_ids; 424 __u32 prog_cnt; 425 } query; 426 427 struct { 428 __u64 name; 429 __u32 prog_fd; 430 } raw_tracepoint; 431 432 struct { /* anonymous struct for BPF_BTF_LOAD */ 433 __aligned_u64 btf; 434 __aligned_u64 btf_log_buf; 435 __u32 btf_size; 436 __u32 btf_log_size; 437 __u32 btf_log_level; 438 }; 439 440 struct { 441 __u32 pid; /* input: pid */ 442 __u32 fd; /* input: fd */ 443 __u32 flags; /* input: flags */ 444 __u32 buf_len; /* input/output: buf len */ 445 __aligned_u64 buf; /* input/output: 446 * tp_name for tracepoint 447 * symbol for kprobe 448 * filename for uprobe 449 */ 450 __u32 prog_id; /* output: prod_id */ 451 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 452 __u64 probe_offset; /* output: probe_offset */ 453 __u64 probe_addr; /* output: probe_addr */ 454 } task_fd_query; 455 } __attribute__((aligned(8))); 456 457 /* The description below is an attempt at providing documentation to eBPF 458 * developers about the multiple available eBPF helper functions. It can be 459 * parsed and used to produce a manual page. The workflow is the following, 460 * and requires the rst2man utility: 461 * 462 * $ ./scripts/bpf_helpers_doc.py \ 463 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 464 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 465 * $ man /tmp/bpf-helpers.7 466 * 467 * Note that in order to produce this external documentation, some RST 468 * formatting is used in the descriptions to get "bold" and "italics" in 469 * manual pages. Also note that the few trailing white spaces are 470 * intentional, removing them would break paragraphs for rst2man. 471 * 472 * Start of BPF helper function descriptions: 473 * 474 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 475 * Description 476 * Perform a lookup in *map* for an entry associated to *key*. 477 * Return 478 * Map value associated to *key*, or **NULL** if no entry was 479 * found. 480 * 481 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 482 * Description 483 * Add or update the value of the entry associated to *key* in 484 * *map* with *value*. *flags* is one of: 485 * 486 * **BPF_NOEXIST** 487 * The entry for *key* must not exist in the map. 488 * **BPF_EXIST** 489 * The entry for *key* must already exist in the map. 490 * **BPF_ANY** 491 * No condition on the existence of the entry for *key*. 492 * 493 * Flag value **BPF_NOEXIST** cannot be used for maps of types 494 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 495 * elements always exist), the helper would return an error. 496 * Return 497 * 0 on success, or a negative error in case of failure. 498 * 499 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 500 * Description 501 * Delete entry with *key* from *map*. 502 * Return 503 * 0 on success, or a negative error in case of failure. 504 * 505 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 506 * Description 507 * Push an element *value* in *map*. *flags* is one of: 508 * 509 * **BPF_EXIST** 510 * If the queue/stack is full, the oldest element is removed to 511 * make room for this. 512 * Return 513 * 0 on success, or a negative error in case of failure. 514 * 515 * int bpf_probe_read(void *dst, u32 size, const void *src) 516 * Description 517 * For tracing programs, safely attempt to read *size* bytes from 518 * address *src* and store the data in *dst*. 519 * Return 520 * 0 on success, or a negative error in case of failure. 521 * 522 * u64 bpf_ktime_get_ns(void) 523 * Description 524 * Return the time elapsed since system boot, in nanoseconds. 525 * Return 526 * Current *ktime*. 527 * 528 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 529 * Description 530 * This helper is a "printk()-like" facility for debugging. It 531 * prints a message defined by format *fmt* (of size *fmt_size*) 532 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 533 * available. It can take up to three additional **u64** 534 * arguments (as an eBPF helpers, the total number of arguments is 535 * limited to five). 536 * 537 * Each time the helper is called, it appends a line to the trace. 538 * The format of the trace is customizable, and the exact output 539 * one will get depends on the options set in 540 * *\/sys/kernel/debug/tracing/trace_options* (see also the 541 * *README* file under the same directory). However, it usually 542 * defaults to something like: 543 * 544 * :: 545 * 546 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 547 * 548 * In the above: 549 * 550 * * ``telnet`` is the name of the current task. 551 * * ``470`` is the PID of the current task. 552 * * ``001`` is the CPU number on which the task is 553 * running. 554 * * In ``.N..``, each character refers to a set of 555 * options (whether irqs are enabled, scheduling 556 * options, whether hard/softirqs are running, level of 557 * preempt_disabled respectively). **N** means that 558 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 559 * are set. 560 * * ``419421.045894`` is a timestamp. 561 * * ``0x00000001`` is a fake value used by BPF for the 562 * instruction pointer register. 563 * * ``<formatted msg>`` is the message formatted with 564 * *fmt*. 565 * 566 * The conversion specifiers supported by *fmt* are similar, but 567 * more limited than for printk(). They are **%d**, **%i**, 568 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 569 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 570 * of field, padding with zeroes, etc.) is available, and the 571 * helper will return **-EINVAL** (but print nothing) if it 572 * encounters an unknown specifier. 573 * 574 * Also, note that **bpf_trace_printk**\ () is slow, and should 575 * only be used for debugging purposes. For this reason, a notice 576 * bloc (spanning several lines) is printed to kernel logs and 577 * states that the helper should not be used "for production use" 578 * the first time this helper is used (or more precisely, when 579 * **trace_printk**\ () buffers are allocated). For passing values 580 * to user space, perf events should be preferred. 581 * Return 582 * The number of bytes written to the buffer, or a negative error 583 * in case of failure. 584 * 585 * u32 bpf_get_prandom_u32(void) 586 * Description 587 * Get a pseudo-random number. 588 * 589 * From a security point of view, this helper uses its own 590 * pseudo-random internal state, and cannot be used to infer the 591 * seed of other random functions in the kernel. However, it is 592 * essential to note that the generator used by the helper is not 593 * cryptographically secure. 594 * Return 595 * A random 32-bit unsigned value. 596 * 597 * u32 bpf_get_smp_processor_id(void) 598 * Description 599 * Get the SMP (symmetric multiprocessing) processor id. Note that 600 * all programs run with preemption disabled, which means that the 601 * SMP processor id is stable during all the execution of the 602 * program. 603 * Return 604 * The SMP id of the processor running the program. 605 * 606 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 607 * Description 608 * Store *len* bytes from address *from* into the packet 609 * associated to *skb*, at *offset*. *flags* are a combination of 610 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 611 * checksum for the packet after storing the bytes) and 612 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 613 * **->swhash** and *skb*\ **->l4hash** to 0). 614 * 615 * A call to this helper is susceptible to change the underlaying 616 * packet buffer. Therefore, at load time, all checks on pointers 617 * previously done by the verifier are invalidated and must be 618 * performed again, if the helper is used in combination with 619 * direct packet access. 620 * Return 621 * 0 on success, or a negative error in case of failure. 622 * 623 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 624 * Description 625 * Recompute the layer 3 (e.g. IP) checksum for the packet 626 * associated to *skb*. Computation is incremental, so the helper 627 * must know the former value of the header field that was 628 * modified (*from*), the new value of this field (*to*), and the 629 * number of bytes (2 or 4) for this field, stored in *size*. 630 * Alternatively, it is possible to store the difference between 631 * the previous and the new values of the header field in *to*, by 632 * setting *from* and *size* to 0. For both methods, *offset* 633 * indicates the location of the IP checksum within the packet. 634 * 635 * This helper works in combination with **bpf_csum_diff**\ (), 636 * which does not update the checksum in-place, but offers more 637 * flexibility and can handle sizes larger than 2 or 4 for the 638 * checksum to update. 639 * 640 * A call to this helper is susceptible to change the underlaying 641 * packet buffer. Therefore, at load time, all checks on pointers 642 * previously done by the verifier are invalidated and must be 643 * performed again, if the helper is used in combination with 644 * direct packet access. 645 * Return 646 * 0 on success, or a negative error in case of failure. 647 * 648 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 649 * Description 650 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 651 * packet associated to *skb*. Computation is incremental, so the 652 * helper must know the former value of the header field that was 653 * modified (*from*), the new value of this field (*to*), and the 654 * number of bytes (2 or 4) for this field, stored on the lowest 655 * four bits of *flags*. Alternatively, it is possible to store 656 * the difference between the previous and the new values of the 657 * header field in *to*, by setting *from* and the four lowest 658 * bits of *flags* to 0. For both methods, *offset* indicates the 659 * location of the IP checksum within the packet. In addition to 660 * the size of the field, *flags* can be added (bitwise OR) actual 661 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 662 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 663 * for updates resulting in a null checksum the value is set to 664 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 665 * the checksum is to be computed against a pseudo-header. 666 * 667 * This helper works in combination with **bpf_csum_diff**\ (), 668 * which does not update the checksum in-place, but offers more 669 * flexibility and can handle sizes larger than 2 or 4 for the 670 * checksum to update. 671 * 672 * A call to this helper is susceptible to change the underlaying 673 * packet buffer. Therefore, at load time, all checks on pointers 674 * previously done by the verifier are invalidated and must be 675 * performed again, if the helper is used in combination with 676 * direct packet access. 677 * Return 678 * 0 on success, or a negative error in case of failure. 679 * 680 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 681 * Description 682 * This special helper is used to trigger a "tail call", or in 683 * other words, to jump into another eBPF program. The same stack 684 * frame is used (but values on stack and in registers for the 685 * caller are not accessible to the callee). This mechanism allows 686 * for program chaining, either for raising the maximum number of 687 * available eBPF instructions, or to execute given programs in 688 * conditional blocks. For security reasons, there is an upper 689 * limit to the number of successive tail calls that can be 690 * performed. 691 * 692 * Upon call of this helper, the program attempts to jump into a 693 * program referenced at index *index* in *prog_array_map*, a 694 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 695 * *ctx*, a pointer to the context. 696 * 697 * If the call succeeds, the kernel immediately runs the first 698 * instruction of the new program. This is not a function call, 699 * and it never returns to the previous program. If the call 700 * fails, then the helper has no effect, and the caller continues 701 * to run its subsequent instructions. A call can fail if the 702 * destination program for the jump does not exist (i.e. *index* 703 * is superior to the number of entries in *prog_array_map*), or 704 * if the maximum number of tail calls has been reached for this 705 * chain of programs. This limit is defined in the kernel by the 706 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 707 * which is currently set to 32. 708 * Return 709 * 0 on success, or a negative error in case of failure. 710 * 711 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 712 * Description 713 * Clone and redirect the packet associated to *skb* to another 714 * net device of index *ifindex*. Both ingress and egress 715 * interfaces can be used for redirection. The **BPF_F_INGRESS** 716 * value in *flags* is used to make the distinction (ingress path 717 * is selected if the flag is present, egress path otherwise). 718 * This is the only flag supported for now. 719 * 720 * In comparison with **bpf_redirect**\ () helper, 721 * **bpf_clone_redirect**\ () has the associated cost of 722 * duplicating the packet buffer, but this can be executed out of 723 * the eBPF program. Conversely, **bpf_redirect**\ () is more 724 * efficient, but it is handled through an action code where the 725 * redirection happens only after the eBPF program has returned. 726 * 727 * A call to this helper is susceptible to change the underlaying 728 * packet buffer. Therefore, at load time, all checks on pointers 729 * previously done by the verifier are invalidated and must be 730 * performed again, if the helper is used in combination with 731 * direct packet access. 732 * Return 733 * 0 on success, or a negative error in case of failure. 734 * 735 * u64 bpf_get_current_pid_tgid(void) 736 * Return 737 * A 64-bit integer containing the current tgid and pid, and 738 * created as such: 739 * *current_task*\ **->tgid << 32 \|** 740 * *current_task*\ **->pid**. 741 * 742 * u64 bpf_get_current_uid_gid(void) 743 * Return 744 * A 64-bit integer containing the current GID and UID, and 745 * created as such: *current_gid* **<< 32 \|** *current_uid*. 746 * 747 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 748 * Description 749 * Copy the **comm** attribute of the current task into *buf* of 750 * *size_of_buf*. The **comm** attribute contains the name of 751 * the executable (excluding the path) for the current task. The 752 * *size_of_buf* must be strictly positive. On success, the 753 * helper makes sure that the *buf* is NUL-terminated. On failure, 754 * it is filled with zeroes. 755 * Return 756 * 0 on success, or a negative error in case of failure. 757 * 758 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 759 * Description 760 * Retrieve the classid for the current task, i.e. for the net_cls 761 * cgroup to which *skb* belongs. 762 * 763 * This helper can be used on TC egress path, but not on ingress. 764 * 765 * The net_cls cgroup provides an interface to tag network packets 766 * based on a user-provided identifier for all traffic coming from 767 * the tasks belonging to the related cgroup. See also the related 768 * kernel documentation, available from the Linux sources in file 769 * *Documentation/cgroup-v1/net_cls.txt*. 770 * 771 * The Linux kernel has two versions for cgroups: there are 772 * cgroups v1 and cgroups v2. Both are available to users, who can 773 * use a mixture of them, but note that the net_cls cgroup is for 774 * cgroup v1 only. This makes it incompatible with BPF programs 775 * run on cgroups, which is a cgroup-v2-only feature (a socket can 776 * only hold data for one version of cgroups at a time). 777 * 778 * This helper is only available is the kernel was compiled with 779 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 780 * "**y**" or to "**m**". 781 * Return 782 * The classid, or 0 for the default unconfigured classid. 783 * 784 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 785 * Description 786 * Push a *vlan_tci* (VLAN tag control information) of protocol 787 * *vlan_proto* to the packet associated to *skb*, then update 788 * the checksum. Note that if *vlan_proto* is different from 789 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 790 * be **ETH_P_8021Q**. 791 * 792 * A call to this helper is susceptible to change the underlaying 793 * packet buffer. Therefore, at load time, all checks on pointers 794 * previously done by the verifier are invalidated and must be 795 * performed again, if the helper is used in combination with 796 * direct packet access. 797 * Return 798 * 0 on success, or a negative error in case of failure. 799 * 800 * int bpf_skb_vlan_pop(struct sk_buff *skb) 801 * Description 802 * Pop a VLAN header from the packet associated to *skb*. 803 * 804 * A call to this helper is susceptible to change the underlaying 805 * packet buffer. Therefore, at load time, all checks on pointers 806 * previously done by the verifier are invalidated and must be 807 * performed again, if the helper is used in combination with 808 * direct packet access. 809 * Return 810 * 0 on success, or a negative error in case of failure. 811 * 812 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 813 * Description 814 * Get tunnel metadata. This helper takes a pointer *key* to an 815 * empty **struct bpf_tunnel_key** of **size**, that will be 816 * filled with tunnel metadata for the packet associated to *skb*. 817 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 818 * indicates that the tunnel is based on IPv6 protocol instead of 819 * IPv4. 820 * 821 * The **struct bpf_tunnel_key** is an object that generalizes the 822 * principal parameters used by various tunneling protocols into a 823 * single struct. This way, it can be used to easily make a 824 * decision based on the contents of the encapsulation header, 825 * "summarized" in this struct. In particular, it holds the IP 826 * address of the remote end (IPv4 or IPv6, depending on the case) 827 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 828 * this struct exposes the *key*\ **->tunnel_id**, which is 829 * generally mapped to a VNI (Virtual Network Identifier), making 830 * it programmable together with the **bpf_skb_set_tunnel_key**\ 831 * () helper. 832 * 833 * Let's imagine that the following code is part of a program 834 * attached to the TC ingress interface, on one end of a GRE 835 * tunnel, and is supposed to filter out all messages coming from 836 * remote ends with IPv4 address other than 10.0.0.1: 837 * 838 * :: 839 * 840 * int ret; 841 * struct bpf_tunnel_key key = {}; 842 * 843 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 844 * if (ret < 0) 845 * return TC_ACT_SHOT; // drop packet 846 * 847 * if (key.remote_ipv4 != 0x0a000001) 848 * return TC_ACT_SHOT; // drop packet 849 * 850 * return TC_ACT_OK; // accept packet 851 * 852 * This interface can also be used with all encapsulation devices 853 * that can operate in "collect metadata" mode: instead of having 854 * one network device per specific configuration, the "collect 855 * metadata" mode only requires a single device where the 856 * configuration can be extracted from this helper. 857 * 858 * This can be used together with various tunnels such as VXLan, 859 * Geneve, GRE or IP in IP (IPIP). 860 * Return 861 * 0 on success, or a negative error in case of failure. 862 * 863 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 864 * Description 865 * Populate tunnel metadata for packet associated to *skb.* The 866 * tunnel metadata is set to the contents of *key*, of *size*. The 867 * *flags* can be set to a combination of the following values: 868 * 869 * **BPF_F_TUNINFO_IPV6** 870 * Indicate that the tunnel is based on IPv6 protocol 871 * instead of IPv4. 872 * **BPF_F_ZERO_CSUM_TX** 873 * For IPv4 packets, add a flag to tunnel metadata 874 * indicating that checksum computation should be skipped 875 * and checksum set to zeroes. 876 * **BPF_F_DONT_FRAGMENT** 877 * Add a flag to tunnel metadata indicating that the 878 * packet should not be fragmented. 879 * **BPF_F_SEQ_NUMBER** 880 * Add a flag to tunnel metadata indicating that a 881 * sequence number should be added to tunnel header before 882 * sending the packet. This flag was added for GRE 883 * encapsulation, but might be used with other protocols 884 * as well in the future. 885 * 886 * Here is a typical usage on the transmit path: 887 * 888 * :: 889 * 890 * struct bpf_tunnel_key key; 891 * populate key ... 892 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 893 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 894 * 895 * See also the description of the **bpf_skb_get_tunnel_key**\ () 896 * helper for additional information. 897 * Return 898 * 0 on success, or a negative error in case of failure. 899 * 900 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 901 * Description 902 * Read the value of a perf event counter. This helper relies on a 903 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 904 * the perf event counter is selected when *map* is updated with 905 * perf event file descriptors. The *map* is an array whose size 906 * is the number of available CPUs, and each cell contains a value 907 * relative to one CPU. The value to retrieve is indicated by 908 * *flags*, that contains the index of the CPU to look up, masked 909 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 910 * **BPF_F_CURRENT_CPU** to indicate that the value for the 911 * current CPU should be retrieved. 912 * 913 * Note that before Linux 4.13, only hardware perf event can be 914 * retrieved. 915 * 916 * Also, be aware that the newer helper 917 * **bpf_perf_event_read_value**\ () is recommended over 918 * **bpf_perf_event_read**\ () in general. The latter has some ABI 919 * quirks where error and counter value are used as a return code 920 * (which is wrong to do since ranges may overlap). This issue is 921 * fixed with **bpf_perf_event_read_value**\ (), which at the same 922 * time provides more features over the **bpf_perf_event_read**\ 923 * () interface. Please refer to the description of 924 * **bpf_perf_event_read_value**\ () for details. 925 * Return 926 * The value of the perf event counter read from the map, or a 927 * negative error code in case of failure. 928 * 929 * int bpf_redirect(u32 ifindex, u64 flags) 930 * Description 931 * Redirect the packet to another net device of index *ifindex*. 932 * This helper is somewhat similar to **bpf_clone_redirect**\ 933 * (), except that the packet is not cloned, which provides 934 * increased performance. 935 * 936 * Except for XDP, both ingress and egress interfaces can be used 937 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 938 * to make the distinction (ingress path is selected if the flag 939 * is present, egress path otherwise). Currently, XDP only 940 * supports redirection to the egress interface, and accepts no 941 * flag at all. 942 * 943 * The same effect can be attained with the more generic 944 * **bpf_redirect_map**\ (), which requires specific maps to be 945 * used but offers better performance. 946 * Return 947 * For XDP, the helper returns **XDP_REDIRECT** on success or 948 * **XDP_ABORTED** on error. For other program types, the values 949 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 950 * error. 951 * 952 * u32 bpf_get_route_realm(struct sk_buff *skb) 953 * Description 954 * Retrieve the realm or the route, that is to say the 955 * **tclassid** field of the destination for the *skb*. The 956 * indentifier retrieved is a user-provided tag, similar to the 957 * one used with the net_cls cgroup (see description for 958 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 959 * held by a route (a destination entry), not by a task. 960 * 961 * Retrieving this identifier works with the clsact TC egress hook 962 * (see also **tc-bpf(8)**), or alternatively on conventional 963 * classful egress qdiscs, but not on TC ingress path. In case of 964 * clsact TC egress hook, this has the advantage that, internally, 965 * the destination entry has not been dropped yet in the transmit 966 * path. Therefore, the destination entry does not need to be 967 * artificially held via **netif_keep_dst**\ () for a classful 968 * qdisc until the *skb* is freed. 969 * 970 * This helper is available only if the kernel was compiled with 971 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 972 * Return 973 * The realm of the route for the packet associated to *skb*, or 0 974 * if none was found. 975 * 976 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 977 * Description 978 * Write raw *data* blob into a special BPF perf event held by 979 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 980 * event must have the following attributes: **PERF_SAMPLE_RAW** 981 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 982 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 983 * 984 * The *flags* are used to indicate the index in *map* for which 985 * the value must be put, masked with **BPF_F_INDEX_MASK**. 986 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 987 * to indicate that the index of the current CPU core should be 988 * used. 989 * 990 * The value to write, of *size*, is passed through eBPF stack and 991 * pointed by *data*. 992 * 993 * The context of the program *ctx* needs also be passed to the 994 * helper. 995 * 996 * On user space, a program willing to read the values needs to 997 * call **perf_event_open**\ () on the perf event (either for 998 * one or for all CPUs) and to store the file descriptor into the 999 * *map*. This must be done before the eBPF program can send data 1000 * into it. An example is available in file 1001 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1002 * tree (the eBPF program counterpart is in 1003 * *samples/bpf/trace_output_kern.c*). 1004 * 1005 * **bpf_perf_event_output**\ () achieves better performance 1006 * than **bpf_trace_printk**\ () for sharing data with user 1007 * space, and is much better suitable for streaming data from eBPF 1008 * programs. 1009 * 1010 * Note that this helper is not restricted to tracing use cases 1011 * and can be used with programs attached to TC or XDP as well, 1012 * where it allows for passing data to user space listeners. Data 1013 * can be: 1014 * 1015 * * Only custom structs, 1016 * * Only the packet payload, or 1017 * * A combination of both. 1018 * Return 1019 * 0 on success, or a negative error in case of failure. 1020 * 1021 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 1022 * Description 1023 * This helper was provided as an easy way to load data from a 1024 * packet. It can be used to load *len* bytes from *offset* from 1025 * the packet associated to *skb*, into the buffer pointed by 1026 * *to*. 1027 * 1028 * Since Linux 4.7, usage of this helper has mostly been replaced 1029 * by "direct packet access", enabling packet data to be 1030 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1031 * pointing respectively to the first byte of packet data and to 1032 * the byte after the last byte of packet data. However, it 1033 * remains useful if one wishes to read large quantities of data 1034 * at once from a packet into the eBPF stack. 1035 * Return 1036 * 0 on success, or a negative error in case of failure. 1037 * 1038 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 1039 * Description 1040 * Walk a user or a kernel stack and return its id. To achieve 1041 * this, the helper needs *ctx*, which is a pointer to the context 1042 * on which the tracing program is executed, and a pointer to a 1043 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1044 * 1045 * The last argument, *flags*, holds the number of stack frames to 1046 * skip (from 0 to 255), masked with 1047 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1048 * a combination of the following flags: 1049 * 1050 * **BPF_F_USER_STACK** 1051 * Collect a user space stack instead of a kernel stack. 1052 * **BPF_F_FAST_STACK_CMP** 1053 * Compare stacks by hash only. 1054 * **BPF_F_REUSE_STACKID** 1055 * If two different stacks hash into the same *stackid*, 1056 * discard the old one. 1057 * 1058 * The stack id retrieved is a 32 bit long integer handle which 1059 * can be further combined with other data (including other stack 1060 * ids) and used as a key into maps. This can be useful for 1061 * generating a variety of graphs (such as flame graphs or off-cpu 1062 * graphs). 1063 * 1064 * For walking a stack, this helper is an improvement over 1065 * **bpf_probe_read**\ (), which can be used with unrolled loops 1066 * but is not efficient and consumes a lot of eBPF instructions. 1067 * Instead, **bpf_get_stackid**\ () can collect up to 1068 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1069 * this limit can be controlled with the **sysctl** program, and 1070 * that it should be manually increased in order to profile long 1071 * user stacks (such as stacks for Java programs). To do so, use: 1072 * 1073 * :: 1074 * 1075 * # sysctl kernel.perf_event_max_stack=<new value> 1076 * Return 1077 * The positive or null stack id on success, or a negative error 1078 * in case of failure. 1079 * 1080 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1081 * Description 1082 * Compute a checksum difference, from the raw buffer pointed by 1083 * *from*, of length *from_size* (that must be a multiple of 4), 1084 * towards the raw buffer pointed by *to*, of size *to_size* 1085 * (same remark). An optional *seed* can be added to the value 1086 * (this can be cascaded, the seed may come from a previous call 1087 * to the helper). 1088 * 1089 * This is flexible enough to be used in several ways: 1090 * 1091 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1092 * checksum, it can be used when pushing new data. 1093 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1094 * checksum, it can be used when removing data from a packet. 1095 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1096 * can be used to compute a diff. Note that *from_size* and 1097 * *to_size* do not need to be equal. 1098 * 1099 * This helper can be used in combination with 1100 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1101 * which one can feed in the difference computed with 1102 * **bpf_csum_diff**\ (). 1103 * Return 1104 * The checksum result, or a negative error code in case of 1105 * failure. 1106 * 1107 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1108 * Description 1109 * Retrieve tunnel options metadata for the packet associated to 1110 * *skb*, and store the raw tunnel option data to the buffer *opt* 1111 * of *size*. 1112 * 1113 * This helper can be used with encapsulation devices that can 1114 * operate in "collect metadata" mode (please refer to the related 1115 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1116 * more details). A particular example where this can be used is 1117 * in combination with the Geneve encapsulation protocol, where it 1118 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1119 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1120 * the eBPF program. This allows for full customization of these 1121 * headers. 1122 * Return 1123 * The size of the option data retrieved. 1124 * 1125 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1126 * Description 1127 * Set tunnel options metadata for the packet associated to *skb* 1128 * to the option data contained in the raw buffer *opt* of *size*. 1129 * 1130 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1131 * helper for additional information. 1132 * Return 1133 * 0 on success, or a negative error in case of failure. 1134 * 1135 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1136 * Description 1137 * Change the protocol of the *skb* to *proto*. Currently 1138 * supported are transition from IPv4 to IPv6, and from IPv6 to 1139 * IPv4. The helper takes care of the groundwork for the 1140 * transition, including resizing the socket buffer. The eBPF 1141 * program is expected to fill the new headers, if any, via 1142 * **skb_store_bytes**\ () and to recompute the checksums with 1143 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1144 * (). The main case for this helper is to perform NAT64 1145 * operations out of an eBPF program. 1146 * 1147 * Internally, the GSO type is marked as dodgy so that headers are 1148 * checked and segments are recalculated by the GSO/GRO engine. 1149 * The size for GSO target is adapted as well. 1150 * 1151 * All values for *flags* are reserved for future usage, and must 1152 * be left at zero. 1153 * 1154 * A call to this helper is susceptible to change the underlaying 1155 * packet buffer. Therefore, at load time, all checks on pointers 1156 * previously done by the verifier are invalidated and must be 1157 * performed again, if the helper is used in combination with 1158 * direct packet access. 1159 * Return 1160 * 0 on success, or a negative error in case of failure. 1161 * 1162 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1163 * Description 1164 * Change the packet type for the packet associated to *skb*. This 1165 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1166 * the eBPF program does not have a write access to *skb*\ 1167 * **->pkt_type** beside this helper. Using a helper here allows 1168 * for graceful handling of errors. 1169 * 1170 * The major use case is to change incoming *skb*s to 1171 * **PACKET_HOST** in a programmatic way instead of having to 1172 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1173 * example. 1174 * 1175 * Note that *type* only allows certain values. At this time, they 1176 * are: 1177 * 1178 * **PACKET_HOST** 1179 * Packet is for us. 1180 * **PACKET_BROADCAST** 1181 * Send packet to all. 1182 * **PACKET_MULTICAST** 1183 * Send packet to group. 1184 * **PACKET_OTHERHOST** 1185 * Send packet to someone else. 1186 * Return 1187 * 0 on success, or a negative error in case of failure. 1188 * 1189 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1190 * Description 1191 * Check whether *skb* is a descendant of the cgroup2 held by 1192 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1193 * Return 1194 * The return value depends on the result of the test, and can be: 1195 * 1196 * * 0, if the *skb* failed the cgroup2 descendant test. 1197 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1198 * * A negative error code, if an error occurred. 1199 * 1200 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1201 * Description 1202 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1203 * not set, in particular if the hash was cleared due to mangling, 1204 * recompute this hash. Later accesses to the hash can be done 1205 * directly with *skb*\ **->hash**. 1206 * 1207 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1208 * prototype with **bpf_skb_change_proto**\ (), or calling 1209 * **bpf_skb_store_bytes**\ () with the 1210 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1211 * the hash and to trigger a new computation for the next call to 1212 * **bpf_get_hash_recalc**\ (). 1213 * Return 1214 * The 32-bit hash. 1215 * 1216 * u64 bpf_get_current_task(void) 1217 * Return 1218 * A pointer to the current task struct. 1219 * 1220 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1221 * Description 1222 * Attempt in a safe way to write *len* bytes from the buffer 1223 * *src* to *dst* in memory. It only works for threads that are in 1224 * user context, and *dst* must be a valid user space address. 1225 * 1226 * This helper should not be used to implement any kind of 1227 * security mechanism because of TOC-TOU attacks, but rather to 1228 * debug, divert, and manipulate execution of semi-cooperative 1229 * processes. 1230 * 1231 * Keep in mind that this feature is meant for experiments, and it 1232 * has a risk of crashing the system and running programs. 1233 * Therefore, when an eBPF program using this helper is attached, 1234 * a warning including PID and process name is printed to kernel 1235 * logs. 1236 * Return 1237 * 0 on success, or a negative error in case of failure. 1238 * 1239 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1240 * Description 1241 * Check whether the probe is being run is the context of a given 1242 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1243 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1244 * Return 1245 * The return value depends on the result of the test, and can be: 1246 * 1247 * * 0, if the *skb* task belongs to the cgroup2. 1248 * * 1, if the *skb* task does not belong to the cgroup2. 1249 * * A negative error code, if an error occurred. 1250 * 1251 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1252 * Description 1253 * Resize (trim or grow) the packet associated to *skb* to the 1254 * new *len*. The *flags* are reserved for future usage, and must 1255 * be left at zero. 1256 * 1257 * The basic idea is that the helper performs the needed work to 1258 * change the size of the packet, then the eBPF program rewrites 1259 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1260 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1261 * and others. This helper is a slow path utility intended for 1262 * replies with control messages. And because it is targeted for 1263 * slow path, the helper itself can afford to be slow: it 1264 * implicitly linearizes, unclones and drops offloads from the 1265 * *skb*. 1266 * 1267 * A call to this helper is susceptible to change the underlaying 1268 * packet buffer. Therefore, at load time, all checks on pointers 1269 * previously done by the verifier are invalidated and must be 1270 * performed again, if the helper is used in combination with 1271 * direct packet access. 1272 * Return 1273 * 0 on success, or a negative error in case of failure. 1274 * 1275 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1276 * Description 1277 * Pull in non-linear data in case the *skb* is non-linear and not 1278 * all of *len* are part of the linear section. Make *len* bytes 1279 * from *skb* readable and writable. If a zero value is passed for 1280 * *len*, then the whole length of the *skb* is pulled. 1281 * 1282 * This helper is only needed for reading and writing with direct 1283 * packet access. 1284 * 1285 * For direct packet access, testing that offsets to access 1286 * are within packet boundaries (test on *skb*\ **->data_end**) is 1287 * susceptible to fail if offsets are invalid, or if the requested 1288 * data is in non-linear parts of the *skb*. On failure the 1289 * program can just bail out, or in the case of a non-linear 1290 * buffer, use a helper to make the data available. The 1291 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1292 * the data. Another one consists in using **bpf_skb_pull_data** 1293 * to pull in once the non-linear parts, then retesting and 1294 * eventually access the data. 1295 * 1296 * At the same time, this also makes sure the *skb* is uncloned, 1297 * which is a necessary condition for direct write. As this needs 1298 * to be an invariant for the write part only, the verifier 1299 * detects writes and adds a prologue that is calling 1300 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1301 * the very beginning in case it is indeed cloned. 1302 * 1303 * A call to this helper is susceptible to change the underlaying 1304 * packet buffer. Therefore, at load time, all checks on pointers 1305 * previously done by the verifier are invalidated and must be 1306 * performed again, if the helper is used in combination with 1307 * direct packet access. 1308 * Return 1309 * 0 on success, or a negative error in case of failure. 1310 * 1311 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1312 * Description 1313 * Add the checksum *csum* into *skb*\ **->csum** in case the 1314 * driver has supplied a checksum for the entire packet into that 1315 * field. Return an error otherwise. This helper is intended to be 1316 * used in combination with **bpf_csum_diff**\ (), in particular 1317 * when the checksum needs to be updated after data has been 1318 * written into the packet through direct packet access. 1319 * Return 1320 * The checksum on success, or a negative error code in case of 1321 * failure. 1322 * 1323 * void bpf_set_hash_invalid(struct sk_buff *skb) 1324 * Description 1325 * Invalidate the current *skb*\ **->hash**. It can be used after 1326 * mangling on headers through direct packet access, in order to 1327 * indicate that the hash is outdated and to trigger a 1328 * recalculation the next time the kernel tries to access this 1329 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1330 * 1331 * int bpf_get_numa_node_id(void) 1332 * Description 1333 * Return the id of the current NUMA node. The primary use case 1334 * for this helper is the selection of sockets for the local NUMA 1335 * node, when the program is attached to sockets using the 1336 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1337 * but the helper is also available to other eBPF program types, 1338 * similarly to **bpf_get_smp_processor_id**\ (). 1339 * Return 1340 * The id of current NUMA node. 1341 * 1342 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1343 * Description 1344 * Grows headroom of packet associated to *skb* and adjusts the 1345 * offset of the MAC header accordingly, adding *len* bytes of 1346 * space. It automatically extends and reallocates memory as 1347 * required. 1348 * 1349 * This helper can be used on a layer 3 *skb* to push a MAC header 1350 * for redirection into a layer 2 device. 1351 * 1352 * All values for *flags* are reserved for future usage, and must 1353 * be left at zero. 1354 * 1355 * A call to this helper is susceptible to change the underlaying 1356 * packet buffer. Therefore, at load time, all checks on pointers 1357 * previously done by the verifier are invalidated and must be 1358 * performed again, if the helper is used in combination with 1359 * direct packet access. 1360 * Return 1361 * 0 on success, or a negative error in case of failure. 1362 * 1363 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1364 * Description 1365 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1366 * it is possible to use a negative value for *delta*. This helper 1367 * can be used to prepare the packet for pushing or popping 1368 * headers. 1369 * 1370 * A call to this helper is susceptible to change the underlaying 1371 * packet buffer. Therefore, at load time, all checks on pointers 1372 * previously done by the verifier are invalidated and must be 1373 * performed again, if the helper is used in combination with 1374 * direct packet access. 1375 * Return 1376 * 0 on success, or a negative error in case of failure. 1377 * 1378 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1379 * Description 1380 * Copy a NUL terminated string from an unsafe address 1381 * *unsafe_ptr* to *dst*. The *size* should include the 1382 * terminating NUL byte. In case the string length is smaller than 1383 * *size*, the target is not padded with further NUL bytes. If the 1384 * string length is larger than *size*, just *size*-1 bytes are 1385 * copied and the last byte is set to NUL. 1386 * 1387 * On success, the length of the copied string is returned. This 1388 * makes this helper useful in tracing programs for reading 1389 * strings, and more importantly to get its length at runtime. See 1390 * the following snippet: 1391 * 1392 * :: 1393 * 1394 * SEC("kprobe/sys_open") 1395 * void bpf_sys_open(struct pt_regs *ctx) 1396 * { 1397 * char buf[PATHLEN]; // PATHLEN is defined to 256 1398 * int res = bpf_probe_read_str(buf, sizeof(buf), 1399 * ctx->di); 1400 * 1401 * // Consume buf, for example push it to 1402 * // userspace via bpf_perf_event_output(); we 1403 * // can use res (the string length) as event 1404 * // size, after checking its boundaries. 1405 * } 1406 * 1407 * In comparison, using **bpf_probe_read()** helper here instead 1408 * to read the string would require to estimate the length at 1409 * compile time, and would often result in copying more memory 1410 * than necessary. 1411 * 1412 * Another useful use case is when parsing individual process 1413 * arguments or individual environment variables navigating 1414 * *current*\ **->mm->arg_start** and *current*\ 1415 * **->mm->env_start**: using this helper and the return value, 1416 * one can quickly iterate at the right offset of the memory area. 1417 * Return 1418 * On success, the strictly positive length of the string, 1419 * including the trailing NUL character. On error, a negative 1420 * value. 1421 * 1422 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1423 * Description 1424 * If the **struct sk_buff** pointed by *skb* has a known socket, 1425 * retrieve the cookie (generated by the kernel) of this socket. 1426 * If no cookie has been set yet, generate a new cookie. Once 1427 * generated, the socket cookie remains stable for the life of the 1428 * socket. This helper can be useful for monitoring per socket 1429 * networking traffic statistics as it provides a unique socket 1430 * identifier per namespace. 1431 * Return 1432 * A 8-byte long non-decreasing number on success, or 0 if the 1433 * socket field is missing inside *skb*. 1434 * 1435 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1436 * Description 1437 * Equivalent to bpf_get_socket_cookie() helper that accepts 1438 * *skb*, but gets socket from **struct bpf_sock_addr** contex. 1439 * Return 1440 * A 8-byte long non-decreasing number. 1441 * 1442 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1443 * Description 1444 * Equivalent to bpf_get_socket_cookie() helper that accepts 1445 * *skb*, but gets socket from **struct bpf_sock_ops** contex. 1446 * Return 1447 * A 8-byte long non-decreasing number. 1448 * 1449 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1450 * Return 1451 * The owner UID of the socket associated to *skb*. If the socket 1452 * is **NULL**, or if it is not a full socket (i.e. if it is a 1453 * time-wait or a request socket instead), **overflowuid** value 1454 * is returned (note that **overflowuid** might also be the actual 1455 * UID value for the socket). 1456 * 1457 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1458 * Description 1459 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1460 * to value *hash*. 1461 * Return 1462 * 0 1463 * 1464 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1465 * Description 1466 * Emulate a call to **setsockopt()** on the socket associated to 1467 * *bpf_socket*, which must be a full socket. The *level* at 1468 * which the option resides and the name *optname* of the option 1469 * must be specified, see **setsockopt(2)** for more information. 1470 * The option value of length *optlen* is pointed by *optval*. 1471 * 1472 * This helper actually implements a subset of **setsockopt()**. 1473 * It supports the following *level*\ s: 1474 * 1475 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1476 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1477 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1478 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1479 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1480 * **TCP_BPF_SNDCWND_CLAMP**. 1481 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1482 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1483 * Return 1484 * 0 on success, or a negative error in case of failure. 1485 * 1486 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1487 * Description 1488 * Grow or shrink the room for data in the packet associated to 1489 * *skb* by *len_diff*, and according to the selected *mode*. 1490 * 1491 * There is a single supported mode at this time: 1492 * 1493 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1494 * (room space is added or removed below the layer 3 header). 1495 * 1496 * All values for *flags* are reserved for future usage, and must 1497 * be left at zero. 1498 * 1499 * A call to this helper is susceptible to change the underlaying 1500 * packet buffer. Therefore, at load time, all checks on pointers 1501 * previously done by the verifier are invalidated and must be 1502 * performed again, if the helper is used in combination with 1503 * direct packet access. 1504 * Return 1505 * 0 on success, or a negative error in case of failure. 1506 * 1507 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1508 * Description 1509 * Redirect the packet to the endpoint referenced by *map* at 1510 * index *key*. Depending on its type, this *map* can contain 1511 * references to net devices (for forwarding packets through other 1512 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1513 * but this is only implemented for native XDP (with driver 1514 * support) as of this writing). 1515 * 1516 * All values for *flags* are reserved for future usage, and must 1517 * be left at zero. 1518 * 1519 * When used to redirect packets to net devices, this helper 1520 * provides a high performance increase over **bpf_redirect**\ (). 1521 * This is due to various implementation details of the underlying 1522 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1523 * () tries to send packet as a "bulk" to the device. 1524 * Return 1525 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1526 * 1527 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1528 * Description 1529 * Redirect the packet to the socket referenced by *map* (of type 1530 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1531 * egress interfaces can be used for redirection. The 1532 * **BPF_F_INGRESS** value in *flags* is used to make the 1533 * distinction (ingress path is selected if the flag is present, 1534 * egress path otherwise). This is the only flag supported for now. 1535 * Return 1536 * **SK_PASS** on success, or **SK_DROP** on error. 1537 * 1538 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1539 * Description 1540 * Add an entry to, or update a *map* referencing sockets. The 1541 * *skops* is used as a new value for the entry associated to 1542 * *key*. *flags* is one of: 1543 * 1544 * **BPF_NOEXIST** 1545 * The entry for *key* must not exist in the map. 1546 * **BPF_EXIST** 1547 * The entry for *key* must already exist in the map. 1548 * **BPF_ANY** 1549 * No condition on the existence of the entry for *key*. 1550 * 1551 * If the *map* has eBPF programs (parser and verdict), those will 1552 * be inherited by the socket being added. If the socket is 1553 * already attached to eBPF programs, this results in an error. 1554 * Return 1555 * 0 on success, or a negative error in case of failure. 1556 * 1557 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1558 * Description 1559 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1560 * *delta* (which can be positive or negative). Note that this 1561 * operation modifies the address stored in *xdp_md*\ **->data**, 1562 * so the latter must be loaded only after the helper has been 1563 * called. 1564 * 1565 * The use of *xdp_md*\ **->data_meta** is optional and programs 1566 * are not required to use it. The rationale is that when the 1567 * packet is processed with XDP (e.g. as DoS filter), it is 1568 * possible to push further meta data along with it before passing 1569 * to the stack, and to give the guarantee that an ingress eBPF 1570 * program attached as a TC classifier on the same device can pick 1571 * this up for further post-processing. Since TC works with socket 1572 * buffers, it remains possible to set from XDP the **mark** or 1573 * **priority** pointers, or other pointers for the socket buffer. 1574 * Having this scratch space generic and programmable allows for 1575 * more flexibility as the user is free to store whatever meta 1576 * data they need. 1577 * 1578 * A call to this helper is susceptible to change the underlaying 1579 * packet buffer. Therefore, at load time, all checks on pointers 1580 * previously done by the verifier are invalidated and must be 1581 * performed again, if the helper is used in combination with 1582 * direct packet access. 1583 * Return 1584 * 0 on success, or a negative error in case of failure. 1585 * 1586 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1587 * Description 1588 * Read the value of a perf event counter, and store it into *buf* 1589 * of size *buf_size*. This helper relies on a *map* of type 1590 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1591 * counter is selected when *map* is updated with perf event file 1592 * descriptors. The *map* is an array whose size is the number of 1593 * available CPUs, and each cell contains a value relative to one 1594 * CPU. The value to retrieve is indicated by *flags*, that 1595 * contains the index of the CPU to look up, masked with 1596 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1597 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1598 * current CPU should be retrieved. 1599 * 1600 * This helper behaves in a way close to 1601 * **bpf_perf_event_read**\ () helper, save that instead of 1602 * just returning the value observed, it fills the *buf* 1603 * structure. This allows for additional data to be retrieved: in 1604 * particular, the enabled and running times (in *buf*\ 1605 * **->enabled** and *buf*\ **->running**, respectively) are 1606 * copied. In general, **bpf_perf_event_read_value**\ () is 1607 * recommended over **bpf_perf_event_read**\ (), which has some 1608 * ABI issues and provides fewer functionalities. 1609 * 1610 * These values are interesting, because hardware PMU (Performance 1611 * Monitoring Unit) counters are limited resources. When there are 1612 * more PMU based perf events opened than available counters, 1613 * kernel will multiplex these events so each event gets certain 1614 * percentage (but not all) of the PMU time. In case that 1615 * multiplexing happens, the number of samples or counter value 1616 * will not reflect the case compared to when no multiplexing 1617 * occurs. This makes comparison between different runs difficult. 1618 * Typically, the counter value should be normalized before 1619 * comparing to other experiments. The usual normalization is done 1620 * as follows. 1621 * 1622 * :: 1623 * 1624 * normalized_counter = counter * t_enabled / t_running 1625 * 1626 * Where t_enabled is the time enabled for event and t_running is 1627 * the time running for event since last normalization. The 1628 * enabled and running times are accumulated since the perf event 1629 * open. To achieve scaling factor between two invocations of an 1630 * eBPF program, users can can use CPU id as the key (which is 1631 * typical for perf array usage model) to remember the previous 1632 * value and do the calculation inside the eBPF program. 1633 * Return 1634 * 0 on success, or a negative error in case of failure. 1635 * 1636 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1637 * Description 1638 * For en eBPF program attached to a perf event, retrieve the 1639 * value of the event counter associated to *ctx* and store it in 1640 * the structure pointed by *buf* and of size *buf_size*. Enabled 1641 * and running times are also stored in the structure (see 1642 * description of helper **bpf_perf_event_read_value**\ () for 1643 * more details). 1644 * Return 1645 * 0 on success, or a negative error in case of failure. 1646 * 1647 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1648 * Description 1649 * Emulate a call to **getsockopt()** on the socket associated to 1650 * *bpf_socket*, which must be a full socket. The *level* at 1651 * which the option resides and the name *optname* of the option 1652 * must be specified, see **getsockopt(2)** for more information. 1653 * The retrieved value is stored in the structure pointed by 1654 * *opval* and of length *optlen*. 1655 * 1656 * This helper actually implements a subset of **getsockopt()**. 1657 * It supports the following *level*\ s: 1658 * 1659 * * **IPPROTO_TCP**, which supports *optname* 1660 * **TCP_CONGESTION**. 1661 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1662 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1663 * Return 1664 * 0 on success, or a negative error in case of failure. 1665 * 1666 * int bpf_override_return(struct pt_reg *regs, u64 rc) 1667 * Description 1668 * Used for error injection, this helper uses kprobes to override 1669 * the return value of the probed function, and to set it to *rc*. 1670 * The first argument is the context *regs* on which the kprobe 1671 * works. 1672 * 1673 * This helper works by setting setting the PC (program counter) 1674 * to an override function which is run in place of the original 1675 * probed function. This means the probed function is not run at 1676 * all. The replacement function just returns with the required 1677 * value. 1678 * 1679 * This helper has security implications, and thus is subject to 1680 * restrictions. It is only available if the kernel was compiled 1681 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1682 * option, and in this case it only works on functions tagged with 1683 * **ALLOW_ERROR_INJECTION** in the kernel code. 1684 * 1685 * Also, the helper is only available for the architectures having 1686 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1687 * x86 architecture is the only one to support this feature. 1688 * Return 1689 * 0 1690 * 1691 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1692 * Description 1693 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1694 * for the full TCP socket associated to *bpf_sock_ops* to 1695 * *argval*. 1696 * 1697 * The primary use of this field is to determine if there should 1698 * be calls to eBPF programs of type 1699 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1700 * code. A program of the same type can change its value, per 1701 * connection and as necessary, when the connection is 1702 * established. This field is directly accessible for reading, but 1703 * this helper must be used for updates in order to return an 1704 * error if an eBPF program tries to set a callback that is not 1705 * supported in the current kernel. 1706 * 1707 * The supported callback values that *argval* can combine are: 1708 * 1709 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1710 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1711 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1712 * 1713 * Here are some examples of where one could call such eBPF 1714 * program: 1715 * 1716 * * When RTO fires. 1717 * * When a packet is retransmitted. 1718 * * When the connection terminates. 1719 * * When a packet is sent. 1720 * * When a packet is received. 1721 * Return 1722 * Code **-EINVAL** if the socket is not a full TCP socket; 1723 * otherwise, a positive number containing the bits that could not 1724 * be set is returned (which comes down to 0 if all bits were set 1725 * as required). 1726 * 1727 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1728 * Description 1729 * This helper is used in programs implementing policies at the 1730 * socket level. If the message *msg* is allowed to pass (i.e. if 1731 * the verdict eBPF program returns **SK_PASS**), redirect it to 1732 * the socket referenced by *map* (of type 1733 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1734 * egress interfaces can be used for redirection. The 1735 * **BPF_F_INGRESS** value in *flags* is used to make the 1736 * distinction (ingress path is selected if the flag is present, 1737 * egress path otherwise). This is the only flag supported for now. 1738 * Return 1739 * **SK_PASS** on success, or **SK_DROP** on error. 1740 * 1741 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1742 * Description 1743 * For socket policies, apply the verdict of the eBPF program to 1744 * the next *bytes* (number of bytes) of message *msg*. 1745 * 1746 * For example, this helper can be used in the following cases: 1747 * 1748 * * A single **sendmsg**\ () or **sendfile**\ () system call 1749 * contains multiple logical messages that the eBPF program is 1750 * supposed to read and for which it should apply a verdict. 1751 * * An eBPF program only cares to read the first *bytes* of a 1752 * *msg*. If the message has a large payload, then setting up 1753 * and calling the eBPF program repeatedly for all bytes, even 1754 * though the verdict is already known, would create unnecessary 1755 * overhead. 1756 * 1757 * When called from within an eBPF program, the helper sets a 1758 * counter internal to the BPF infrastructure, that is used to 1759 * apply the last verdict to the next *bytes*. If *bytes* is 1760 * smaller than the current data being processed from a 1761 * **sendmsg**\ () or **sendfile**\ () system call, the first 1762 * *bytes* will be sent and the eBPF program will be re-run with 1763 * the pointer for start of data pointing to byte number *bytes* 1764 * **+ 1**. If *bytes* is larger than the current data being 1765 * processed, then the eBPF verdict will be applied to multiple 1766 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1767 * consumed. 1768 * 1769 * Note that if a socket closes with the internal counter holding 1770 * a non-zero value, this is not a problem because data is not 1771 * being buffered for *bytes* and is sent as it is received. 1772 * Return 1773 * 0 1774 * 1775 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1776 * Description 1777 * For socket policies, prevent the execution of the verdict eBPF 1778 * program for message *msg* until *bytes* (byte number) have been 1779 * accumulated. 1780 * 1781 * This can be used when one needs a specific number of bytes 1782 * before a verdict can be assigned, even if the data spans 1783 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1784 * case would be a user calling **sendmsg**\ () repeatedly with 1785 * 1-byte long message segments. Obviously, this is bad for 1786 * performance, but it is still valid. If the eBPF program needs 1787 * *bytes* bytes to validate a header, this helper can be used to 1788 * prevent the eBPF program to be called again until *bytes* have 1789 * been accumulated. 1790 * Return 1791 * 0 1792 * 1793 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1794 * Description 1795 * For socket policies, pull in non-linear data from user space 1796 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1797 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1798 * respectively. 1799 * 1800 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1801 * *msg* it can only parse data that the (**data**, **data_end**) 1802 * pointers have already consumed. For **sendmsg**\ () hooks this 1803 * is likely the first scatterlist element. But for calls relying 1804 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1805 * be the range (**0**, **0**) because the data is shared with 1806 * user space and by default the objective is to avoid allowing 1807 * user space to modify data while (or after) eBPF verdict is 1808 * being decided. This helper can be used to pull in data and to 1809 * set the start and end pointer to given values. Data will be 1810 * copied if necessary (i.e. if data was not linear and if start 1811 * and end pointers do not point to the same chunk). 1812 * 1813 * A call to this helper is susceptible to change the underlaying 1814 * packet buffer. Therefore, at load time, all checks on pointers 1815 * previously done by the verifier are invalidated and must be 1816 * performed again, if the helper is used in combination with 1817 * direct packet access. 1818 * 1819 * All values for *flags* are reserved for future usage, and must 1820 * be left at zero. 1821 * Return 1822 * 0 on success, or a negative error in case of failure. 1823 * 1824 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1825 * Description 1826 * Bind the socket associated to *ctx* to the address pointed by 1827 * *addr*, of length *addr_len*. This allows for making outgoing 1828 * connection from the desired IP address, which can be useful for 1829 * example when all processes inside a cgroup should use one 1830 * single IP address on a host that has multiple IP configured. 1831 * 1832 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1833 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1834 * **AF_INET6**). Looking for a free port to bind to can be 1835 * expensive, therefore binding to port is not permitted by the 1836 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1837 * must be set to zero. 1838 * Return 1839 * 0 on success, or a negative error in case of failure. 1840 * 1841 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1842 * Description 1843 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1844 * only possible to shrink the packet as of this writing, 1845 * therefore *delta* must be a negative integer. 1846 * 1847 * A call to this helper is susceptible to change the underlaying 1848 * packet buffer. Therefore, at load time, all checks on pointers 1849 * previously done by the verifier are invalidated and must be 1850 * performed again, if the helper is used in combination with 1851 * direct packet access. 1852 * Return 1853 * 0 on success, or a negative error in case of failure. 1854 * 1855 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1856 * Description 1857 * Retrieve the XFRM state (IP transform framework, see also 1858 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1859 * 1860 * The retrieved value is stored in the **struct bpf_xfrm_state** 1861 * pointed by *xfrm_state* and of length *size*. 1862 * 1863 * All values for *flags* are reserved for future usage, and must 1864 * be left at zero. 1865 * 1866 * This helper is available only if the kernel was compiled with 1867 * **CONFIG_XFRM** configuration option. 1868 * Return 1869 * 0 on success, or a negative error in case of failure. 1870 * 1871 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1872 * Description 1873 * Return a user or a kernel stack in bpf program provided buffer. 1874 * To achieve this, the helper needs *ctx*, which is a pointer 1875 * to the context on which the tracing program is executed. 1876 * To store the stacktrace, the bpf program provides *buf* with 1877 * a nonnegative *size*. 1878 * 1879 * The last argument, *flags*, holds the number of stack frames to 1880 * skip (from 0 to 255), masked with 1881 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1882 * the following flags: 1883 * 1884 * **BPF_F_USER_STACK** 1885 * Collect a user space stack instead of a kernel stack. 1886 * **BPF_F_USER_BUILD_ID** 1887 * Collect buildid+offset instead of ips for user stack, 1888 * only valid if **BPF_F_USER_STACK** is also specified. 1889 * 1890 * **bpf_get_stack**\ () can collect up to 1891 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1892 * to sufficient large buffer size. Note that 1893 * this limit can be controlled with the **sysctl** program, and 1894 * that it should be manually increased in order to profile long 1895 * user stacks (such as stacks for Java programs). To do so, use: 1896 * 1897 * :: 1898 * 1899 * # sysctl kernel.perf_event_max_stack=<new value> 1900 * Return 1901 * A non-negative value equal to or less than *size* on success, 1902 * or a negative error in case of failure. 1903 * 1904 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1905 * Description 1906 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1907 * it provides an easy way to load *len* bytes from *offset* 1908 * from the packet associated to *skb*, into the buffer pointed 1909 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1910 * a fifth argument *start_header* exists in order to select a 1911 * base offset to start from. *start_header* can be one of: 1912 * 1913 * **BPF_HDR_START_MAC** 1914 * Base offset to load data from is *skb*'s mac header. 1915 * **BPF_HDR_START_NET** 1916 * Base offset to load data from is *skb*'s network header. 1917 * 1918 * In general, "direct packet access" is the preferred method to 1919 * access packet data, however, this helper is in particular useful 1920 * in socket filters where *skb*\ **->data** does not always point 1921 * to the start of the mac header and where "direct packet access" 1922 * is not available. 1923 * Return 1924 * 0 on success, or a negative error in case of failure. 1925 * 1926 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 1927 * Description 1928 * Do FIB lookup in kernel tables using parameters in *params*. 1929 * If lookup is successful and result shows packet is to be 1930 * forwarded, the neighbor tables are searched for the nexthop. 1931 * If successful (ie., FIB lookup shows forwarding and nexthop 1932 * is resolved), the nexthop address is returned in ipv4_dst 1933 * or ipv6_dst based on family, smac is set to mac address of 1934 * egress device, dmac is set to nexthop mac address, rt_metric 1935 * is set to metric from route (IPv4/IPv6 only), and ifindex 1936 * is set to the device index of the nexthop from the FIB lookup. 1937 * 1938 * *plen* argument is the size of the passed in struct. 1939 * *flags* argument can be a combination of one or more of the 1940 * following values: 1941 * 1942 * **BPF_FIB_LOOKUP_DIRECT** 1943 * Do a direct table lookup vs full lookup using FIB 1944 * rules. 1945 * **BPF_FIB_LOOKUP_OUTPUT** 1946 * Perform lookup from an egress perspective (default is 1947 * ingress). 1948 * 1949 * *ctx* is either **struct xdp_md** for XDP programs or 1950 * **struct sk_buff** tc cls_act programs. 1951 * Return 1952 * * < 0 if any input argument is invalid 1953 * * 0 on success (packet is forwarded, nexthop neighbor exists) 1954 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 1955 * packet is not forwarded or needs assist from full stack 1956 * 1957 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 1958 * Description 1959 * Add an entry to, or update a sockhash *map* referencing sockets. 1960 * The *skops* is used as a new value for the entry associated to 1961 * *key*. *flags* is one of: 1962 * 1963 * **BPF_NOEXIST** 1964 * The entry for *key* must not exist in the map. 1965 * **BPF_EXIST** 1966 * The entry for *key* must already exist in the map. 1967 * **BPF_ANY** 1968 * No condition on the existence of the entry for *key*. 1969 * 1970 * If the *map* has eBPF programs (parser and verdict), those will 1971 * be inherited by the socket being added. If the socket is 1972 * already attached to eBPF programs, this results in an error. 1973 * Return 1974 * 0 on success, or a negative error in case of failure. 1975 * 1976 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 1977 * Description 1978 * This helper is used in programs implementing policies at the 1979 * socket level. If the message *msg* is allowed to pass (i.e. if 1980 * the verdict eBPF program returns **SK_PASS**), redirect it to 1981 * the socket referenced by *map* (of type 1982 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1983 * egress interfaces can be used for redirection. The 1984 * **BPF_F_INGRESS** value in *flags* is used to make the 1985 * distinction (ingress path is selected if the flag is present, 1986 * egress path otherwise). This is the only flag supported for now. 1987 * Return 1988 * **SK_PASS** on success, or **SK_DROP** on error. 1989 * 1990 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 1991 * Description 1992 * This helper is used in programs implementing policies at the 1993 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 1994 * if the verdeict eBPF program returns **SK_PASS**), redirect it 1995 * to the socket referenced by *map* (of type 1996 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1997 * egress interfaces can be used for redirection. The 1998 * **BPF_F_INGRESS** value in *flags* is used to make the 1999 * distinction (ingress path is selected if the flag is present, 2000 * egress otherwise). This is the only flag supported for now. 2001 * Return 2002 * **SK_PASS** on success, or **SK_DROP** on error. 2003 * 2004 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2005 * Description 2006 * Encapsulate the packet associated to *skb* within a Layer 3 2007 * protocol header. This header is provided in the buffer at 2008 * address *hdr*, with *len* its size in bytes. *type* indicates 2009 * the protocol of the header and can be one of: 2010 * 2011 * **BPF_LWT_ENCAP_SEG6** 2012 * IPv6 encapsulation with Segment Routing Header 2013 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2014 * the IPv6 header is computed by the kernel. 2015 * **BPF_LWT_ENCAP_SEG6_INLINE** 2016 * Only works if *skb* contains an IPv6 packet. Insert a 2017 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2018 * the IPv6 header. 2019 * **BPF_LWT_ENCAP_IP** 2020 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2021 * must be IPv4 or IPv6, followed by zero or more 2022 * additional headers, up to LWT_BPF_MAX_HEADROOM total 2023 * bytes in all prepended headers. Please note that 2024 * if skb_is_gso(skb) is true, no more than two headers 2025 * can be prepended, and the inner header, if present, 2026 * should be either GRE or UDP/GUE. 2027 * 2028 * BPF_LWT_ENCAP_SEG6*** types can be called by bpf programs of 2029 * type BPF_PROG_TYPE_LWT_IN; BPF_LWT_ENCAP_IP type can be called 2030 * by bpf programs of types BPF_PROG_TYPE_LWT_IN and 2031 * BPF_PROG_TYPE_LWT_XMIT. 2032 * 2033 * A call to this helper is susceptible to change the underlaying 2034 * packet buffer. Therefore, at load time, all checks on pointers 2035 * previously done by the verifier are invalidated and must be 2036 * performed again, if the helper is used in combination with 2037 * direct packet access. 2038 * Return 2039 * 0 on success, or a negative error in case of failure. 2040 * 2041 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2042 * Description 2043 * Store *len* bytes from address *from* into the packet 2044 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2045 * inside the outermost IPv6 Segment Routing Header can be 2046 * modified through this helper. 2047 * 2048 * A call to this helper is susceptible to change the underlaying 2049 * packet buffer. Therefore, at load time, all checks on pointers 2050 * previously done by the verifier are invalidated and must be 2051 * performed again, if the helper is used in combination with 2052 * direct packet access. 2053 * Return 2054 * 0 on success, or a negative error in case of failure. 2055 * 2056 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2057 * Description 2058 * Adjust the size allocated to TLVs in the outermost IPv6 2059 * Segment Routing Header contained in the packet associated to 2060 * *skb*, at position *offset* by *delta* bytes. Only offsets 2061 * after the segments are accepted. *delta* can be as well 2062 * positive (growing) as negative (shrinking). 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_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2073 * Description 2074 * Apply an IPv6 Segment Routing action of type *action* to the 2075 * packet associated to *skb*. Each action takes a parameter 2076 * contained at address *param*, and of length *param_len* bytes. 2077 * *action* can be one of: 2078 * 2079 * **SEG6_LOCAL_ACTION_END_X** 2080 * End.X action: Endpoint with Layer-3 cross-connect. 2081 * Type of *param*: **struct in6_addr**. 2082 * **SEG6_LOCAL_ACTION_END_T** 2083 * End.T action: Endpoint with specific IPv6 table lookup. 2084 * Type of *param*: **int**. 2085 * **SEG6_LOCAL_ACTION_END_B6** 2086 * End.B6 action: Endpoint bound to an SRv6 policy. 2087 * Type of param: **struct ipv6_sr_hdr**. 2088 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2089 * End.B6.Encap action: Endpoint bound to an SRv6 2090 * encapsulation policy. 2091 * Type of param: **struct ipv6_sr_hdr**. 2092 * 2093 * A call to this helper is susceptible to change the underlaying 2094 * packet buffer. Therefore, at load time, all checks on pointers 2095 * previously done by the verifier are invalidated and must be 2096 * performed again, if the helper is used in combination with 2097 * direct packet access. 2098 * Return 2099 * 0 on success, or a negative error in case of failure. 2100 * 2101 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2102 * Description 2103 * This helper is used in programs implementing IR decoding, to 2104 * report a successfully decoded key press with *scancode*, 2105 * *toggle* value in the given *protocol*. The scancode will be 2106 * translated to a keycode using the rc keymap, and reported as 2107 * an input key down event. After a period a key up event is 2108 * generated. This period can be extended by calling either 2109 * **bpf_rc_keydown**\ () again with the same values, or calling 2110 * **bpf_rc_repeat**\ (). 2111 * 2112 * Some protocols include a toggle bit, in case the button was 2113 * released and pressed again between consecutive scancodes. 2114 * 2115 * The *ctx* should point to the lirc sample as passed into 2116 * the program. 2117 * 2118 * The *protocol* is the decoded protocol number (see 2119 * **enum rc_proto** for some predefined values). 2120 * 2121 * This helper is only available is the kernel was compiled with 2122 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2123 * "**y**". 2124 * Return 2125 * 0 2126 * 2127 * int bpf_rc_repeat(void *ctx) 2128 * Description 2129 * This helper is used in programs implementing IR decoding, to 2130 * report a successfully decoded repeat key message. This delays 2131 * the generation of a key up event for previously generated 2132 * key down event. 2133 * 2134 * Some IR protocols like NEC have a special IR message for 2135 * repeating last button, for when a button is held down. 2136 * 2137 * The *ctx* should point to the lirc sample as passed into 2138 * the program. 2139 * 2140 * This helper is only available is the kernel was compiled with 2141 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2142 * "**y**". 2143 * Return 2144 * 0 2145 * 2146 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb) 2147 * Description 2148 * Return the cgroup v2 id of the socket associated with the *skb*. 2149 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2150 * helper for cgroup v1 by providing a tag resp. identifier that 2151 * can be matched on or used for map lookups e.g. to implement 2152 * policy. The cgroup v2 id of a given path in the hierarchy is 2153 * exposed in user space through the f_handle API in order to get 2154 * to the same 64-bit id. 2155 * 2156 * This helper can be used on TC egress path, but not on ingress, 2157 * and is available only if the kernel was compiled with the 2158 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2159 * Return 2160 * The id is returned or 0 in case the id could not be retrieved. 2161 * 2162 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2163 * Description 2164 * Return id of cgroup v2 that is ancestor of cgroup associated 2165 * with the *skb* at the *ancestor_level*. The root cgroup is at 2166 * *ancestor_level* zero and each step down the hierarchy 2167 * increments the level. If *ancestor_level* == level of cgroup 2168 * associated with *skb*, then return value will be same as that 2169 * of **bpf_skb_cgroup_id**\ (). 2170 * 2171 * The helper is useful to implement policies based on cgroups 2172 * that are upper in hierarchy than immediate cgroup associated 2173 * with *skb*. 2174 * 2175 * The format of returned id and helper limitations are same as in 2176 * **bpf_skb_cgroup_id**\ (). 2177 * Return 2178 * The id is returned or 0 in case the id could not be retrieved. 2179 * 2180 * u64 bpf_get_current_cgroup_id(void) 2181 * Return 2182 * A 64-bit integer containing the current cgroup id based 2183 * on the cgroup within which the current task is running. 2184 * 2185 * void* get_local_storage(void *map, u64 flags) 2186 * Description 2187 * Get the pointer to the local storage area. 2188 * The type and the size of the local storage is defined 2189 * by the *map* argument. 2190 * The *flags* meaning is specific for each map type, 2191 * and has to be 0 for cgroup local storage. 2192 * 2193 * Depending on the BPF program type, a local storage area 2194 * can be shared between multiple instances of the BPF program, 2195 * running simultaneously. 2196 * 2197 * A user should care about the synchronization by himself. 2198 * For example, by using the **BPF_STX_XADD** instruction to alter 2199 * the shared data. 2200 * Return 2201 * A pointer to the local storage area. 2202 * 2203 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2204 * Description 2205 * Select a **SO_REUSEPORT** socket from a 2206 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2207 * It checks the selected socket is matching the incoming 2208 * request in the socket buffer. 2209 * Return 2210 * 0 on success, or a negative error in case of failure. 2211 * 2212 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2213 * Description 2214 * Look for TCP socket matching *tuple*, optionally in a child 2215 * network namespace *netns*. The return value must be checked, 2216 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2217 * 2218 * The *ctx* should point to the context of the program, such as 2219 * the skb or socket (depending on the hook in use). This is used 2220 * to determine the base network namespace for the lookup. 2221 * 2222 * *tuple_size* must be one of: 2223 * 2224 * **sizeof**\ (*tuple*\ **->ipv4**) 2225 * Look for an IPv4 socket. 2226 * **sizeof**\ (*tuple*\ **->ipv6**) 2227 * Look for an IPv6 socket. 2228 * 2229 * If the *netns* is a negative signed 32-bit integer, then the 2230 * socket lookup table in the netns associated with the *ctx* will 2231 * will be used. For the TC hooks, this is the netns of the device 2232 * in the skb. For socket hooks, this is the netns of the socket. 2233 * If *netns* is any other signed 32-bit value greater than or 2234 * equal to zero then it specifies the ID of the netns relative to 2235 * the netns associated with the *ctx*. *netns* values beyond the 2236 * range of 32-bit integers are reserved for future use. 2237 * 2238 * All values for *flags* are reserved for future usage, and must 2239 * be left at zero. 2240 * 2241 * This helper is available only if the kernel was compiled with 2242 * **CONFIG_NET** configuration option. 2243 * Return 2244 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2245 * For sockets with reuseport option, the **struct bpf_sock** 2246 * result is from **reuse->socks**\ [] using the hash of the tuple. 2247 * 2248 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2249 * Description 2250 * Look for UDP socket matching *tuple*, optionally in a child 2251 * network namespace *netns*. The return value must be checked, 2252 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2253 * 2254 * The *ctx* should point to the context of the program, such as 2255 * the skb or socket (depending on the hook in use). This is used 2256 * to determine the base network namespace for the lookup. 2257 * 2258 * *tuple_size* must be one of: 2259 * 2260 * **sizeof**\ (*tuple*\ **->ipv4**) 2261 * Look for an IPv4 socket. 2262 * **sizeof**\ (*tuple*\ **->ipv6**) 2263 * Look for an IPv6 socket. 2264 * 2265 * If the *netns* is a negative signed 32-bit integer, then the 2266 * socket lookup table in the netns associated with the *ctx* will 2267 * will be used. For the TC hooks, this is the netns of the device 2268 * in the skb. For socket hooks, this is the netns of the socket. 2269 * If *netns* is any other signed 32-bit value greater than or 2270 * equal to zero then it specifies the ID of the netns relative to 2271 * the netns associated with the *ctx*. *netns* values beyond the 2272 * range of 32-bit integers are reserved for future use. 2273 * 2274 * All values for *flags* are reserved for future usage, and must 2275 * be left at zero. 2276 * 2277 * This helper is available only if the kernel was compiled with 2278 * **CONFIG_NET** configuration option. 2279 * Return 2280 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2281 * For sockets with reuseport option, the **struct bpf_sock** 2282 * result is from **reuse->socks**\ [] using the hash of the tuple. 2283 * 2284 * int bpf_sk_release(struct bpf_sock *sock) 2285 * Description 2286 * Release the reference held by *sock*. *sock* must be a 2287 * non-**NULL** pointer that was returned from 2288 * **bpf_sk_lookup_xxx**\ (). 2289 * Return 2290 * 0 on success, or a negative error in case of failure. 2291 * 2292 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2293 * Description 2294 * Pop an element from *map*. 2295 * Return 2296 * 0 on success, or a negative error in case of failure. 2297 * 2298 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2299 * Description 2300 * Get an element from *map* without removing it. 2301 * Return 2302 * 0 on success, or a negative error in case of failure. 2303 * 2304 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 2305 * Description 2306 * For socket policies, insert *len* bytes into *msg* at offset 2307 * *start*. 2308 * 2309 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2310 * *msg* it may want to insert metadata or options into the *msg*. 2311 * This can later be read and used by any of the lower layer BPF 2312 * hooks. 2313 * 2314 * This helper may fail if under memory pressure (a malloc 2315 * fails) in these cases BPF programs will get an appropriate 2316 * error and BPF programs will need to handle them. 2317 * Return 2318 * 0 on success, or a negative error in case of failure. 2319 * 2320 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags) 2321 * Description 2322 * Will remove *pop* bytes from a *msg* starting at byte *start*. 2323 * This may result in **ENOMEM** errors under certain situations if 2324 * an allocation and copy are required due to a full ring buffer. 2325 * However, the helper will try to avoid doing the allocation 2326 * if possible. Other errors can occur if input parameters are 2327 * invalid either due to *start* byte not being valid part of *msg* 2328 * payload and/or *pop* value being to large. 2329 * Return 2330 * 0 on success, or a negative error in case of failure. 2331 * 2332 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2333 * Description 2334 * This helper is used in programs implementing IR decoding, to 2335 * report a successfully decoded pointer movement. 2336 * 2337 * The *ctx* should point to the lirc sample as passed into 2338 * the program. 2339 * 2340 * This helper is only available is the kernel was compiled with 2341 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2342 * "**y**". 2343 * Return 2344 * 0 2345 * 2346 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2347 * Description 2348 * This helper gets a **struct bpf_sock** pointer such 2349 * that all the fields in bpf_sock can be accessed. 2350 * Return 2351 * A **struct bpf_sock** pointer on success, or NULL in 2352 * case of failure. 2353 * 2354 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2355 * Description 2356 * This helper gets a **struct bpf_tcp_sock** pointer from a 2357 * **struct bpf_sock** pointer. 2358 * 2359 * Return 2360 * A **struct bpf_tcp_sock** pointer on success, or NULL in 2361 * case of failure. 2362 * 2363 * int bpf_skb_ecn_set_ce(struct sk_buf *skb) 2364 * Description 2365 * Sets ECN of IP header to ce (congestion encountered) if 2366 * current value is ect (ECN capable). Works with IPv6 and IPv4. 2367 * Return 2368 * 1 if set, 0 if not set. 2369 */ 2370 #define __BPF_FUNC_MAPPER(FN) \ 2371 FN(unspec), \ 2372 FN(map_lookup_elem), \ 2373 FN(map_update_elem), \ 2374 FN(map_delete_elem), \ 2375 FN(probe_read), \ 2376 FN(ktime_get_ns), \ 2377 FN(trace_printk), \ 2378 FN(get_prandom_u32), \ 2379 FN(get_smp_processor_id), \ 2380 FN(skb_store_bytes), \ 2381 FN(l3_csum_replace), \ 2382 FN(l4_csum_replace), \ 2383 FN(tail_call), \ 2384 FN(clone_redirect), \ 2385 FN(get_current_pid_tgid), \ 2386 FN(get_current_uid_gid), \ 2387 FN(get_current_comm), \ 2388 FN(get_cgroup_classid), \ 2389 FN(skb_vlan_push), \ 2390 FN(skb_vlan_pop), \ 2391 FN(skb_get_tunnel_key), \ 2392 FN(skb_set_tunnel_key), \ 2393 FN(perf_event_read), \ 2394 FN(redirect), \ 2395 FN(get_route_realm), \ 2396 FN(perf_event_output), \ 2397 FN(skb_load_bytes), \ 2398 FN(get_stackid), \ 2399 FN(csum_diff), \ 2400 FN(skb_get_tunnel_opt), \ 2401 FN(skb_set_tunnel_opt), \ 2402 FN(skb_change_proto), \ 2403 FN(skb_change_type), \ 2404 FN(skb_under_cgroup), \ 2405 FN(get_hash_recalc), \ 2406 FN(get_current_task), \ 2407 FN(probe_write_user), \ 2408 FN(current_task_under_cgroup), \ 2409 FN(skb_change_tail), \ 2410 FN(skb_pull_data), \ 2411 FN(csum_update), \ 2412 FN(set_hash_invalid), \ 2413 FN(get_numa_node_id), \ 2414 FN(skb_change_head), \ 2415 FN(xdp_adjust_head), \ 2416 FN(probe_read_str), \ 2417 FN(get_socket_cookie), \ 2418 FN(get_socket_uid), \ 2419 FN(set_hash), \ 2420 FN(setsockopt), \ 2421 FN(skb_adjust_room), \ 2422 FN(redirect_map), \ 2423 FN(sk_redirect_map), \ 2424 FN(sock_map_update), \ 2425 FN(xdp_adjust_meta), \ 2426 FN(perf_event_read_value), \ 2427 FN(perf_prog_read_value), \ 2428 FN(getsockopt), \ 2429 FN(override_return), \ 2430 FN(sock_ops_cb_flags_set), \ 2431 FN(msg_redirect_map), \ 2432 FN(msg_apply_bytes), \ 2433 FN(msg_cork_bytes), \ 2434 FN(msg_pull_data), \ 2435 FN(bind), \ 2436 FN(xdp_adjust_tail), \ 2437 FN(skb_get_xfrm_state), \ 2438 FN(get_stack), \ 2439 FN(skb_load_bytes_relative), \ 2440 FN(fib_lookup), \ 2441 FN(sock_hash_update), \ 2442 FN(msg_redirect_hash), \ 2443 FN(sk_redirect_hash), \ 2444 FN(lwt_push_encap), \ 2445 FN(lwt_seg6_store_bytes), \ 2446 FN(lwt_seg6_adjust_srh), \ 2447 FN(lwt_seg6_action), \ 2448 FN(rc_repeat), \ 2449 FN(rc_keydown), \ 2450 FN(skb_cgroup_id), \ 2451 FN(get_current_cgroup_id), \ 2452 FN(get_local_storage), \ 2453 FN(sk_select_reuseport), \ 2454 FN(skb_ancestor_cgroup_id), \ 2455 FN(sk_lookup_tcp), \ 2456 FN(sk_lookup_udp), \ 2457 FN(sk_release), \ 2458 FN(map_push_elem), \ 2459 FN(map_pop_elem), \ 2460 FN(map_peek_elem), \ 2461 FN(msg_push_data), \ 2462 FN(msg_pop_data), \ 2463 FN(rc_pointer_rel), \ 2464 FN(spin_lock), \ 2465 FN(spin_unlock), \ 2466 FN(sk_fullsock), \ 2467 FN(tcp_sock), \ 2468 FN(skb_ecn_set_ce), 2469 2470 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2471 * function eBPF program intends to call 2472 */ 2473 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2474 enum bpf_func_id { 2475 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2476 __BPF_FUNC_MAX_ID, 2477 }; 2478 #undef __BPF_ENUM_FN 2479 2480 /* All flags used by eBPF helper functions, placed here. */ 2481 2482 /* BPF_FUNC_skb_store_bytes flags. */ 2483 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2484 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2485 2486 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2487 * First 4 bits are for passing the header field size. 2488 */ 2489 #define BPF_F_HDR_FIELD_MASK 0xfULL 2490 2491 /* BPF_FUNC_l4_csum_replace flags. */ 2492 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2493 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2494 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2495 2496 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2497 #define BPF_F_INGRESS (1ULL << 0) 2498 2499 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2500 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2501 2502 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2503 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2504 #define BPF_F_USER_STACK (1ULL << 8) 2505 /* flags used by BPF_FUNC_get_stackid only. */ 2506 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2507 #define BPF_F_REUSE_STACKID (1ULL << 10) 2508 /* flags used by BPF_FUNC_get_stack only. */ 2509 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2510 2511 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2512 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2513 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2514 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2515 2516 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2517 * BPF_FUNC_perf_event_read_value flags. 2518 */ 2519 #define BPF_F_INDEX_MASK 0xffffffffULL 2520 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2521 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2522 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2523 2524 /* Current network namespace */ 2525 #define BPF_F_CURRENT_NETNS (-1L) 2526 2527 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 2528 enum bpf_adj_room_mode { 2529 BPF_ADJ_ROOM_NET, 2530 }; 2531 2532 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 2533 enum bpf_hdr_start_off { 2534 BPF_HDR_START_MAC, 2535 BPF_HDR_START_NET, 2536 }; 2537 2538 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2539 enum bpf_lwt_encap_mode { 2540 BPF_LWT_ENCAP_SEG6, 2541 BPF_LWT_ENCAP_SEG6_INLINE, 2542 BPF_LWT_ENCAP_IP, 2543 }; 2544 2545 #define __bpf_md_ptr(type, name) \ 2546 union { \ 2547 type name; \ 2548 __u64 :64; \ 2549 } __attribute__((aligned(8))) 2550 2551 /* user accessible mirror of in-kernel sk_buff. 2552 * new fields can only be added to the end of this structure 2553 */ 2554 struct __sk_buff { 2555 __u32 len; 2556 __u32 pkt_type; 2557 __u32 mark; 2558 __u32 queue_mapping; 2559 __u32 protocol; 2560 __u32 vlan_present; 2561 __u32 vlan_tci; 2562 __u32 vlan_proto; 2563 __u32 priority; 2564 __u32 ingress_ifindex; 2565 __u32 ifindex; 2566 __u32 tc_index; 2567 __u32 cb[5]; 2568 __u32 hash; 2569 __u32 tc_classid; 2570 __u32 data; 2571 __u32 data_end; 2572 __u32 napi_id; 2573 2574 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 2575 __u32 family; 2576 __u32 remote_ip4; /* Stored in network byte order */ 2577 __u32 local_ip4; /* Stored in network byte order */ 2578 __u32 remote_ip6[4]; /* Stored in network byte order */ 2579 __u32 local_ip6[4]; /* Stored in network byte order */ 2580 __u32 remote_port; /* Stored in network byte order */ 2581 __u32 local_port; /* stored in host byte order */ 2582 /* ... here. */ 2583 2584 __u32 data_meta; 2585 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 2586 __u64 tstamp; 2587 __u32 wire_len; 2588 __u32 gso_segs; 2589 __bpf_md_ptr(struct bpf_sock *, sk); 2590 }; 2591 2592 struct bpf_tunnel_key { 2593 __u32 tunnel_id; 2594 union { 2595 __u32 remote_ipv4; 2596 __u32 remote_ipv6[4]; 2597 }; 2598 __u8 tunnel_tos; 2599 __u8 tunnel_ttl; 2600 __u16 tunnel_ext; /* Padding, future use. */ 2601 __u32 tunnel_label; 2602 }; 2603 2604 /* user accessible mirror of in-kernel xfrm_state. 2605 * new fields can only be added to the end of this structure 2606 */ 2607 struct bpf_xfrm_state { 2608 __u32 reqid; 2609 __u32 spi; /* Stored in network byte order */ 2610 __u16 family; 2611 __u16 ext; /* Padding, future use. */ 2612 union { 2613 __u32 remote_ipv4; /* Stored in network byte order */ 2614 __u32 remote_ipv6[4]; /* Stored in network byte order */ 2615 }; 2616 }; 2617 2618 /* Generic BPF return codes which all BPF program types may support. 2619 * The values are binary compatible with their TC_ACT_* counter-part to 2620 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 2621 * programs. 2622 * 2623 * XDP is handled seprately, see XDP_*. 2624 */ 2625 enum bpf_ret_code { 2626 BPF_OK = 0, 2627 /* 1 reserved */ 2628 BPF_DROP = 2, 2629 /* 3-6 reserved */ 2630 BPF_REDIRECT = 7, 2631 /* >127 are reserved for prog type specific return codes. 2632 * 2633 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 2634 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 2635 * changed and should be routed based on its new L3 header. 2636 * (This is an L3 redirect, as opposed to L2 redirect 2637 * represented by BPF_REDIRECT above). 2638 */ 2639 BPF_LWT_REROUTE = 128, 2640 }; 2641 2642 struct bpf_sock { 2643 __u32 bound_dev_if; 2644 __u32 family; 2645 __u32 type; 2646 __u32 protocol; 2647 __u32 mark; 2648 __u32 priority; 2649 /* IP address also allows 1 and 2 bytes access */ 2650 __u32 src_ip4; 2651 __u32 src_ip6[4]; 2652 __u32 src_port; /* host byte order */ 2653 __u32 dst_port; /* network byte order */ 2654 __u32 dst_ip4; 2655 __u32 dst_ip6[4]; 2656 __u32 state; 2657 }; 2658 2659 struct bpf_tcp_sock { 2660 __u32 snd_cwnd; /* Sending congestion window */ 2661 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 2662 __u32 rtt_min; 2663 __u32 snd_ssthresh; /* Slow start size threshold */ 2664 __u32 rcv_nxt; /* What we want to receive next */ 2665 __u32 snd_nxt; /* Next sequence we send */ 2666 __u32 snd_una; /* First byte we want an ack for */ 2667 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 2668 __u32 ecn_flags; /* ECN status bits. */ 2669 __u32 rate_delivered; /* saved rate sample: packets delivered */ 2670 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 2671 __u32 packets_out; /* Packets which are "in flight" */ 2672 __u32 retrans_out; /* Retransmitted packets out */ 2673 __u32 total_retrans; /* Total retransmits for entire connection */ 2674 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 2675 * total number of segments in. 2676 */ 2677 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 2678 * total number of data segments in. 2679 */ 2680 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 2681 * The total number of segments sent. 2682 */ 2683 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 2684 * total number of data segments sent. 2685 */ 2686 __u32 lost_out; /* Lost packets */ 2687 __u32 sacked_out; /* SACK'd packets */ 2688 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 2689 * sum(delta(rcv_nxt)), or how many bytes 2690 * were acked. 2691 */ 2692 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 2693 * sum(delta(snd_una)), or how many bytes 2694 * were acked. 2695 */ 2696 }; 2697 2698 struct bpf_sock_tuple { 2699 union { 2700 struct { 2701 __be32 saddr; 2702 __be32 daddr; 2703 __be16 sport; 2704 __be16 dport; 2705 } ipv4; 2706 struct { 2707 __be32 saddr[4]; 2708 __be32 daddr[4]; 2709 __be16 sport; 2710 __be16 dport; 2711 } ipv6; 2712 }; 2713 }; 2714 2715 #define XDP_PACKET_HEADROOM 256 2716 2717 /* User return codes for XDP prog type. 2718 * A valid XDP program must return one of these defined values. All other 2719 * return codes are reserved for future use. Unknown return codes will 2720 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 2721 */ 2722 enum xdp_action { 2723 XDP_ABORTED = 0, 2724 XDP_DROP, 2725 XDP_PASS, 2726 XDP_TX, 2727 XDP_REDIRECT, 2728 }; 2729 2730 /* user accessible metadata for XDP packet hook 2731 * new fields must be added to the end of this structure 2732 */ 2733 struct xdp_md { 2734 __u32 data; 2735 __u32 data_end; 2736 __u32 data_meta; 2737 /* Below access go through struct xdp_rxq_info */ 2738 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 2739 __u32 rx_queue_index; /* rxq->queue_index */ 2740 }; 2741 2742 enum sk_action { 2743 SK_DROP = 0, 2744 SK_PASS, 2745 }; 2746 2747 /* user accessible metadata for SK_MSG packet hook, new fields must 2748 * be added to the end of this structure 2749 */ 2750 struct sk_msg_md { 2751 __bpf_md_ptr(void *, data); 2752 __bpf_md_ptr(void *, data_end); 2753 2754 __u32 family; 2755 __u32 remote_ip4; /* Stored in network byte order */ 2756 __u32 local_ip4; /* Stored in network byte order */ 2757 __u32 remote_ip6[4]; /* Stored in network byte order */ 2758 __u32 local_ip6[4]; /* Stored in network byte order */ 2759 __u32 remote_port; /* Stored in network byte order */ 2760 __u32 local_port; /* stored in host byte order */ 2761 __u32 size; /* Total size of sk_msg */ 2762 }; 2763 2764 struct sk_reuseport_md { 2765 /* 2766 * Start of directly accessible data. It begins from 2767 * the tcp/udp header. 2768 */ 2769 __bpf_md_ptr(void *, data); 2770 /* End of directly accessible data */ 2771 __bpf_md_ptr(void *, data_end); 2772 /* 2773 * Total length of packet (starting from the tcp/udp header). 2774 * Note that the directly accessible bytes (data_end - data) 2775 * could be less than this "len". Those bytes could be 2776 * indirectly read by a helper "bpf_skb_load_bytes()". 2777 */ 2778 __u32 len; 2779 /* 2780 * Eth protocol in the mac header (network byte order). e.g. 2781 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 2782 */ 2783 __u32 eth_protocol; 2784 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 2785 __u32 bind_inany; /* Is sock bound to an INANY address? */ 2786 __u32 hash; /* A hash of the packet 4 tuples */ 2787 }; 2788 2789 #define BPF_TAG_SIZE 8 2790 2791 struct bpf_prog_info { 2792 __u32 type; 2793 __u32 id; 2794 __u8 tag[BPF_TAG_SIZE]; 2795 __u32 jited_prog_len; 2796 __u32 xlated_prog_len; 2797 __aligned_u64 jited_prog_insns; 2798 __aligned_u64 xlated_prog_insns; 2799 __u64 load_time; /* ns since boottime */ 2800 __u32 created_by_uid; 2801 __u32 nr_map_ids; 2802 __aligned_u64 map_ids; 2803 char name[BPF_OBJ_NAME_LEN]; 2804 __u32 ifindex; 2805 __u32 gpl_compatible:1; 2806 __u64 netns_dev; 2807 __u64 netns_ino; 2808 __u32 nr_jited_ksyms; 2809 __u32 nr_jited_func_lens; 2810 __aligned_u64 jited_ksyms; 2811 __aligned_u64 jited_func_lens; 2812 __u32 btf_id; 2813 __u32 func_info_rec_size; 2814 __aligned_u64 func_info; 2815 __u32 nr_func_info; 2816 __u32 nr_line_info; 2817 __aligned_u64 line_info; 2818 __aligned_u64 jited_line_info; 2819 __u32 nr_jited_line_info; 2820 __u32 line_info_rec_size; 2821 __u32 jited_line_info_rec_size; 2822 __u32 nr_prog_tags; 2823 __aligned_u64 prog_tags; 2824 __u64 run_time_ns; 2825 __u64 run_cnt; 2826 } __attribute__((aligned(8))); 2827 2828 struct bpf_map_info { 2829 __u32 type; 2830 __u32 id; 2831 __u32 key_size; 2832 __u32 value_size; 2833 __u32 max_entries; 2834 __u32 map_flags; 2835 char name[BPF_OBJ_NAME_LEN]; 2836 __u32 ifindex; 2837 __u32 :32; 2838 __u64 netns_dev; 2839 __u64 netns_ino; 2840 __u32 btf_id; 2841 __u32 btf_key_type_id; 2842 __u32 btf_value_type_id; 2843 } __attribute__((aligned(8))); 2844 2845 struct bpf_btf_info { 2846 __aligned_u64 btf; 2847 __u32 btf_size; 2848 __u32 id; 2849 } __attribute__((aligned(8))); 2850 2851 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 2852 * by user and intended to be used by socket (e.g. to bind to, depends on 2853 * attach attach type). 2854 */ 2855 struct bpf_sock_addr { 2856 __u32 user_family; /* Allows 4-byte read, but no write. */ 2857 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 2858 * Stored in network byte order. 2859 */ 2860 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 2861 * Stored in network byte order. 2862 */ 2863 __u32 user_port; /* Allows 4-byte read and write. 2864 * Stored in network byte order 2865 */ 2866 __u32 family; /* Allows 4-byte read, but no write */ 2867 __u32 type; /* Allows 4-byte read, but no write */ 2868 __u32 protocol; /* Allows 4-byte read, but no write */ 2869 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write. 2870 * Stored in network byte order. 2871 */ 2872 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 2873 * Stored in network byte order. 2874 */ 2875 }; 2876 2877 /* User bpf_sock_ops struct to access socket values and specify request ops 2878 * and their replies. 2879 * Some of this fields are in network (bigendian) byte order and may need 2880 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 2881 * New fields can only be added at the end of this structure 2882 */ 2883 struct bpf_sock_ops { 2884 __u32 op; 2885 union { 2886 __u32 args[4]; /* Optionally passed to bpf program */ 2887 __u32 reply; /* Returned by bpf program */ 2888 __u32 replylong[4]; /* Optionally returned by bpf prog */ 2889 }; 2890 __u32 family; 2891 __u32 remote_ip4; /* Stored in network byte order */ 2892 __u32 local_ip4; /* Stored in network byte order */ 2893 __u32 remote_ip6[4]; /* Stored in network byte order */ 2894 __u32 local_ip6[4]; /* Stored in network byte order */ 2895 __u32 remote_port; /* Stored in network byte order */ 2896 __u32 local_port; /* stored in host byte order */ 2897 __u32 is_fullsock; /* Some TCP fields are only valid if 2898 * there is a full socket. If not, the 2899 * fields read as zero. 2900 */ 2901 __u32 snd_cwnd; 2902 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 2903 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 2904 __u32 state; 2905 __u32 rtt_min; 2906 __u32 snd_ssthresh; 2907 __u32 rcv_nxt; 2908 __u32 snd_nxt; 2909 __u32 snd_una; 2910 __u32 mss_cache; 2911 __u32 ecn_flags; 2912 __u32 rate_delivered; 2913 __u32 rate_interval_us; 2914 __u32 packets_out; 2915 __u32 retrans_out; 2916 __u32 total_retrans; 2917 __u32 segs_in; 2918 __u32 data_segs_in; 2919 __u32 segs_out; 2920 __u32 data_segs_out; 2921 __u32 lost_out; 2922 __u32 sacked_out; 2923 __u32 sk_txhash; 2924 __u64 bytes_received; 2925 __u64 bytes_acked; 2926 }; 2927 2928 /* Definitions for bpf_sock_ops_cb_flags */ 2929 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 2930 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 2931 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 2932 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 2933 * supported cb flags 2934 */ 2935 2936 /* List of known BPF sock_ops operators. 2937 * New entries can only be added at the end 2938 */ 2939 enum { 2940 BPF_SOCK_OPS_VOID, 2941 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 2942 * -1 if default value should be used 2943 */ 2944 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 2945 * window (in packets) or -1 if default 2946 * value should be used 2947 */ 2948 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 2949 * active connection is initialized 2950 */ 2951 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 2952 * active connection is 2953 * established 2954 */ 2955 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 2956 * passive connection is 2957 * established 2958 */ 2959 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 2960 * needs ECN 2961 */ 2962 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 2963 * based on the path and may be 2964 * dependent on the congestion control 2965 * algorithm. In general it indicates 2966 * a congestion threshold. RTTs above 2967 * this indicate congestion 2968 */ 2969 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 2970 * Arg1: value of icsk_retransmits 2971 * Arg2: value of icsk_rto 2972 * Arg3: whether RTO has expired 2973 */ 2974 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 2975 * Arg1: sequence number of 1st byte 2976 * Arg2: # segments 2977 * Arg3: return value of 2978 * tcp_transmit_skb (0 => success) 2979 */ 2980 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 2981 * Arg1: old_state 2982 * Arg2: new_state 2983 */ 2984 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 2985 * socket transition to LISTEN state. 2986 */ 2987 }; 2988 2989 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 2990 * changes between the TCP and BPF versions. Ideally this should never happen. 2991 * If it does, we need to add code to convert them before calling 2992 * the BPF sock_ops function. 2993 */ 2994 enum { 2995 BPF_TCP_ESTABLISHED = 1, 2996 BPF_TCP_SYN_SENT, 2997 BPF_TCP_SYN_RECV, 2998 BPF_TCP_FIN_WAIT1, 2999 BPF_TCP_FIN_WAIT2, 3000 BPF_TCP_TIME_WAIT, 3001 BPF_TCP_CLOSE, 3002 BPF_TCP_CLOSE_WAIT, 3003 BPF_TCP_LAST_ACK, 3004 BPF_TCP_LISTEN, 3005 BPF_TCP_CLOSING, /* Now a valid state */ 3006 BPF_TCP_NEW_SYN_RECV, 3007 3008 BPF_TCP_MAX_STATES /* Leave at the end! */ 3009 }; 3010 3011 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3012 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3013 3014 struct bpf_perf_event_value { 3015 __u64 counter; 3016 __u64 enabled; 3017 __u64 running; 3018 }; 3019 3020 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3021 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3022 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3023 3024 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3025 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3026 3027 struct bpf_cgroup_dev_ctx { 3028 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3029 __u32 access_type; 3030 __u32 major; 3031 __u32 minor; 3032 }; 3033 3034 struct bpf_raw_tracepoint_args { 3035 __u64 args[0]; 3036 }; 3037 3038 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3039 * OUTPUT: Do lookup from egress perspective; default is ingress 3040 */ 3041 #define BPF_FIB_LOOKUP_DIRECT BIT(0) 3042 #define BPF_FIB_LOOKUP_OUTPUT BIT(1) 3043 3044 enum { 3045 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3046 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3047 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3048 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3049 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3050 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3051 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3052 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3053 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3054 }; 3055 3056 struct bpf_fib_lookup { 3057 /* input: network family for lookup (AF_INET, AF_INET6) 3058 * output: network family of egress nexthop 3059 */ 3060 __u8 family; 3061 3062 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3063 __u8 l4_protocol; 3064 __be16 sport; 3065 __be16 dport; 3066 3067 /* total length of packet from network header - used for MTU check */ 3068 __u16 tot_len; 3069 3070 /* input: L3 device index for lookup 3071 * output: device index from FIB lookup 3072 */ 3073 __u32 ifindex; 3074 3075 union { 3076 /* inputs to lookup */ 3077 __u8 tos; /* AF_INET */ 3078 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3079 3080 /* output: metric of fib result (IPv4/IPv6 only) */ 3081 __u32 rt_metric; 3082 }; 3083 3084 union { 3085 __be32 ipv4_src; 3086 __u32 ipv6_src[4]; /* in6_addr; network order */ 3087 }; 3088 3089 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3090 * network header. output: bpf_fib_lookup sets to gateway address 3091 * if FIB lookup returns gateway route 3092 */ 3093 union { 3094 __be32 ipv4_dst; 3095 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3096 }; 3097 3098 /* output */ 3099 __be16 h_vlan_proto; 3100 __be16 h_vlan_TCI; 3101 __u8 smac[6]; /* ETH_ALEN */ 3102 __u8 dmac[6]; /* ETH_ALEN */ 3103 }; 3104 3105 enum bpf_task_fd_type { 3106 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3107 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3108 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3109 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3110 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3111 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3112 }; 3113 3114 struct bpf_flow_keys { 3115 __u16 nhoff; 3116 __u16 thoff; 3117 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3118 __u8 is_frag; 3119 __u8 is_first_frag; 3120 __u8 is_encap; 3121 __u8 ip_proto; 3122 __be16 n_proto; 3123 __be16 sport; 3124 __be16 dport; 3125 union { 3126 struct { 3127 __be32 ipv4_src; 3128 __be32 ipv4_dst; 3129 }; 3130 struct { 3131 __u32 ipv6_src[4]; /* in6_addr; network order */ 3132 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3133 }; 3134 }; 3135 }; 3136 3137 struct bpf_func_info { 3138 __u32 insn_off; 3139 __u32 type_id; 3140 }; 3141 3142 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3143 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3144 3145 struct bpf_line_info { 3146 __u32 insn_off; 3147 __u32 file_name_off; 3148 __u32 line_off; 3149 __u32 line_col; 3150 }; 3151 3152 struct bpf_spin_lock { 3153 __u32 val; 3154 }; 3155 #endif /* _UAPI__LINUX_BPF_H__ */ 3156