1 R"********( 2 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 3 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 4 * 5 * This program is free software; you can redistribute it and/or 6 * modify it under the terms of version 2 of the GNU General Public 7 * License as published by the Free Software Foundation. 8 */ 9 #ifndef _UAPI__LINUX_BPF_H__ 10 #define _UAPI__LINUX_BPF_H__ 11 12 #include <linux/types.h> 13 #include <linux/bpf_common.h> 14 15 /* Extended instruction set based on top of classic BPF */ 16 17 /* instruction classes */ 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 enum bpf_prog_type { 138 BPF_PROG_TYPE_UNSPEC, 139 BPF_PROG_TYPE_SOCKET_FILTER, 140 BPF_PROG_TYPE_KPROBE, 141 BPF_PROG_TYPE_SCHED_CLS, 142 BPF_PROG_TYPE_SCHED_ACT, 143 BPF_PROG_TYPE_TRACEPOINT, 144 BPF_PROG_TYPE_XDP, 145 BPF_PROG_TYPE_PERF_EVENT, 146 BPF_PROG_TYPE_CGROUP_SKB, 147 BPF_PROG_TYPE_CGROUP_SOCK, 148 BPF_PROG_TYPE_LWT_IN, 149 BPF_PROG_TYPE_LWT_OUT, 150 BPF_PROG_TYPE_LWT_XMIT, 151 BPF_PROG_TYPE_SOCK_OPS, 152 BPF_PROG_TYPE_SK_SKB, 153 BPF_PROG_TYPE_CGROUP_DEVICE, 154 BPF_PROG_TYPE_SK_MSG, 155 BPF_PROG_TYPE_RAW_TRACEPOINT, 156 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 157 BPF_PROG_TYPE_LWT_SEG6LOCAL, 158 BPF_PROG_TYPE_LIRC_MODE2, 159 BPF_PROG_TYPE_SK_REUSEPORT, 160 BPF_PROG_TYPE_FLOW_DISSECTOR, 161 }; 162 163 enum bpf_attach_type { 164 BPF_CGROUP_INET_INGRESS, 165 BPF_CGROUP_INET_EGRESS, 166 BPF_CGROUP_INET_SOCK_CREATE, 167 BPF_CGROUP_SOCK_OPS, 168 BPF_SK_SKB_STREAM_PARSER, 169 BPF_SK_SKB_STREAM_VERDICT, 170 BPF_CGROUP_DEVICE, 171 BPF_SK_MSG_VERDICT, 172 BPF_CGROUP_INET4_BIND, 173 BPF_CGROUP_INET6_BIND, 174 BPF_CGROUP_INET4_CONNECT, 175 BPF_CGROUP_INET6_CONNECT, 176 BPF_CGROUP_INET4_POST_BIND, 177 BPF_CGROUP_INET6_POST_BIND, 178 BPF_CGROUP_UDP4_SENDMSG, 179 BPF_CGROUP_UDP6_SENDMSG, 180 BPF_LIRC_MODE2, 181 BPF_FLOW_DISSECTOR, 182 __MAX_BPF_ATTACH_TYPE 183 }; 184 185 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 186 187 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 188 * 189 * NONE(default): No further bpf programs allowed in the subtree. 190 * 191 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 192 * the program in this cgroup yields to sub-cgroup program. 193 * 194 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 195 * that cgroup program gets run in addition to the program in this cgroup. 196 * 197 * Only one program is allowed to be attached to a cgroup with 198 * NONE or BPF_F_ALLOW_OVERRIDE flag. 199 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 200 * release old program and attach the new one. Attach flags has to match. 201 * 202 * Multiple programs are allowed to be attached to a cgroup with 203 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 204 * (those that were attached first, run first) 205 * The programs of sub-cgroup are executed first, then programs of 206 * this cgroup and then programs of parent cgroup. 207 * When children program makes decision (like picking TCP CA or sock bind) 208 * parent program has a chance to override it. 209 * 210 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 211 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 212 * Ex1: 213 * cgrp1 (MULTI progs A, B) -> 214 * cgrp2 (OVERRIDE prog C) -> 215 * cgrp3 (MULTI prog D) -> 216 * cgrp4 (OVERRIDE prog E) -> 217 * cgrp5 (NONE prog F) 218 * the event in cgrp5 triggers execution of F,D,A,B in that order. 219 * if prog F is detached, the execution is E,D,A,B 220 * if prog F and D are detached, the execution is E,A,B 221 * if prog F, E and D are detached, the execution is C,A,B 222 * 223 * All eligible programs are executed regardless of return code from 224 * earlier programs. 225 */ 226 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 227 #define BPF_F_ALLOW_MULTI (1U << 1) 228 229 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 230 * verifier will perform strict alignment checking as if the kernel 231 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 232 * and NET_IP_ALIGN defined to 2. 233 */ 234 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 235 236 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */ 237 #define BPF_PSEUDO_MAP_FD 1 238 239 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 240 * offset to another bpf function 241 */ 242 #define BPF_PSEUDO_CALL 1 243 244 /* flags for BPF_MAP_UPDATE_ELEM command */ 245 #define BPF_ANY 0 /* create new element or update existing */ 246 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 247 #define BPF_EXIST 2 /* update existing element */ 248 249 /* flags for BPF_MAP_CREATE command */ 250 #define BPF_F_NO_PREALLOC (1U << 0) 251 /* Instead of having one common LRU list in the 252 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 253 * which can scale and perform better. 254 * Note, the LRU nodes (including free nodes) cannot be moved 255 * across different LRU lists. 256 */ 257 #define BPF_F_NO_COMMON_LRU (1U << 1) 258 /* Specify numa node during map creation */ 259 #define BPF_F_NUMA_NODE (1U << 2) 260 261 /* flags for BPF_PROG_QUERY */ 262 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 263 264 #define BPF_OBJ_NAME_LEN 16U 265 266 /* Flags for accessing BPF object */ 267 #define BPF_F_RDONLY (1U << 3) 268 #define BPF_F_WRONLY (1U << 4) 269 270 /* Flag for stack_map, store build_id+offset instead of pointer */ 271 #define BPF_F_STACK_BUILD_ID (1U << 5) 272 273 enum bpf_stack_build_id_status { 274 /* user space need an empty entry to identify end of a trace */ 275 BPF_STACK_BUILD_ID_EMPTY = 0, 276 /* with valid build_id and offset */ 277 BPF_STACK_BUILD_ID_VALID = 1, 278 /* couldn't get build_id, fallback to ip */ 279 BPF_STACK_BUILD_ID_IP = 2, 280 }; 281 282 #define BPF_BUILD_ID_SIZE 20 283 struct bpf_stack_build_id { 284 __s32 status; 285 unsigned char build_id[BPF_BUILD_ID_SIZE]; 286 union { 287 __u64 offset; 288 __u64 ip; 289 }; 290 }; 291 292 union bpf_attr { 293 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 294 __u32 map_type; /* one of enum bpf_map_type */ 295 __u32 key_size; /* size of key in bytes */ 296 __u32 value_size; /* size of value in bytes */ 297 __u32 max_entries; /* max number of entries in a map */ 298 __u32 map_flags; /* BPF_MAP_CREATE related 299 * flags defined above. 300 */ 301 __u32 inner_map_fd; /* fd pointing to the inner map */ 302 __u32 numa_node; /* numa node (effective only if 303 * BPF_F_NUMA_NODE is set). 304 */ 305 char map_name[BPF_OBJ_NAME_LEN]; 306 __u32 map_ifindex; /* ifindex of netdev to create on */ 307 __u32 btf_fd; /* fd pointing to a BTF type data */ 308 __u32 btf_key_type_id; /* BTF type_id of the key */ 309 __u32 btf_value_type_id; /* BTF type_id of the value */ 310 }; 311 312 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 313 __u32 map_fd; 314 __aligned_u64 key; 315 union { 316 __aligned_u64 value; 317 __aligned_u64 next_key; 318 }; 319 __u64 flags; 320 }; 321 322 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 323 __u32 prog_type; /* one of enum bpf_prog_type */ 324 __u32 insn_cnt; 325 __aligned_u64 insns; 326 __aligned_u64 license; 327 __u32 log_level; /* verbosity level of verifier */ 328 __u32 log_size; /* size of user buffer */ 329 __aligned_u64 log_buf; /* user supplied buffer */ 330 __u32 kern_version; /* checked when prog_type=kprobe */ 331 __u32 prog_flags; 332 char prog_name[BPF_OBJ_NAME_LEN]; 333 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 334 /* For some prog types expected attach type must be known at 335 * load time to verify attach type specific parts of prog 336 * (context accesses, allowed helpers, etc). 337 */ 338 __u32 expected_attach_type; 339 }; 340 341 struct { /* anonymous struct used by BPF_OBJ_* commands */ 342 __aligned_u64 pathname; 343 __u32 bpf_fd; 344 __u32 file_flags; 345 }; 346 347 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 348 __u32 target_fd; /* container object to attach to */ 349 __u32 attach_bpf_fd; /* eBPF program to attach */ 350 __u32 attach_type; 351 __u32 attach_flags; 352 }; 353 354 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 355 __u32 prog_fd; 356 __u32 retval; 357 __u32 data_size_in; 358 __u32 data_size_out; 359 __aligned_u64 data_in; 360 __aligned_u64 data_out; 361 __u32 repeat; 362 __u32 duration; 363 } test; 364 365 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 366 union { 367 __u32 start_id; 368 __u32 prog_id; 369 __u32 map_id; 370 __u32 btf_id; 371 }; 372 __u32 next_id; 373 __u32 open_flags; 374 }; 375 376 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 377 __u32 bpf_fd; 378 __u32 info_len; 379 __aligned_u64 info; 380 } info; 381 382 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 383 __u32 target_fd; /* container object to query */ 384 __u32 attach_type; 385 __u32 query_flags; 386 __u32 attach_flags; 387 __aligned_u64 prog_ids; 388 __u32 prog_cnt; 389 } query; 390 391 struct { 392 __u64 name; 393 __u32 prog_fd; 394 } raw_tracepoint; 395 396 struct { /* anonymous struct for BPF_BTF_LOAD */ 397 __aligned_u64 btf; 398 __aligned_u64 btf_log_buf; 399 __u32 btf_size; 400 __u32 btf_log_size; 401 __u32 btf_log_level; 402 }; 403 404 struct { 405 __u32 pid; /* input: pid */ 406 __u32 fd; /* input: fd */ 407 __u32 flags; /* input: flags */ 408 __u32 buf_len; /* input/output: buf len */ 409 __aligned_u64 buf; /* input/output: 410 * tp_name for tracepoint 411 * symbol for kprobe 412 * filename for uprobe 413 */ 414 __u32 prog_id; /* output: prod_id */ 415 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 416 __u64 probe_offset; /* output: probe_offset */ 417 __u64 probe_addr; /* output: probe_addr */ 418 } task_fd_query; 419 } __attribute__((aligned(8))); 420 421 /* The description below is an attempt at providing documentation to eBPF 422 * developers about the multiple available eBPF helper functions. It can be 423 * parsed and used to produce a manual page. The workflow is the following, 424 * and requires the rst2man utility: 425 * 426 * $ ./scripts/bpf_helpers_doc.py \ 427 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 428 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 429 * $ man /tmp/bpf-helpers.7 430 * 431 * Note that in order to produce this external documentation, some RST 432 * formatting is used in the descriptions to get "bold" and "italics" in 433 * manual pages. Also note that the few trailing white spaces are 434 * intentional, removing them would break paragraphs for rst2man. 435 * 436 * Start of BPF helper function descriptions: 437 * 438 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 439 * Description 440 * Perform a lookup in *map* for an entry associated to *key*. 441 * Return 442 * Map value associated to *key*, or **NULL** if no entry was 443 * found. 444 * 445 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 446 * Description 447 * Add or update the value of the entry associated to *key* in 448 * *map* with *value*. *flags* is one of: 449 * 450 * **BPF_NOEXIST** 451 * The entry for *key* must not exist in the map. 452 * **BPF_EXIST** 453 * The entry for *key* must already exist in the map. 454 * **BPF_ANY** 455 * No condition on the existence of the entry for *key*. 456 * 457 * Flag value **BPF_NOEXIST** cannot be used for maps of types 458 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 459 * elements always exist), the helper would return an error. 460 * Return 461 * 0 on success, or a negative error in case of failure. 462 * 463 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 464 * Description 465 * Delete entry with *key* from *map*. 466 * Return 467 * 0 on success, or a negative error in case of failure. 468 * 469 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 470 * Description 471 * Push an element *value* in *map*. *flags* is one of: 472 * 473 * **BPF_EXIST** 474 * If the queue/stack is full, the oldest element is removed to 475 * make room for this. 476 * Return 477 * 0 on success, or a negative error in case of failure. 478 * 479 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 480 * Description 481 * Pop an element from *map*. 482 * Return 483 * 0 on success, or a negative error in case of failure. 484 * 485 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 486 * Description 487 * Get an element from *map* without removing it. 488 * Return 489 * 0 on success, or a negative error in case of failure. 490 * 491 * int bpf_probe_read(void *dst, u32 size, const void *src) 492 * Description 493 * For tracing programs, safely attempt to read *size* bytes from 494 * address *src* and store the data in *dst*. 495 * Return 496 * 0 on success, or a negative error in case of failure. 497 * 498 * u64 bpf_ktime_get_ns(void) 499 * Description 500 * Return the time elapsed since system boot, in nanoseconds. 501 * Return 502 * Current *ktime*. 503 * 504 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 505 * Description 506 * This helper is a "printk()-like" facility for debugging. It 507 * prints a message defined by format *fmt* (of size *fmt_size*) 508 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 509 * available. It can take up to three additional **u64** 510 * arguments (as an eBPF helpers, the total number of arguments is 511 * limited to five). 512 * 513 * Each time the helper is called, it appends a line to the trace. 514 * The format of the trace is customizable, and the exact output 515 * one will get depends on the options set in 516 * *\/sys/kernel/debug/tracing/trace_options* (see also the 517 * *README* file under the same directory). However, it usually 518 * defaults to something like: 519 * 520 * :: 521 * 522 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 523 * 524 * In the above: 525 * 526 * * ``telnet`` is the name of the current task. 527 * * ``470`` is the PID of the current task. 528 * * ``001`` is the CPU number on which the task is 529 * running. 530 * * In ``.N..``, each character refers to a set of 531 * options (whether irqs are enabled, scheduling 532 * options, whether hard/softirqs are running, level of 533 * preempt_disabled respectively). **N** means that 534 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 535 * are set. 536 * * ``419421.045894`` is a timestamp. 537 * * ``0x00000001`` is a fake value used by BPF for the 538 * instruction pointer register. 539 * * ``<formatted msg>`` is the message formatted with 540 * *fmt*. 541 * 542 * The conversion specifiers supported by *fmt* are similar, but 543 * more limited than for printk(). They are **%d**, **%i**, 544 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 545 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 546 * of field, padding with zeroes, etc.) is available, and the 547 * helper will return **-EINVAL** (but print nothing) if it 548 * encounters an unknown specifier. 549 * 550 * Also, note that **bpf_trace_printk**\ () is slow, and should 551 * only be used for debugging purposes. For this reason, a notice 552 * bloc (spanning several lines) is printed to kernel logs and 553 * states that the helper should not be used "for production use" 554 * the first time this helper is used (or more precisely, when 555 * **trace_printk**\ () buffers are allocated). For passing values 556 * to user space, perf events should be preferred. 557 * Return 558 * The number of bytes written to the buffer, or a negative error 559 * in case of failure. 560 * 561 * u32 bpf_get_prandom_u32(void) 562 * Description 563 * Get a pseudo-random number. 564 * 565 * From a security point of view, this helper uses its own 566 * pseudo-random internal state, and cannot be used to infer the 567 * seed of other random functions in the kernel. However, it is 568 * essential to note that the generator used by the helper is not 569 * cryptographically secure. 570 * Return 571 * A random 32-bit unsigned value. 572 * 573 * u32 bpf_get_smp_processor_id(void) 574 * Description 575 * Get the SMP (symmetric multiprocessing) processor id. Note that 576 * all programs run with preemption disabled, which means that the 577 * SMP processor id is stable during all the execution of the 578 * program. 579 * Return 580 * The SMP id of the processor running the program. 581 * 582 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 583 * Description 584 * Store *len* bytes from address *from* into the packet 585 * associated to *skb*, at *offset*. *flags* are a combination of 586 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 587 * checksum for the packet after storing the bytes) and 588 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 589 * **->swhash** and *skb*\ **->l4hash** to 0). 590 * 591 * A call to this helper is susceptible to change the underlaying 592 * packet buffer. Therefore, at load time, all checks on pointers 593 * previously done by the verifier are invalidated and must be 594 * performed again, if the helper is used in combination with 595 * direct packet access. 596 * Return 597 * 0 on success, or a negative error in case of failure. 598 * 599 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 600 * Description 601 * Recompute the layer 3 (e.g. IP) checksum for the packet 602 * associated to *skb*. Computation is incremental, so the helper 603 * must know the former value of the header field that was 604 * modified (*from*), the new value of this field (*to*), and the 605 * number of bytes (2 or 4) for this field, stored in *size*. 606 * Alternatively, it is possible to store the difference between 607 * the previous and the new values of the header field in *to*, by 608 * setting *from* and *size* to 0. For both methods, *offset* 609 * indicates the location of the IP checksum within the packet. 610 * 611 * This helper works in combination with **bpf_csum_diff**\ (), 612 * which does not update the checksum in-place, but offers more 613 * flexibility and can handle sizes larger than 2 or 4 for the 614 * checksum to update. 615 * 616 * A call to this helper is susceptible to change the underlaying 617 * packet buffer. Therefore, at load time, all checks on pointers 618 * previously done by the verifier are invalidated and must be 619 * performed again, if the helper is used in combination with 620 * direct packet access. 621 * Return 622 * 0 on success, or a negative error in case of failure. 623 * 624 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 625 * Description 626 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 627 * packet associated to *skb*. Computation is incremental, so the 628 * helper must know the former value of the header field that was 629 * modified (*from*), the new value of this field (*to*), and the 630 * number of bytes (2 or 4) for this field, stored on the lowest 631 * four bits of *flags*. Alternatively, it is possible to store 632 * the difference between the previous and the new values of the 633 * header field in *to*, by setting *from* and the four lowest 634 * bits of *flags* to 0. For both methods, *offset* indicates the 635 * location of the IP checksum within the packet. In addition to 636 * the size of the field, *flags* can be added (bitwise OR) actual 637 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 638 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 639 * for updates resulting in a null checksum the value is set to 640 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 641 * the checksum is to be computed against a pseudo-header. 642 * 643 * This helper works in combination with **bpf_csum_diff**\ (), 644 * which does not update the checksum in-place, but offers more 645 * flexibility and can handle sizes larger than 2 or 4 for the 646 * checksum to update. 647 * 648 * A call to this helper is susceptible to change the underlaying 649 * packet buffer. Therefore, at load time, all checks on pointers 650 * previously done by the verifier are invalidated and must be 651 * performed again, if the helper is used in combination with 652 * direct packet access. 653 * Return 654 * 0 on success, or a negative error in case of failure. 655 * 656 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 657 * Description 658 * This special helper is used to trigger a "tail call", or in 659 * other words, to jump into another eBPF program. The same stack 660 * frame is used (but values on stack and in registers for the 661 * caller are not accessible to the callee). This mechanism allows 662 * for program chaining, either for raising the maximum number of 663 * available eBPF instructions, or to execute given programs in 664 * conditional blocks. For security reasons, there is an upper 665 * limit to the number of successive tail calls that can be 666 * performed. 667 * 668 * Upon call of this helper, the program attempts to jump into a 669 * program referenced at index *index* in *prog_array_map*, a 670 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 671 * *ctx*, a pointer to the context. 672 * 673 * If the call succeeds, the kernel immediately runs the first 674 * instruction of the new program. This is not a function call, 675 * and it never returns to the previous program. If the call 676 * fails, then the helper has no effect, and the caller continues 677 * to run its subsequent instructions. A call can fail if the 678 * destination program for the jump does not exist (i.e. *index* 679 * is superior to the number of entries in *prog_array_map*), or 680 * if the maximum number of tail calls has been reached for this 681 * chain of programs. This limit is defined in the kernel by the 682 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 683 * which is currently set to 32. 684 * Return 685 * 0 on success, or a negative error in case of failure. 686 * 687 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 688 * Description 689 * Clone and redirect the packet associated to *skb* to another 690 * net device of index *ifindex*. Both ingress and egress 691 * interfaces can be used for redirection. The **BPF_F_INGRESS** 692 * value in *flags* is used to make the distinction (ingress path 693 * is selected if the flag is present, egress path otherwise). 694 * This is the only flag supported for now. 695 * 696 * In comparison with **bpf_redirect**\ () helper, 697 * **bpf_clone_redirect**\ () has the associated cost of 698 * duplicating the packet buffer, but this can be executed out of 699 * the eBPF program. Conversely, **bpf_redirect**\ () is more 700 * efficient, but it is handled through an action code where the 701 * redirection happens only after the eBPF program has returned. 702 * 703 * A call to this helper is susceptible to change the underlaying 704 * packet buffer. Therefore, at load time, all checks on pointers 705 * previously done by the verifier are invalidated and must be 706 * performed again, if the helper is used in combination with 707 * direct packet access. 708 * Return 709 * 0 on success, or a negative error in case of failure. 710 * 711 * u64 bpf_get_current_pid_tgid(void) 712 * Return 713 * A 64-bit integer containing the current tgid and pid, and 714 * created as such: 715 * *current_task*\ **->tgid << 32 \|** 716 * *current_task*\ **->pid**. 717 * 718 * u64 bpf_get_current_uid_gid(void) 719 * Return 720 * A 64-bit integer containing the current GID and UID, and 721 * created as such: *current_gid* **<< 32 \|** *current_uid*. 722 * 723 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 724 * Description 725 * Copy the **comm** attribute of the current task into *buf* of 726 * *size_of_buf*. The **comm** attribute contains the name of 727 * the executable (excluding the path) for the current task. The 728 * *size_of_buf* must be strictly positive. On success, the 729 * helper makes sure that the *buf* is NUL-terminated. On failure, 730 * it is filled with zeroes. 731 * Return 732 * 0 on success, or a negative error in case of failure. 733 * 734 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 735 * Description 736 * Retrieve the classid for the current task, i.e. for the net_cls 737 * cgroup to which *skb* belongs. 738 * 739 * This helper can be used on TC egress path, but not on ingress. 740 * 741 * The net_cls cgroup provides an interface to tag network packets 742 * based on a user-provided identifier for all traffic coming from 743 * the tasks belonging to the related cgroup. See also the related 744 * kernel documentation, available from the Linux sources in file 745 * *Documentation/cgroup-v1/net_cls.txt*. 746 * 747 * The Linux kernel has two versions for cgroups: there are 748 * cgroups v1 and cgroups v2. Both are available to users, who can 749 * use a mixture of them, but note that the net_cls cgroup is for 750 * cgroup v1 only. This makes it incompatible with BPF programs 751 * run on cgroups, which is a cgroup-v2-only feature (a socket can 752 * only hold data for one version of cgroups at a time). 753 * 754 * This helper is only available is the kernel was compiled with 755 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 756 * "**y**" or to "**m**". 757 * Return 758 * The classid, or 0 for the default unconfigured classid. 759 * 760 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 761 * Description 762 * Push a *vlan_tci* (VLAN tag control information) of protocol 763 * *vlan_proto* to the packet associated to *skb*, then update 764 * the checksum. Note that if *vlan_proto* is different from 765 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 766 * be **ETH_P_8021Q**. 767 * 768 * A call to this helper is susceptible to change the underlaying 769 * packet buffer. Therefore, at load time, all checks on pointers 770 * previously done by the verifier are invalidated and must be 771 * performed again, if the helper is used in combination with 772 * direct packet access. 773 * Return 774 * 0 on success, or a negative error in case of failure. 775 * 776 * int bpf_skb_vlan_pop(struct sk_buff *skb) 777 * Description 778 * Pop a VLAN header from the packet associated to *skb*. 779 * 780 * A call to this helper is susceptible to change the underlaying 781 * packet buffer. Therefore, at load time, all checks on pointers 782 * previously done by the verifier are invalidated and must be 783 * performed again, if the helper is used in combination with 784 * direct packet access. 785 * Return 786 * 0 on success, or a negative error in case of failure. 787 * 788 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 789 * Description 790 * Get tunnel metadata. This helper takes a pointer *key* to an 791 * empty **struct bpf_tunnel_key** of **size**, that will be 792 * filled with tunnel metadata for the packet associated to *skb*. 793 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 794 * indicates that the tunnel is based on IPv6 protocol instead of 795 * IPv4. 796 * 797 * The **struct bpf_tunnel_key** is an object that generalizes the 798 * principal parameters used by various tunneling protocols into a 799 * single struct. This way, it can be used to easily make a 800 * decision based on the contents of the encapsulation header, 801 * "summarized" in this struct. In particular, it holds the IP 802 * address of the remote end (IPv4 or IPv6, depending on the case) 803 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 804 * this struct exposes the *key*\ **->tunnel_id**, which is 805 * generally mapped to a VNI (Virtual Network Identifier), making 806 * it programmable together with the **bpf_skb_set_tunnel_key**\ 807 * () helper. 808 * 809 * Let's imagine that the following code is part of a program 810 * attached to the TC ingress interface, on one end of a GRE 811 * tunnel, and is supposed to filter out all messages coming from 812 * remote ends with IPv4 address other than 10.0.0.1: 813 * 814 * :: 815 * 816 * int ret; 817 * struct bpf_tunnel_key key = {}; 818 * 819 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 820 * if (ret < 0) 821 * return TC_ACT_SHOT; // drop packet 822 * 823 * if (key.remote_ipv4 != 0x0a000001) 824 * return TC_ACT_SHOT; // drop packet 825 * 826 * return TC_ACT_OK; // accept packet 827 * 828 * This interface can also be used with all encapsulation devices 829 * that can operate in "collect metadata" mode: instead of having 830 * one network device per specific configuration, the "collect 831 * metadata" mode only requires a single device where the 832 * configuration can be extracted from this helper. 833 * 834 * This can be used together with various tunnels such as VXLan, 835 * Geneve, GRE or IP in IP (IPIP). 836 * Return 837 * 0 on success, or a negative error in case of failure. 838 * 839 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 840 * Description 841 * Populate tunnel metadata for packet associated to *skb.* The 842 * tunnel metadata is set to the contents of *key*, of *size*. The 843 * *flags* can be set to a combination of the following values: 844 * 845 * **BPF_F_TUNINFO_IPV6** 846 * Indicate that the tunnel is based on IPv6 protocol 847 * instead of IPv4. 848 * **BPF_F_ZERO_CSUM_TX** 849 * For IPv4 packets, add a flag to tunnel metadata 850 * indicating that checksum computation should be skipped 851 * and checksum set to zeroes. 852 * **BPF_F_DONT_FRAGMENT** 853 * Add a flag to tunnel metadata indicating that the 854 * packet should not be fragmented. 855 * **BPF_F_SEQ_NUMBER** 856 * Add a flag to tunnel metadata indicating that a 857 * sequence number should be added to tunnel header before 858 * sending the packet. This flag was added for GRE 859 * encapsulation, but might be used with other protocols 860 * as well in the future. 861 * 862 * Here is a typical usage on the transmit path: 863 * 864 * :: 865 * 866 * struct bpf_tunnel_key key; 867 * populate key ... 868 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 869 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 870 * 871 * See also the description of the **bpf_skb_get_tunnel_key**\ () 872 * helper for additional information. 873 * Return 874 * 0 on success, or a negative error in case of failure. 875 * 876 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 877 * Description 878 * Read the value of a perf event counter. This helper relies on a 879 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 880 * the perf event counter is selected when *map* is updated with 881 * perf event file descriptors. The *map* is an array whose size 882 * is the number of available CPUs, and each cell contains a value 883 * relative to one CPU. The value to retrieve is indicated by 884 * *flags*, that contains the index of the CPU to look up, masked 885 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 886 * **BPF_F_CURRENT_CPU** to indicate that the value for the 887 * current CPU should be retrieved. 888 * 889 * Note that before Linux 4.13, only hardware perf event can be 890 * retrieved. 891 * 892 * Also, be aware that the newer helper 893 * **bpf_perf_event_read_value**\ () is recommended over 894 * **bpf_perf_event_read**\ () in general. The latter has some ABI 895 * quirks where error and counter value are used as a return code 896 * (which is wrong to do since ranges may overlap). This issue is 897 * fixed with **bpf_perf_event_read_value**\ (), which at the same 898 * time provides more features over the **bpf_perf_event_read**\ 899 * () interface. Please refer to the description of 900 * **bpf_perf_event_read_value**\ () for details. 901 * Return 902 * The value of the perf event counter read from the map, or a 903 * negative error code in case of failure. 904 * 905 * int bpf_redirect(u32 ifindex, u64 flags) 906 * Description 907 * Redirect the packet to another net device of index *ifindex*. 908 * This helper is somewhat similar to **bpf_clone_redirect**\ 909 * (), except that the packet is not cloned, which provides 910 * increased performance. 911 * 912 * Except for XDP, both ingress and egress interfaces can be used 913 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 914 * to make the distinction (ingress path is selected if the flag 915 * is present, egress path otherwise). Currently, XDP only 916 * supports redirection to the egress interface, and accepts no 917 * flag at all. 918 * 919 * The same effect can be attained with the more generic 920 * **bpf_redirect_map**\ (), which requires specific maps to be 921 * used but offers better performance. 922 * Return 923 * For XDP, the helper returns **XDP_REDIRECT** on success or 924 * **XDP_ABORTED** on error. For other program types, the values 925 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 926 * error. 927 * 928 * u32 bpf_get_route_realm(struct sk_buff *skb) 929 * Description 930 * Retrieve the realm or the route, that is to say the 931 * **tclassid** field of the destination for the *skb*. The 932 * indentifier retrieved is a user-provided tag, similar to the 933 * one used with the net_cls cgroup (see description for 934 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 935 * held by a route (a destination entry), not by a task. 936 * 937 * Retrieving this identifier works with the clsact TC egress hook 938 * (see also **tc-bpf(8)**), or alternatively on conventional 939 * classful egress qdiscs, but not on TC ingress path. In case of 940 * clsact TC egress hook, this has the advantage that, internally, 941 * the destination entry has not been dropped yet in the transmit 942 * path. Therefore, the destination entry does not need to be 943 * artificially held via **netif_keep_dst**\ () for a classful 944 * qdisc until the *skb* is freed. 945 * 946 * This helper is available only if the kernel was compiled with 947 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 948 * Return 949 * The realm of the route for the packet associated to *skb*, or 0 950 * if none was found. 951 * 952 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 953 * Description 954 * Write raw *data* blob into a special BPF perf event held by 955 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 956 * event must have the following attributes: **PERF_SAMPLE_RAW** 957 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 958 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 959 * 960 * The *flags* are used to indicate the index in *map* for which 961 * the value must be put, masked with **BPF_F_INDEX_MASK**. 962 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 963 * to indicate that the index of the current CPU core should be 964 * used. 965 * 966 * The value to write, of *size*, is passed through eBPF stack and 967 * pointed by *data*. 968 * 969 * The context of the program *ctx* needs also be passed to the 970 * helper. 971 * 972 * On user space, a program willing to read the values needs to 973 * call **perf_event_open**\ () on the perf event (either for 974 * one or for all CPUs) and to store the file descriptor into the 975 * *map*. This must be done before the eBPF program can send data 976 * into it. An example is available in file 977 * *samples/bpf/trace_output_user.c* in the Linux kernel source 978 * tree (the eBPF program counterpart is in 979 * *samples/bpf/trace_output_kern.c*). 980 * 981 * **bpf_perf_event_output**\ () achieves better performance 982 * than **bpf_trace_printk**\ () for sharing data with user 983 * space, and is much better suitable for streaming data from eBPF 984 * programs. 985 * 986 * Note that this helper is not restricted to tracing use cases 987 * and can be used with programs attached to TC or XDP as well, 988 * where it allows for passing data to user space listeners. Data 989 * can be: 990 * 991 * * Only custom structs, 992 * * Only the packet payload, or 993 * * A combination of both. 994 * Return 995 * 0 on success, or a negative error in case of failure. 996 * 997 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 998 * Description 999 * This helper was provided as an easy way to load data from a 1000 * packet. It can be used to load *len* bytes from *offset* from 1001 * the packet associated to *skb*, into the buffer pointed by 1002 * *to*. 1003 * 1004 * Since Linux 4.7, usage of this helper has mostly been replaced 1005 * by "direct packet access", enabling packet data to be 1006 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1007 * pointing respectively to the first byte of packet data and to 1008 * the byte after the last byte of packet data. However, it 1009 * remains useful if one wishes to read large quantities of data 1010 * at once from a packet into the eBPF stack. 1011 * Return 1012 * 0 on success, or a negative error in case of failure. 1013 * 1014 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 1015 * Description 1016 * Walk a user or a kernel stack and return its id. To achieve 1017 * this, the helper needs *ctx*, which is a pointer to the context 1018 * on which the tracing program is executed, and a pointer to a 1019 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1020 * 1021 * The last argument, *flags*, holds the number of stack frames to 1022 * skip (from 0 to 255), masked with 1023 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1024 * a combination of the following flags: 1025 * 1026 * **BPF_F_USER_STACK** 1027 * Collect a user space stack instead of a kernel stack. 1028 * **BPF_F_FAST_STACK_CMP** 1029 * Compare stacks by hash only. 1030 * **BPF_F_REUSE_STACKID** 1031 * If two different stacks hash into the same *stackid*, 1032 * discard the old one. 1033 * 1034 * The stack id retrieved is a 32 bit long integer handle which 1035 * can be further combined with other data (including other stack 1036 * ids) and used as a key into maps. This can be useful for 1037 * generating a variety of graphs (such as flame graphs or off-cpu 1038 * graphs). 1039 * 1040 * For walking a stack, this helper is an improvement over 1041 * **bpf_probe_read**\ (), which can be used with unrolled loops 1042 * but is not efficient and consumes a lot of eBPF instructions. 1043 * Instead, **bpf_get_stackid**\ () can collect up to 1044 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1045 * this limit can be controlled with the **sysctl** program, and 1046 * that it should be manually increased in order to profile long 1047 * user stacks (such as stacks for Java programs). To do so, use: 1048 * 1049 * :: 1050 * 1051 * # sysctl kernel.perf_event_max_stack=<new value> 1052 * Return 1053 * The positive or null stack id on success, or a negative error 1054 * in case of failure. 1055 * 1056 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1057 * Description 1058 * Compute a checksum difference, from the raw buffer pointed by 1059 * *from*, of length *from_size* (that must be a multiple of 4), 1060 * towards the raw buffer pointed by *to*, of size *to_size* 1061 * (same remark). An optional *seed* can be added to the value 1062 * (this can be cascaded, the seed may come from a previous call 1063 * to the helper). 1064 * 1065 * This is flexible enough to be used in several ways: 1066 * 1067 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1068 * checksum, it can be used when pushing new data. 1069 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1070 * checksum, it can be used when removing data from a packet. 1071 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1072 * can be used to compute a diff. Note that *from_size* and 1073 * *to_size* do not need to be equal. 1074 * 1075 * This helper can be used in combination with 1076 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1077 * which one can feed in the difference computed with 1078 * **bpf_csum_diff**\ (). 1079 * Return 1080 * The checksum result, or a negative error code in case of 1081 * failure. 1082 * 1083 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1084 * Description 1085 * Retrieve tunnel options metadata for the packet associated to 1086 * *skb*, and store the raw tunnel option data to the buffer *opt* 1087 * of *size*. 1088 * 1089 * This helper can be used with encapsulation devices that can 1090 * operate in "collect metadata" mode (please refer to the related 1091 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1092 * more details). A particular example where this can be used is 1093 * in combination with the Geneve encapsulation protocol, where it 1094 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1095 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1096 * the eBPF program. This allows for full customization of these 1097 * headers. 1098 * Return 1099 * The size of the option data retrieved. 1100 * 1101 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1102 * Description 1103 * Set tunnel options metadata for the packet associated to *skb* 1104 * to the option data contained in the raw buffer *opt* of *size*. 1105 * 1106 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1107 * helper for additional information. 1108 * Return 1109 * 0 on success, or a negative error in case of failure. 1110 * 1111 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1112 * Description 1113 * Change the protocol of the *skb* to *proto*. Currently 1114 * supported are transition from IPv4 to IPv6, and from IPv6 to 1115 * IPv4. The helper takes care of the groundwork for the 1116 * transition, including resizing the socket buffer. The eBPF 1117 * program is expected to fill the new headers, if any, via 1118 * **skb_store_bytes**\ () and to recompute the checksums with 1119 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1120 * (). The main case for this helper is to perform NAT64 1121 * operations out of an eBPF program. 1122 * 1123 * Internally, the GSO type is marked as dodgy so that headers are 1124 * checked and segments are recalculated by the GSO/GRO engine. 1125 * The size for GSO target is adapted as well. 1126 * 1127 * All values for *flags* are reserved for future usage, and must 1128 * be left at zero. 1129 * 1130 * A call to this helper is susceptible to change the underlaying 1131 * packet buffer. Therefore, at load time, all checks on pointers 1132 * previously done by the verifier are invalidated and must be 1133 * performed again, if the helper is used in combination with 1134 * direct packet access. 1135 * Return 1136 * 0 on success, or a negative error in case of failure. 1137 * 1138 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1139 * Description 1140 * Change the packet type for the packet associated to *skb*. This 1141 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1142 * the eBPF program does not have a write access to *skb*\ 1143 * **->pkt_type** beside this helper. Using a helper here allows 1144 * for graceful handling of errors. 1145 * 1146 * The major use case is to change incoming *skb*s to 1147 * **PACKET_HOST** in a programmatic way instead of having to 1148 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1149 * example. 1150 * 1151 * Note that *type* only allows certain values. At this time, they 1152 * are: 1153 * 1154 * **PACKET_HOST** 1155 * Packet is for us. 1156 * **PACKET_BROADCAST** 1157 * Send packet to all. 1158 * **PACKET_MULTICAST** 1159 * Send packet to group. 1160 * **PACKET_OTHERHOST** 1161 * Send packet to someone else. 1162 * Return 1163 * 0 on success, or a negative error in case of failure. 1164 * 1165 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1166 * Description 1167 * Check whether *skb* is a descendant of the cgroup2 held by 1168 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1169 * Return 1170 * The return value depends on the result of the test, and can be: 1171 * 1172 * * 0, if the *skb* failed the cgroup2 descendant test. 1173 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1174 * * A negative error code, if an error occurred. 1175 * 1176 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1177 * Description 1178 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1179 * not set, in particular if the hash was cleared due to mangling, 1180 * recompute this hash. Later accesses to the hash can be done 1181 * directly with *skb*\ **->hash**. 1182 * 1183 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1184 * prototype with **bpf_skb_change_proto**\ (), or calling 1185 * **bpf_skb_store_bytes**\ () with the 1186 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1187 * the hash and to trigger a new computation for the next call to 1188 * **bpf_get_hash_recalc**\ (). 1189 * Return 1190 * The 32-bit hash. 1191 * 1192 * u64 bpf_get_current_task(void) 1193 * Return 1194 * A pointer to the current task struct. 1195 * 1196 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1197 * Description 1198 * Attempt in a safe way to write *len* bytes from the buffer 1199 * *src* to *dst* in memory. It only works for threads that are in 1200 * user context, and *dst* must be a valid user space address. 1201 * 1202 * This helper should not be used to implement any kind of 1203 * security mechanism because of TOC-TOU attacks, but rather to 1204 * debug, divert, and manipulate execution of semi-cooperative 1205 * processes. 1206 * 1207 * Keep in mind that this feature is meant for experiments, and it 1208 * has a risk of crashing the system and running programs. 1209 * Therefore, when an eBPF program using this helper is attached, 1210 * a warning including PID and process name is printed to kernel 1211 * logs. 1212 * Return 1213 * 0 on success, or a negative error in case of failure. 1214 * 1215 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1216 * Description 1217 * Check whether the probe is being run is the context of a given 1218 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1219 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1220 * Return 1221 * The return value depends on the result of the test, and can be: 1222 * 1223 * * 0, if the *skb* task belongs to the cgroup2. 1224 * * 1, if the *skb* task does not belong to the cgroup2. 1225 * * A negative error code, if an error occurred. 1226 * 1227 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1228 * Description 1229 * Resize (trim or grow) the packet associated to *skb* to the 1230 * new *len*. The *flags* are reserved for future usage, and must 1231 * be left at zero. 1232 * 1233 * The basic idea is that the helper performs the needed work to 1234 * change the size of the packet, then the eBPF program rewrites 1235 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1236 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1237 * and others. This helper is a slow path utility intended for 1238 * replies with control messages. And because it is targeted for 1239 * slow path, the helper itself can afford to be slow: it 1240 * implicitly linearizes, unclones and drops offloads from the 1241 * *skb*. 1242 * 1243 * A call to this helper is susceptible to change the underlaying 1244 * packet buffer. Therefore, at load time, all checks on pointers 1245 * previously done by the verifier are invalidated and must be 1246 * performed again, if the helper is used in combination with 1247 * direct packet access. 1248 * Return 1249 * 0 on success, or a negative error in case of failure. 1250 * 1251 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1252 * Description 1253 * Pull in non-linear data in case the *skb* is non-linear and not 1254 * all of *len* are part of the linear section. Make *len* bytes 1255 * from *skb* readable and writable. If a zero value is passed for 1256 * *len*, then the whole length of the *skb* is pulled. 1257 * 1258 * This helper is only needed for reading and writing with direct 1259 * packet access. 1260 * 1261 * For direct packet access, testing that offsets to access 1262 * are within packet boundaries (test on *skb*\ **->data_end**) is 1263 * susceptible to fail if offsets are invalid, or if the requested 1264 * data is in non-linear parts of the *skb*. On failure the 1265 * program can just bail out, or in the case of a non-linear 1266 * buffer, use a helper to make the data available. The 1267 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1268 * the data. Another one consists in using **bpf_skb_pull_data** 1269 * to pull in once the non-linear parts, then retesting and 1270 * eventually access the data. 1271 * 1272 * At the same time, this also makes sure the *skb* is uncloned, 1273 * which is a necessary condition for direct write. As this needs 1274 * to be an invariant for the write part only, the verifier 1275 * detects writes and adds a prologue that is calling 1276 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1277 * the very beginning in case it is indeed cloned. 1278 * 1279 * A call to this helper is susceptible to change the underlaying 1280 * packet buffer. Therefore, at load time, all checks on pointers 1281 * previously done by the verifier are invalidated and must be 1282 * performed again, if the helper is used in combination with 1283 * direct packet access. 1284 * Return 1285 * 0 on success, or a negative error in case of failure. 1286 * 1287 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1288 * Description 1289 * Add the checksum *csum* into *skb*\ **->csum** in case the 1290 * driver has supplied a checksum for the entire packet into that 1291 * field. Return an error otherwise. This helper is intended to be 1292 * used in combination with **bpf_csum_diff**\ (), in particular 1293 * when the checksum needs to be updated after data has been 1294 * written into the packet through direct packet access. 1295 * Return 1296 * The checksum on success, or a negative error code in case of 1297 * failure. 1298 * 1299 * void bpf_set_hash_invalid(struct sk_buff *skb) 1300 * Description 1301 * Invalidate the current *skb*\ **->hash**. It can be used after 1302 * mangling on headers through direct packet access, in order to 1303 * indicate that the hash is outdated and to trigger a 1304 * recalculation the next time the kernel tries to access this 1305 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1306 * 1307 * int bpf_get_numa_node_id(void) 1308 * Description 1309 * Return the id of the current NUMA node. The primary use case 1310 * for this helper is the selection of sockets for the local NUMA 1311 * node, when the program is attached to sockets using the 1312 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1313 * but the helper is also available to other eBPF program types, 1314 * similarly to **bpf_get_smp_processor_id**\ (). 1315 * Return 1316 * The id of current NUMA node. 1317 * 1318 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1319 * Description 1320 * Grows headroom of packet associated to *skb* and adjusts the 1321 * offset of the MAC header accordingly, adding *len* bytes of 1322 * space. It automatically extends and reallocates memory as 1323 * required. 1324 * 1325 * This helper can be used on a layer 3 *skb* to push a MAC header 1326 * for redirection into a layer 2 device. 1327 * 1328 * All values for *flags* are reserved for future usage, and must 1329 * be left at zero. 1330 * 1331 * A call to this helper is susceptible to change the underlaying 1332 * packet buffer. Therefore, at load time, all checks on pointers 1333 * previously done by the verifier are invalidated and must be 1334 * performed again, if the helper is used in combination with 1335 * direct packet access. 1336 * Return 1337 * 0 on success, or a negative error in case of failure. 1338 * 1339 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1340 * Description 1341 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1342 * it is possible to use a negative value for *delta*. This helper 1343 * can be used to prepare the packet for pushing or popping 1344 * headers. 1345 * 1346 * A call to this helper is susceptible to change the underlaying 1347 * packet buffer. Therefore, at load time, all checks on pointers 1348 * previously done by the verifier are invalidated and must be 1349 * performed again, if the helper is used in combination with 1350 * direct packet access. 1351 * Return 1352 * 0 on success, or a negative error in case of failure. 1353 * 1354 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1355 * Description 1356 * Copy a NUL terminated string from an unsafe address 1357 * *unsafe_ptr* to *dst*. The *size* should include the 1358 * terminating NUL byte. In case the string length is smaller than 1359 * *size*, the target is not padded with further NUL bytes. If the 1360 * string length is larger than *size*, just *size*-1 bytes are 1361 * copied and the last byte is set to NUL. 1362 * 1363 * On success, the length of the copied string is returned. This 1364 * makes this helper useful in tracing programs for reading 1365 * strings, and more importantly to get its length at runtime. See 1366 * the following snippet: 1367 * 1368 * :: 1369 * 1370 * SEC("kprobe/sys_open") 1371 * void bpf_sys_open(struct pt_regs *ctx) 1372 * { 1373 * char buf[PATHLEN]; // PATHLEN is defined to 256 1374 * int res = bpf_probe_read_str(buf, sizeof(buf), 1375 * ctx->di); 1376 * 1377 * // Consume buf, for example push it to 1378 * // userspace via bpf_perf_event_output(); we 1379 * // can use res (the string length) as event 1380 * // size, after checking its boundaries. 1381 * } 1382 * 1383 * In comparison, using **bpf_probe_read()** helper here instead 1384 * to read the string would require to estimate the length at 1385 * compile time, and would often result in copying more memory 1386 * than necessary. 1387 * 1388 * Another useful use case is when parsing individual process 1389 * arguments or individual environment variables navigating 1390 * *current*\ **->mm->arg_start** and *current*\ 1391 * **->mm->env_start**: using this helper and the return value, 1392 * one can quickly iterate at the right offset of the memory area. 1393 * Return 1394 * On success, the strictly positive length of the string, 1395 * including the trailing NUL character. On error, a negative 1396 * value. 1397 * 1398 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1399 * Description 1400 * If the **struct sk_buff** pointed by *skb* has a known socket, 1401 * retrieve the cookie (generated by the kernel) of this socket. 1402 * If no cookie has been set yet, generate a new cookie. Once 1403 * generated, the socket cookie remains stable for the life of the 1404 * socket. This helper can be useful for monitoring per socket 1405 * networking traffic statistics as it provides a unique socket 1406 * identifier per namespace. 1407 * Return 1408 * A 8-byte long non-decreasing number on success, or 0 if the 1409 * socket field is missing inside *skb*. 1410 * 1411 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1412 * Description 1413 * Equivalent to bpf_get_socket_cookie() helper that accepts 1414 * *skb*, but gets socket from **struct bpf_sock_addr** contex. 1415 * Return 1416 * A 8-byte long non-decreasing number. 1417 * 1418 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1419 * Description 1420 * Equivalent to bpf_get_socket_cookie() helper that accepts 1421 * *skb*, but gets socket from **struct bpf_sock_ops** contex. 1422 * Return 1423 * A 8-byte long non-decreasing number. 1424 * 1425 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1426 * Return 1427 * The owner UID of the socket associated to *skb*. If the socket 1428 * is **NULL**, or if it is not a full socket (i.e. if it is a 1429 * time-wait or a request socket instead), **overflowuid** value 1430 * is returned (note that **overflowuid** might also be the actual 1431 * UID value for the socket). 1432 * 1433 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1434 * Description 1435 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1436 * to value *hash*. 1437 * Return 1438 * 0 1439 * 1440 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1441 * Description 1442 * Emulate a call to **setsockopt()** on the socket associated to 1443 * *bpf_socket*, which must be a full socket. The *level* at 1444 * which the option resides and the name *optname* of the option 1445 * must be specified, see **setsockopt(2)** for more information. 1446 * The option value of length *optlen* is pointed by *optval*. 1447 * 1448 * This helper actually implements a subset of **setsockopt()**. 1449 * It supports the following *level*\ s: 1450 * 1451 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1452 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1453 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1454 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1455 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1456 * **TCP_BPF_SNDCWND_CLAMP**. 1457 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1458 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1459 * Return 1460 * 0 on success, or a negative error in case of failure. 1461 * 1462 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1463 * Description 1464 * Grow or shrink the room for data in the packet associated to 1465 * *skb* by *len_diff*, and according to the selected *mode*. 1466 * 1467 * There is a single supported mode at this time: 1468 * 1469 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1470 * (room space is added or removed below the layer 3 header). 1471 * 1472 * All values for *flags* are reserved for future usage, and must 1473 * be left at zero. 1474 * 1475 * A call to this helper is susceptible to change the underlaying 1476 * packet buffer. Therefore, at load time, all checks on pointers 1477 * previously done by the verifier are invalidated and must be 1478 * performed again, if the helper is used in combination with 1479 * direct packet access. 1480 * Return 1481 * 0 on success, or a negative error in case of failure. 1482 * 1483 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1484 * Description 1485 * Redirect the packet to the endpoint referenced by *map* at 1486 * index *key*. Depending on its type, this *map* can contain 1487 * references to net devices (for forwarding packets through other 1488 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1489 * but this is only implemented for native XDP (with driver 1490 * support) as of this writing). 1491 * 1492 * All values for *flags* are reserved for future usage, and must 1493 * be left at zero. 1494 * 1495 * When used to redirect packets to net devices, this helper 1496 * provides a high performance increase over **bpf_redirect**\ (). 1497 * This is due to various implementation details of the underlying 1498 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1499 * () tries to send packet as a "bulk" to the device. 1500 * Return 1501 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1502 * 1503 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1504 * Description 1505 * Redirect the packet to the socket referenced by *map* (of type 1506 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1507 * egress interfaces can be used for redirection. The 1508 * **BPF_F_INGRESS** value in *flags* is used to make the 1509 * distinction (ingress path is selected if the flag is present, 1510 * egress path otherwise). This is the only flag supported for now. 1511 * Return 1512 * **SK_PASS** on success, or **SK_DROP** on error. 1513 * 1514 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1515 * Description 1516 * Add an entry to, or update a *map* referencing sockets. The 1517 * *skops* is used as a new value for the entry associated to 1518 * *key*. *flags* is one of: 1519 * 1520 * **BPF_NOEXIST** 1521 * The entry for *key* must not exist in the map. 1522 * **BPF_EXIST** 1523 * The entry for *key* must already exist in the map. 1524 * **BPF_ANY** 1525 * No condition on the existence of the entry for *key*. 1526 * 1527 * If the *map* has eBPF programs (parser and verdict), those will 1528 * be inherited by the socket being added. If the socket is 1529 * already attached to eBPF programs, this results in an error. 1530 * Return 1531 * 0 on success, or a negative error in case of failure. 1532 * 1533 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1534 * Description 1535 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1536 * *delta* (which can be positive or negative). Note that this 1537 * operation modifies the address stored in *xdp_md*\ **->data**, 1538 * so the latter must be loaded only after the helper has been 1539 * called. 1540 * 1541 * The use of *xdp_md*\ **->data_meta** is optional and programs 1542 * are not required to use it. The rationale is that when the 1543 * packet is processed with XDP (e.g. as DoS filter), it is 1544 * possible to push further meta data along with it before passing 1545 * to the stack, and to give the guarantee that an ingress eBPF 1546 * program attached as a TC classifier on the same device can pick 1547 * this up for further post-processing. Since TC works with socket 1548 * buffers, it remains possible to set from XDP the **mark** or 1549 * **priority** pointers, or other pointers for the socket buffer. 1550 * Having this scratch space generic and programmable allows for 1551 * more flexibility as the user is free to store whatever meta 1552 * data they need. 1553 * 1554 * A call to this helper is susceptible to change the underlaying 1555 * packet buffer. Therefore, at load time, all checks on pointers 1556 * previously done by the verifier are invalidated and must be 1557 * performed again, if the helper is used in combination with 1558 * direct packet access. 1559 * Return 1560 * 0 on success, or a negative error in case of failure. 1561 * 1562 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1563 * Description 1564 * Read the value of a perf event counter, and store it into *buf* 1565 * of size *buf_size*. This helper relies on a *map* of type 1566 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1567 * counter is selected when *map* is updated with perf event file 1568 * descriptors. The *map* is an array whose size is the number of 1569 * available CPUs, and each cell contains a value relative to one 1570 * CPU. The value to retrieve is indicated by *flags*, that 1571 * contains the index of the CPU to look up, masked with 1572 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1573 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1574 * current CPU should be retrieved. 1575 * 1576 * This helper behaves in a way close to 1577 * **bpf_perf_event_read**\ () helper, save that instead of 1578 * just returning the value observed, it fills the *buf* 1579 * structure. This allows for additional data to be retrieved: in 1580 * particular, the enabled and running times (in *buf*\ 1581 * **->enabled** and *buf*\ **->running**, respectively) are 1582 * copied. In general, **bpf_perf_event_read_value**\ () is 1583 * recommended over **bpf_perf_event_read**\ (), which has some 1584 * ABI issues and provides fewer functionalities. 1585 * 1586 * These values are interesting, because hardware PMU (Performance 1587 * Monitoring Unit) counters are limited resources. When there are 1588 * more PMU based perf events opened than available counters, 1589 * kernel will multiplex these events so each event gets certain 1590 * percentage (but not all) of the PMU time. In case that 1591 * multiplexing happens, the number of samples or counter value 1592 * will not reflect the case compared to when no multiplexing 1593 * occurs. This makes comparison between different runs difficult. 1594 * Typically, the counter value should be normalized before 1595 * comparing to other experiments. The usual normalization is done 1596 * as follows. 1597 * 1598 * :: 1599 * 1600 * normalized_counter = counter * t_enabled / t_running 1601 * 1602 * Where t_enabled is the time enabled for event and t_running is 1603 * the time running for event since last normalization. The 1604 * enabled and running times are accumulated since the perf event 1605 * open. To achieve scaling factor between two invocations of an 1606 * eBPF program, users can can use CPU id as the key (which is 1607 * typical for perf array usage model) to remember the previous 1608 * value and do the calculation inside the eBPF program. 1609 * Return 1610 * 0 on success, or a negative error in case of failure. 1611 * 1612 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1613 * Description 1614 * For en eBPF program attached to a perf event, retrieve the 1615 * value of the event counter associated to *ctx* and store it in 1616 * the structure pointed by *buf* and of size *buf_size*. Enabled 1617 * and running times are also stored in the structure (see 1618 * description of helper **bpf_perf_event_read_value**\ () for 1619 * more details). 1620 * Return 1621 * 0 on success, or a negative error in case of failure. 1622 * 1623 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1624 * Description 1625 * Emulate a call to **getsockopt()** on the socket associated to 1626 * *bpf_socket*, which must be a full socket. The *level* at 1627 * which the option resides and the name *optname* of the option 1628 * must be specified, see **getsockopt(2)** for more information. 1629 * The retrieved value is stored in the structure pointed by 1630 * *opval* and of length *optlen*. 1631 * 1632 * This helper actually implements a subset of **getsockopt()**. 1633 * It supports the following *level*\ s: 1634 * 1635 * * **IPPROTO_TCP**, which supports *optname* 1636 * **TCP_CONGESTION**. 1637 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1638 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1639 * Return 1640 * 0 on success, or a negative error in case of failure. 1641 * 1642 * int bpf_override_return(struct pt_reg *regs, u64 rc) 1643 * Description 1644 * Used for error injection, this helper uses kprobes to override 1645 * the return value of the probed function, and to set it to *rc*. 1646 * The first argument is the context *regs* on which the kprobe 1647 * works. 1648 * 1649 * This helper works by setting setting the PC (program counter) 1650 * to an override function which is run in place of the original 1651 * probed function. This means the probed function is not run at 1652 * all. The replacement function just returns with the required 1653 * value. 1654 * 1655 * This helper has security implications, and thus is subject to 1656 * restrictions. It is only available if the kernel was compiled 1657 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1658 * option, and in this case it only works on functions tagged with 1659 * **ALLOW_ERROR_INJECTION** in the kernel code. 1660 * 1661 * Also, the helper is only available for the architectures having 1662 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1663 * x86 architecture is the only one to support this feature. 1664 * Return 1665 * 0 1666 * 1667 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1668 * Description 1669 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1670 * for the full TCP socket associated to *bpf_sock_ops* to 1671 * *argval*. 1672 * 1673 * The primary use of this field is to determine if there should 1674 * be calls to eBPF programs of type 1675 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1676 * code. A program of the same type can change its value, per 1677 * connection and as necessary, when the connection is 1678 * established. This field is directly accessible for reading, but 1679 * this helper must be used for updates in order to return an 1680 * error if an eBPF program tries to set a callback that is not 1681 * supported in the current kernel. 1682 * 1683 * The supported callback values that *argval* can combine are: 1684 * 1685 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1686 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1687 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1688 * 1689 * Here are some examples of where one could call such eBPF 1690 * program: 1691 * 1692 * * When RTO fires. 1693 * * When a packet is retransmitted. 1694 * * When the connection terminates. 1695 * * When a packet is sent. 1696 * * When a packet is received. 1697 * Return 1698 * Code **-EINVAL** if the socket is not a full TCP socket; 1699 * otherwise, a positive number containing the bits that could not 1700 * be set is returned (which comes down to 0 if all bits were set 1701 * as required). 1702 * 1703 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1704 * Description 1705 * This helper is used in programs implementing policies at the 1706 * socket level. If the message *msg* is allowed to pass (i.e. if 1707 * the verdict eBPF program returns **SK_PASS**), redirect it to 1708 * the socket referenced by *map* (of type 1709 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1710 * egress interfaces can be used for redirection. The 1711 * **BPF_F_INGRESS** value in *flags* is used to make the 1712 * distinction (ingress path is selected if the flag is present, 1713 * egress path otherwise). This is the only flag supported for now. 1714 * Return 1715 * **SK_PASS** on success, or **SK_DROP** on error. 1716 * 1717 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1718 * Description 1719 * For socket policies, apply the verdict of the eBPF program to 1720 * the next *bytes* (number of bytes) of message *msg*. 1721 * 1722 * For example, this helper can be used in the following cases: 1723 * 1724 * * A single **sendmsg**\ () or **sendfile**\ () system call 1725 * contains multiple logical messages that the eBPF program is 1726 * supposed to read and for which it should apply a verdict. 1727 * * An eBPF program only cares to read the first *bytes* of a 1728 * *msg*. If the message has a large payload, then setting up 1729 * and calling the eBPF program repeatedly for all bytes, even 1730 * though the verdict is already known, would create unnecessary 1731 * overhead. 1732 * 1733 * When called from within an eBPF program, the helper sets a 1734 * counter internal to the BPF infrastructure, that is used to 1735 * apply the last verdict to the next *bytes*. If *bytes* is 1736 * smaller than the current data being processed from a 1737 * **sendmsg**\ () or **sendfile**\ () system call, the first 1738 * *bytes* will be sent and the eBPF program will be re-run with 1739 * the pointer for start of data pointing to byte number *bytes* 1740 * **+ 1**. If *bytes* is larger than the current data being 1741 * processed, then the eBPF verdict will be applied to multiple 1742 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1743 * consumed. 1744 * 1745 * Note that if a socket closes with the internal counter holding 1746 * a non-zero value, this is not a problem because data is not 1747 * being buffered for *bytes* and is sent as it is received. 1748 * Return 1749 * 0 1750 * 1751 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1752 * Description 1753 * For socket policies, prevent the execution of the verdict eBPF 1754 * program for message *msg* until *bytes* (byte number) have been 1755 * accumulated. 1756 * 1757 * This can be used when one needs a specific number of bytes 1758 * before a verdict can be assigned, even if the data spans 1759 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1760 * case would be a user calling **sendmsg**\ () repeatedly with 1761 * 1-byte long message segments. Obviously, this is bad for 1762 * performance, but it is still valid. If the eBPF program needs 1763 * *bytes* bytes to validate a header, this helper can be used to 1764 * prevent the eBPF program to be called again until *bytes* have 1765 * been accumulated. 1766 * Return 1767 * 0 1768 * 1769 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1770 * Description 1771 * For socket policies, pull in non-linear data from user space 1772 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1773 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1774 * respectively. 1775 * 1776 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1777 * *msg* it can only parse data that the (**data**, **data_end**) 1778 * pointers have already consumed. For **sendmsg**\ () hooks this 1779 * is likely the first scatterlist element. But for calls relying 1780 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1781 * be the range (**0**, **0**) because the data is shared with 1782 * user space and by default the objective is to avoid allowing 1783 * user space to modify data while (or after) eBPF verdict is 1784 * being decided. This helper can be used to pull in data and to 1785 * set the start and end pointer to given values. Data will be 1786 * copied if necessary (i.e. if data was not linear and if start 1787 * and end pointers do not point to the same chunk). 1788 * 1789 * A call to this helper is susceptible to change the underlaying 1790 * packet buffer. Therefore, at load time, all checks on pointers 1791 * previously done by the verifier are invalidated and must be 1792 * performed again, if the helper is used in combination with 1793 * direct packet access. 1794 * 1795 * All values for *flags* are reserved for future usage, and must 1796 * be left at zero. 1797 * Return 1798 * 0 on success, or a negative error in case of failure. 1799 * 1800 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1801 * Description 1802 * Bind the socket associated to *ctx* to the address pointed by 1803 * *addr*, of length *addr_len*. This allows for making outgoing 1804 * connection from the desired IP address, which can be useful for 1805 * example when all processes inside a cgroup should use one 1806 * single IP address on a host that has multiple IP configured. 1807 * 1808 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1809 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1810 * **AF_INET6**). Looking for a free port to bind to can be 1811 * expensive, therefore binding to port is not permitted by the 1812 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1813 * must be set to zero. 1814 * Return 1815 * 0 on success, or a negative error in case of failure. 1816 * 1817 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1818 * Description 1819 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1820 * only possible to shrink the packet as of this writing, 1821 * therefore *delta* must be a negative integer. 1822 * 1823 * A call to this helper is susceptible to change the underlaying 1824 * packet buffer. Therefore, at load time, all checks on pointers 1825 * previously done by the verifier are invalidated and must be 1826 * performed again, if the helper is used in combination with 1827 * direct packet access. 1828 * Return 1829 * 0 on success, or a negative error in case of failure. 1830 * 1831 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1832 * Description 1833 * Retrieve the XFRM state (IP transform framework, see also 1834 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1835 * 1836 * The retrieved value is stored in the **struct bpf_xfrm_state** 1837 * pointed by *xfrm_state* and of length *size*. 1838 * 1839 * All values for *flags* are reserved for future usage, and must 1840 * be left at zero. 1841 * 1842 * This helper is available only if the kernel was compiled with 1843 * **CONFIG_XFRM** configuration option. 1844 * Return 1845 * 0 on success, or a negative error in case of failure. 1846 * 1847 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1848 * Description 1849 * Return a user or a kernel stack in bpf program provided buffer. 1850 * To achieve this, the helper needs *ctx*, which is a pointer 1851 * to the context on which the tracing program is executed. 1852 * To store the stacktrace, the bpf program provides *buf* with 1853 * a nonnegative *size*. 1854 * 1855 * The last argument, *flags*, holds the number of stack frames to 1856 * skip (from 0 to 255), masked with 1857 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1858 * the following flags: 1859 * 1860 * **BPF_F_USER_STACK** 1861 * Collect a user space stack instead of a kernel stack. 1862 * **BPF_F_USER_BUILD_ID** 1863 * Collect buildid+offset instead of ips for user stack, 1864 * only valid if **BPF_F_USER_STACK** is also specified. 1865 * 1866 * **bpf_get_stack**\ () can collect up to 1867 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1868 * to sufficient large buffer size. Note that 1869 * this limit can be controlled with the **sysctl** program, and 1870 * that it should be manually increased in order to profile long 1871 * user stacks (such as stacks for Java programs). To do so, use: 1872 * 1873 * :: 1874 * 1875 * # sysctl kernel.perf_event_max_stack=<new value> 1876 * Return 1877 * A non-negative value equal to or less than *size* on success, 1878 * or a negative error in case of failure. 1879 * 1880 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1881 * Description 1882 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1883 * it provides an easy way to load *len* bytes from *offset* 1884 * from the packet associated to *skb*, into the buffer pointed 1885 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1886 * a fifth argument *start_header* exists in order to select a 1887 * base offset to start from. *start_header* can be one of: 1888 * 1889 * **BPF_HDR_START_MAC** 1890 * Base offset to load data from is *skb*'s mac header. 1891 * **BPF_HDR_START_NET** 1892 * Base offset to load data from is *skb*'s network header. 1893 * 1894 * In general, "direct packet access" is the preferred method to 1895 * access packet data, however, this helper is in particular useful 1896 * in socket filters where *skb*\ **->data** does not always point 1897 * to the start of the mac header and where "direct packet access" 1898 * is not available. 1899 * Return 1900 * 0 on success, or a negative error in case of failure. 1901 * 1902 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 1903 * Description 1904 * Do FIB lookup in kernel tables using parameters in *params*. 1905 * If lookup is successful and result shows packet is to be 1906 * forwarded, the neighbor tables are searched for the nexthop. 1907 * If successful (ie., FIB lookup shows forwarding and nexthop 1908 * is resolved), the nexthop address is returned in ipv4_dst 1909 * or ipv6_dst based on family, smac is set to mac address of 1910 * egress device, dmac is set to nexthop mac address, rt_metric 1911 * is set to metric from route (IPv4/IPv6 only), and ifindex 1912 * is set to the device index of the nexthop from the FIB lookup. 1913 * 1914 * *plen* argument is the size of the passed in struct. 1915 * *flags* argument can be a combination of one or more of the 1916 * following values: 1917 * 1918 * **BPF_FIB_LOOKUP_DIRECT** 1919 * Do a direct table lookup vs full lookup using FIB 1920 * rules. 1921 * **BPF_FIB_LOOKUP_OUTPUT** 1922 * Perform lookup from an egress perspective (default is 1923 * ingress). 1924 * 1925 * *ctx* is either **struct xdp_md** for XDP programs or 1926 * **struct sk_buff** tc cls_act programs. 1927 * Return 1928 * * < 0 if any input argument is invalid 1929 * * 0 on success (packet is forwarded, nexthop neighbor exists) 1930 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 1931 * packet is not forwarded or needs assist from full stack 1932 * 1933 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 1934 * Description 1935 * Add an entry to, or update a sockhash *map* referencing sockets. 1936 * The *skops* is used as a new value for the entry associated to 1937 * *key*. *flags* is one of: 1938 * 1939 * **BPF_NOEXIST** 1940 * The entry for *key* must not exist in the map. 1941 * **BPF_EXIST** 1942 * The entry for *key* must already exist in the map. 1943 * **BPF_ANY** 1944 * No condition on the existence of the entry for *key*. 1945 * 1946 * If the *map* has eBPF programs (parser and verdict), those will 1947 * be inherited by the socket being added. If the socket is 1948 * already attached to eBPF programs, this results in an error. 1949 * Return 1950 * 0 on success, or a negative error in case of failure. 1951 * 1952 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 1953 * Description 1954 * This helper is used in programs implementing policies at the 1955 * socket level. If the message *msg* is allowed to pass (i.e. if 1956 * the verdict eBPF program returns **SK_PASS**), redirect it to 1957 * the socket referenced by *map* (of type 1958 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1959 * egress interfaces can be used for redirection. The 1960 * **BPF_F_INGRESS** value in *flags* is used to make the 1961 * distinction (ingress path is selected if the flag is present, 1962 * egress path otherwise). This is the only flag supported for now. 1963 * Return 1964 * **SK_PASS** on success, or **SK_DROP** on error. 1965 * 1966 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 1967 * Description 1968 * This helper is used in programs implementing policies at the 1969 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 1970 * if the verdeict eBPF program returns **SK_PASS**), redirect it 1971 * to the socket referenced by *map* (of type 1972 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1973 * egress interfaces can be used for redirection. The 1974 * **BPF_F_INGRESS** value in *flags* is used to make the 1975 * distinction (ingress path is selected if the flag is present, 1976 * egress otherwise). This is the only flag supported for now. 1977 * Return 1978 * **SK_PASS** on success, or **SK_DROP** on error. 1979 * 1980 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 1981 * Description 1982 * Encapsulate the packet associated to *skb* within a Layer 3 1983 * protocol header. This header is provided in the buffer at 1984 * address *hdr*, with *len* its size in bytes. *type* indicates 1985 * the protocol of the header and can be one of: 1986 * 1987 * **BPF_LWT_ENCAP_SEG6** 1988 * IPv6 encapsulation with Segment Routing Header 1989 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 1990 * the IPv6 header is computed by the kernel. 1991 * **BPF_LWT_ENCAP_SEG6_INLINE** 1992 * Only works if *skb* contains an IPv6 packet. Insert a 1993 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 1994 * the IPv6 header. 1995 * 1996 * A call to this helper is susceptible to change the underlaying 1997 * packet buffer. Therefore, at load time, all checks on pointers 1998 * previously done by the verifier are invalidated and must be 1999 * performed again, if the helper is used in combination with 2000 * direct packet access. 2001 * Return 2002 * 0 on success, or a negative error in case of failure. 2003 * 2004 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2005 * Description 2006 * Store *len* bytes from address *from* into the packet 2007 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2008 * inside the outermost IPv6 Segment Routing Header can be 2009 * modified through this helper. 2010 * 2011 * A call to this helper is susceptible to change the underlaying 2012 * packet buffer. Therefore, at load time, all checks on pointers 2013 * previously done by the verifier are invalidated and must be 2014 * performed again, if the helper is used in combination with 2015 * direct packet access. 2016 * Return 2017 * 0 on success, or a negative error in case of failure. 2018 * 2019 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2020 * Description 2021 * Adjust the size allocated to TLVs in the outermost IPv6 2022 * Segment Routing Header contained in the packet associated to 2023 * *skb*, at position *offset* by *delta* bytes. Only offsets 2024 * after the segments are accepted. *delta* can be as well 2025 * positive (growing) as negative (shrinking). 2026 * 2027 * A call to this helper is susceptible to change the underlaying 2028 * packet buffer. Therefore, at load time, all checks on pointers 2029 * previously done by the verifier are invalidated and must be 2030 * performed again, if the helper is used in combination with 2031 * direct packet access. 2032 * Return 2033 * 0 on success, or a negative error in case of failure. 2034 * 2035 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2036 * Description 2037 * Apply an IPv6 Segment Routing action of type *action* to the 2038 * packet associated to *skb*. Each action takes a parameter 2039 * contained at address *param*, and of length *param_len* bytes. 2040 * *action* can be one of: 2041 * 2042 * **SEG6_LOCAL_ACTION_END_X** 2043 * End.X action: Endpoint with Layer-3 cross-connect. 2044 * Type of *param*: **struct in6_addr**. 2045 * **SEG6_LOCAL_ACTION_END_T** 2046 * End.T action: Endpoint with specific IPv6 table lookup. 2047 * Type of *param*: **int**. 2048 * **SEG6_LOCAL_ACTION_END_B6** 2049 * End.B6 action: Endpoint bound to an SRv6 policy. 2050 * Type of param: **struct ipv6_sr_hdr**. 2051 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2052 * End.B6.Encap action: Endpoint bound to an SRv6 2053 * encapsulation policy. 2054 * Type of param: **struct ipv6_sr_hdr**. 2055 * 2056 * A call to this helper is susceptible to change the underlaying 2057 * packet buffer. Therefore, at load time, all checks on pointers 2058 * previously done by the verifier are invalidated and must be 2059 * performed again, if the helper is used in combination with 2060 * direct packet access. 2061 * Return 2062 * 0 on success, or a negative error in case of failure. 2063 * 2064 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2065 * Description 2066 * This helper is used in programs implementing IR decoding, to 2067 * report a successfully decoded key press with *scancode*, 2068 * *toggle* value in the given *protocol*. The scancode will be 2069 * translated to a keycode using the rc keymap, and reported as 2070 * an input key down event. After a period a key up event is 2071 * generated. This period can be extended by calling either 2072 * **bpf_rc_keydown** () again with the same values, or calling 2073 * **bpf_rc_repeat** (). 2074 * 2075 * Some protocols include a toggle bit, in case the button was 2076 * released and pressed again between consecutive scancodes. 2077 * 2078 * The *ctx* should point to the lirc sample as passed into 2079 * the program. 2080 * 2081 * The *protocol* is the decoded protocol number (see 2082 * **enum rc_proto** for some predefined values). 2083 * 2084 * This helper is only available is the kernel was compiled with 2085 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2086 * "**y**". 2087 * Return 2088 * 0 2089 * 2090 * int bpf_rc_repeat(void *ctx) 2091 * Description 2092 * This helper is used in programs implementing IR decoding, to 2093 * report a successfully decoded repeat key message. This delays 2094 * the generation of a key up event for previously generated 2095 * key down event. 2096 * 2097 * Some IR protocols like NEC have a special IR message for 2098 * repeating last button, for when a button is held down. 2099 * 2100 * The *ctx* should point to the lirc sample as passed into 2101 * the program. 2102 * 2103 * This helper is only available is the kernel was compiled with 2104 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2105 * "**y**". 2106 * Return 2107 * 0 2108 * 2109 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb) 2110 * Description 2111 * Return the cgroup v2 id of the socket associated with the *skb*. 2112 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2113 * helper for cgroup v1 by providing a tag resp. identifier that 2114 * can be matched on or used for map lookups e.g. to implement 2115 * policy. The cgroup v2 id of a given path in the hierarchy is 2116 * exposed in user space through the f_handle API in order to get 2117 * to the same 64-bit id. 2118 * 2119 * This helper can be used on TC egress path, but not on ingress, 2120 * and is available only if the kernel was compiled with the 2121 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2122 * Return 2123 * The id is returned or 0 in case the id could not be retrieved. 2124 * 2125 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2126 * Description 2127 * Return id of cgroup v2 that is ancestor of cgroup associated 2128 * with the *skb* at the *ancestor_level*. The root cgroup is at 2129 * *ancestor_level* zero and each step down the hierarchy 2130 * increments the level. If *ancestor_level* == level of cgroup 2131 * associated with *skb*, then return value will be same as that 2132 * of **bpf_skb_cgroup_id**\ (). 2133 * 2134 * The helper is useful to implement policies based on cgroups 2135 * that are upper in hierarchy than immediate cgroup associated 2136 * with *skb*. 2137 * 2138 * The format of returned id and helper limitations are same as in 2139 * **bpf_skb_cgroup_id**\ (). 2140 * Return 2141 * The id is returned or 0 in case the id could not be retrieved. 2142 * 2143 * u64 bpf_get_current_cgroup_id(void) 2144 * Return 2145 * A 64-bit integer containing the current cgroup id based 2146 * on the cgroup within which the current task is running. 2147 * 2148 * void* get_local_storage(void *map, u64 flags) 2149 * Description 2150 * Get the pointer to the local storage area. 2151 * The type and the size of the local storage is defined 2152 * by the *map* argument. 2153 * The *flags* meaning is specific for each map type, 2154 * and has to be 0 for cgroup local storage. 2155 * 2156 * Depending on the bpf program type, a local storage area 2157 * can be shared between multiple instances of the bpf program, 2158 * running simultaneously. 2159 * 2160 * A user should care about the synchronization by himself. 2161 * For example, by using the BPF_STX_XADD instruction to alter 2162 * the shared data. 2163 * Return 2164 * Pointer to the local storage area. 2165 * 2166 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2167 * Description 2168 * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map 2169 * It checks the selected sk is matching the incoming 2170 * request in the skb. 2171 * Return 2172 * 0 on success, or a negative error in case of failure. 2173 * 2174 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags) 2175 * Description 2176 * Look for TCP socket matching *tuple*, optionally in a child 2177 * network namespace *netns*. The return value must be checked, 2178 * and if non-NULL, released via **bpf_sk_release**\ (). 2179 * 2180 * The *ctx* should point to the context of the program, such as 2181 * the skb or socket (depending on the hook in use). This is used 2182 * to determine the base network namespace for the lookup. 2183 * 2184 * *tuple_size* must be one of: 2185 * 2186 * **sizeof**\ (*tuple*\ **->ipv4**) 2187 * Look for an IPv4 socket. 2188 * **sizeof**\ (*tuple*\ **->ipv6**) 2189 * Look for an IPv6 socket. 2190 * 2191 * If the *netns* is zero, then the socket lookup table in the 2192 * netns associated with the *ctx* will be used. For the TC hooks, 2193 * this in the netns of the device in the skb. For socket hooks, 2194 * this in the netns of the socket. If *netns* is non-zero, then 2195 * it specifies the ID of the netns relative to the netns 2196 * associated with the *ctx*. 2197 * 2198 * All values for *flags* are reserved for future usage, and must 2199 * be left at zero. 2200 * 2201 * This helper is available only if the kernel was compiled with 2202 * **CONFIG_NET** configuration option. 2203 * Return 2204 * Pointer to *struct bpf_sock*, or NULL in case of failure. 2205 * 2206 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags) 2207 * Description 2208 * Look for UDP socket matching *tuple*, optionally in a child 2209 * network namespace *netns*. The return value must be checked, 2210 * and if non-NULL, released via **bpf_sk_release**\ (). 2211 * 2212 * The *ctx* should point to the context of the program, such as 2213 * the skb or socket (depending on the hook in use). This is used 2214 * to determine the base network namespace for the lookup. 2215 * 2216 * *tuple_size* must be one of: 2217 * 2218 * **sizeof**\ (*tuple*\ **->ipv4**) 2219 * Look for an IPv4 socket. 2220 * **sizeof**\ (*tuple*\ **->ipv6**) 2221 * Look for an IPv6 socket. 2222 * 2223 * If the *netns* is zero, then the socket lookup table in the 2224 * netns associated with the *ctx* will be used. For the TC hooks, 2225 * this in the netns of the device in the skb. For socket hooks, 2226 * this in the netns of the socket. If *netns* is non-zero, then 2227 * it specifies the ID of the netns relative to the netns 2228 * associated with the *ctx*. 2229 * 2230 * All values for *flags* are reserved for future usage, and must 2231 * be left at zero. 2232 * 2233 * This helper is available only if the kernel was compiled with 2234 * **CONFIG_NET** configuration option. 2235 * Return 2236 * Pointer to *struct bpf_sock*, or NULL in case of failure. 2237 * 2238 * int bpf_sk_release(struct bpf_sock *sk) 2239 * Description 2240 * Release the reference held by *sock*. *sock* must be a non-NULL 2241 * pointer that was returned from bpf_sk_lookup_xxx\ (). 2242 * Return 2243 * 0 on success, or a negative error in case of failure. 2244 * 2245 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 2246 * Description 2247 * For socket policies, insert *len* bytes into msg at offset 2248 * *start*. 2249 * 2250 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2251 * *msg* it may want to insert metadata or options into the msg. 2252 * This can later be read and used by any of the lower layer BPF 2253 * hooks. 2254 * 2255 * This helper may fail if under memory pressure (a malloc 2256 * fails) in these cases BPF programs will get an appropriate 2257 * error and BPF programs will need to handle them. 2258 * 2259 * Return 2260 * 0 on success, or a negative error in case of failure. 2261 */ 2262 #define __BPF_FUNC_MAPPER(FN) \ 2263 FN(unspec), \ 2264 FN(map_lookup_elem), \ 2265 FN(map_update_elem), \ 2266 FN(map_delete_elem), \ 2267 FN(probe_read), \ 2268 FN(ktime_get_ns), \ 2269 FN(trace_printk), \ 2270 FN(get_prandom_u32), \ 2271 FN(get_smp_processor_id), \ 2272 FN(skb_store_bytes), \ 2273 FN(l3_csum_replace), \ 2274 FN(l4_csum_replace), \ 2275 FN(tail_call), \ 2276 FN(clone_redirect), \ 2277 FN(get_current_pid_tgid), \ 2278 FN(get_current_uid_gid), \ 2279 FN(get_current_comm), \ 2280 FN(get_cgroup_classid), \ 2281 FN(skb_vlan_push), \ 2282 FN(skb_vlan_pop), \ 2283 FN(skb_get_tunnel_key), \ 2284 FN(skb_set_tunnel_key), \ 2285 FN(perf_event_read), \ 2286 FN(redirect), \ 2287 FN(get_route_realm), \ 2288 FN(perf_event_output), \ 2289 FN(skb_load_bytes), \ 2290 FN(get_stackid), \ 2291 FN(csum_diff), \ 2292 FN(skb_get_tunnel_opt), \ 2293 FN(skb_set_tunnel_opt), \ 2294 FN(skb_change_proto), \ 2295 FN(skb_change_type), \ 2296 FN(skb_under_cgroup), \ 2297 FN(get_hash_recalc), \ 2298 FN(get_current_task), \ 2299 FN(probe_write_user), \ 2300 FN(current_task_under_cgroup), \ 2301 FN(skb_change_tail), \ 2302 FN(skb_pull_data), \ 2303 FN(csum_update), \ 2304 FN(set_hash_invalid), \ 2305 FN(get_numa_node_id), \ 2306 FN(skb_change_head), \ 2307 FN(xdp_adjust_head), \ 2308 FN(probe_read_str), \ 2309 FN(get_socket_cookie), \ 2310 FN(get_socket_uid), \ 2311 FN(set_hash), \ 2312 FN(setsockopt), \ 2313 FN(skb_adjust_room), \ 2314 FN(redirect_map), \ 2315 FN(sk_redirect_map), \ 2316 FN(sock_map_update), \ 2317 FN(xdp_adjust_meta), \ 2318 FN(perf_event_read_value), \ 2319 FN(perf_prog_read_value), \ 2320 FN(getsockopt), \ 2321 FN(override_return), \ 2322 FN(sock_ops_cb_flags_set), \ 2323 FN(msg_redirect_map), \ 2324 FN(msg_apply_bytes), \ 2325 FN(msg_cork_bytes), \ 2326 FN(msg_pull_data), \ 2327 FN(bind), \ 2328 FN(xdp_adjust_tail), \ 2329 FN(skb_get_xfrm_state), \ 2330 FN(get_stack), \ 2331 FN(skb_load_bytes_relative), \ 2332 FN(fib_lookup), \ 2333 FN(sock_hash_update), \ 2334 FN(msg_redirect_hash), \ 2335 FN(sk_redirect_hash), \ 2336 FN(lwt_push_encap), \ 2337 FN(lwt_seg6_store_bytes), \ 2338 FN(lwt_seg6_adjust_srh), \ 2339 FN(lwt_seg6_action), \ 2340 FN(rc_repeat), \ 2341 FN(rc_keydown), \ 2342 FN(skb_cgroup_id), \ 2343 FN(get_current_cgroup_id), \ 2344 FN(get_local_storage), \ 2345 FN(sk_select_reuseport), \ 2346 FN(skb_ancestor_cgroup_id), \ 2347 FN(sk_lookup_tcp), \ 2348 FN(sk_lookup_udp), \ 2349 FN(sk_release), \ 2350 FN(map_push_elem), \ 2351 FN(map_pop_elem), \ 2352 FN(map_peek_elem), \ 2353 FN(msg_push_data), 2354 2355 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2356 * function eBPF program intends to call 2357 */ 2358 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2359 enum bpf_func_id { 2360 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 2361 __BPF_FUNC_MAX_ID, 2362 }; 2363 #undef __BPF_ENUM_FN 2364 2365 /* All flags used by eBPF helper functions, placed here. */ 2366 2367 /* BPF_FUNC_skb_store_bytes flags. */ 2368 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 2369 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2370 2371 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2372 * First 4 bits are for passing the header field size. 2373 */ 2374 #define BPF_F_HDR_FIELD_MASK 0xfULL 2375 2376 /* BPF_FUNC_l4_csum_replace flags. */ 2377 #define BPF_F_PSEUDO_HDR (1ULL << 4) 2378 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2379 #define BPF_F_MARK_ENFORCE (1ULL << 6) 2380 2381 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2382 #define BPF_F_INGRESS (1ULL << 0) 2383 2384 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2385 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2386 2387 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2388 #define BPF_F_SKIP_FIELD_MASK 0xffULL 2389 #define BPF_F_USER_STACK (1ULL << 8) 2390 /* flags used by BPF_FUNC_get_stackid only. */ 2391 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2392 #define BPF_F_REUSE_STACKID (1ULL << 10) 2393 /* flags used by BPF_FUNC_get_stack only. */ 2394 #define BPF_F_USER_BUILD_ID (1ULL << 11) 2395 2396 /* BPF_FUNC_skb_set_tunnel_key flags. */ 2397 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 2398 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 2399 #define BPF_F_SEQ_NUMBER (1ULL << 3) 2400 2401 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2402 * BPF_FUNC_perf_event_read_value flags. 2403 */ 2404 #define BPF_F_INDEX_MASK 0xffffffffULL 2405 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2406 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2407 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 2408 2409 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 2410 enum bpf_adj_room_mode { 2411 BPF_ADJ_ROOM_NET, 2412 }; 2413 2414 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 2415 enum bpf_hdr_start_off { 2416 BPF_HDR_START_MAC, 2417 BPF_HDR_START_NET, 2418 }; 2419 2420 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2421 enum bpf_lwt_encap_mode { 2422 BPF_LWT_ENCAP_SEG6, 2423 BPF_LWT_ENCAP_SEG6_INLINE 2424 }; 2425 2426 /* user accessible mirror of in-kernel sk_buff. 2427 * new fields can only be added to the end of this structure 2428 */ 2429 struct __sk_buff { 2430 __u32 len; 2431 __u32 pkt_type; 2432 __u32 mark; 2433 __u32 queue_mapping; 2434 __u32 protocol; 2435 __u32 vlan_present; 2436 __u32 vlan_tci; 2437 __u32 vlan_proto; 2438 __u32 priority; 2439 __u32 ingress_ifindex; 2440 __u32 ifindex; 2441 __u32 tc_index; 2442 __u32 cb[5]; 2443 __u32 hash; 2444 __u32 tc_classid; 2445 __u32 data; 2446 __u32 data_end; 2447 __u32 napi_id; 2448 2449 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 2450 __u32 family; 2451 __u32 remote_ip4; /* Stored in network byte order */ 2452 __u32 local_ip4; /* Stored in network byte order */ 2453 __u32 remote_ip6[4]; /* Stored in network byte order */ 2454 __u32 local_ip6[4]; /* Stored in network byte order */ 2455 __u32 remote_port; /* Stored in network byte order */ 2456 __u32 local_port; /* stored in host byte order */ 2457 /* ... here. */ 2458 2459 __u32 data_meta; 2460 struct bpf_flow_keys *flow_keys; 2461 }; 2462 2463 struct bpf_tunnel_key { 2464 __u32 tunnel_id; 2465 union { 2466 __u32 remote_ipv4; 2467 __u32 remote_ipv6[4]; 2468 }; 2469 __u8 tunnel_tos; 2470 __u8 tunnel_ttl; 2471 __u16 tunnel_ext; /* Padding, future use. */ 2472 __u32 tunnel_label; 2473 }; 2474 2475 /* user accessible mirror of in-kernel xfrm_state. 2476 * new fields can only be added to the end of this structure 2477 */ 2478 struct bpf_xfrm_state { 2479 __u32 reqid; 2480 __u32 spi; /* Stored in network byte order */ 2481 __u16 family; 2482 __u16 ext; /* Padding, future use. */ 2483 union { 2484 __u32 remote_ipv4; /* Stored in network byte order */ 2485 __u32 remote_ipv6[4]; /* Stored in network byte order */ 2486 }; 2487 }; 2488 2489 /* Generic BPF return codes which all BPF program types may support. 2490 * The values are binary compatible with their TC_ACT_* counter-part to 2491 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 2492 * programs. 2493 * 2494 * XDP is handled seprately, see XDP_*. 2495 */ 2496 enum bpf_ret_code { 2497 BPF_OK = 0, 2498 /* 1 reserved */ 2499 BPF_DROP = 2, 2500 /* 3-6 reserved */ 2501 BPF_REDIRECT = 7, 2502 /* >127 are reserved for prog type specific return codes */ 2503 }; 2504 2505 struct bpf_sock { 2506 __u32 bound_dev_if; 2507 __u32 family; 2508 __u32 type; 2509 __u32 protocol; 2510 __u32 mark; 2511 __u32 priority; 2512 __u32 src_ip4; /* Allows 1,2,4-byte read. 2513 * Stored in network byte order. 2514 */ 2515 __u32 src_ip6[4]; /* Allows 1,2,4-byte read. 2516 * Stored in network byte order. 2517 */ 2518 __u32 src_port; /* Allows 4-byte read. 2519 * Stored in host byte order 2520 */ 2521 }; 2522 2523 struct bpf_sock_tuple { 2524 union { 2525 struct { 2526 __be32 saddr; 2527 __be32 daddr; 2528 __be16 sport; 2529 __be16 dport; 2530 } ipv4; 2531 struct { 2532 __be32 saddr[4]; 2533 __be32 daddr[4]; 2534 __be16 sport; 2535 __be16 dport; 2536 } ipv6; 2537 }; 2538 }; 2539 2540 #define XDP_PACKET_HEADROOM 256 2541 2542 /* User return codes for XDP prog type. 2543 * A valid XDP program must return one of these defined values. All other 2544 * return codes are reserved for future use. Unknown return codes will 2545 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 2546 */ 2547 enum xdp_action { 2548 XDP_ABORTED = 0, 2549 XDP_DROP, 2550 XDP_PASS, 2551 XDP_TX, 2552 XDP_REDIRECT, 2553 }; 2554 2555 /* user accessible metadata for XDP packet hook 2556 * new fields must be added to the end of this structure 2557 */ 2558 struct xdp_md { 2559 __u32 data; 2560 __u32 data_end; 2561 __u32 data_meta; 2562 /* Below access go through struct xdp_rxq_info */ 2563 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 2564 __u32 rx_queue_index; /* rxq->queue_index */ 2565 }; 2566 2567 enum sk_action { 2568 SK_DROP = 0, 2569 SK_PASS, 2570 }; 2571 2572 /* user accessible metadata for SK_MSG packet hook, new fields must 2573 * be added to the end of this structure 2574 */ 2575 struct sk_msg_md { 2576 void *data; 2577 void *data_end; 2578 2579 __u32 family; 2580 __u32 remote_ip4; /* Stored in network byte order */ 2581 __u32 local_ip4; /* Stored in network byte order */ 2582 __u32 remote_ip6[4]; /* Stored in network byte order */ 2583 __u32 local_ip6[4]; /* Stored in network byte order */ 2584 __u32 remote_port; /* Stored in network byte order */ 2585 __u32 local_port; /* stored in host byte order */ 2586 }; 2587 2588 struct sk_reuseport_md { 2589 /* 2590 * Start of directly accessible data. It begins from 2591 * the tcp/udp header. 2592 */ 2593 void *data; 2594 void *data_end; /* End of directly accessible data */ 2595 /* 2596 * Total length of packet (starting from the tcp/udp header). 2597 * Note that the directly accessible bytes (data_end - data) 2598 * could be less than this "len". Those bytes could be 2599 * indirectly read by a helper "bpf_skb_load_bytes()". 2600 */ 2601 __u32 len; 2602 /* 2603 * Eth protocol in the mac header (network byte order). e.g. 2604 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 2605 */ 2606 __u32 eth_protocol; 2607 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 2608 __u32 bind_inany; /* Is sock bound to an INANY address? */ 2609 __u32 hash; /* A hash of the packet 4 tuples */ 2610 }; 2611 2612 #define BPF_TAG_SIZE 8 2613 2614 struct bpf_prog_info { 2615 __u32 type; 2616 __u32 id; 2617 __u8 tag[BPF_TAG_SIZE]; 2618 __u32 jited_prog_len; 2619 __u32 xlated_prog_len; 2620 __aligned_u64 jited_prog_insns; 2621 __aligned_u64 xlated_prog_insns; 2622 __u64 load_time; /* ns since boottime */ 2623 __u32 created_by_uid; 2624 __u32 nr_map_ids; 2625 __aligned_u64 map_ids; 2626 char name[BPF_OBJ_NAME_LEN]; 2627 __u32 ifindex; 2628 __u32 gpl_compatible:1; 2629 __u64 netns_dev; 2630 __u64 netns_ino; 2631 __u32 nr_jited_ksyms; 2632 __u32 nr_jited_func_lens; 2633 __aligned_u64 jited_ksyms; 2634 __aligned_u64 jited_func_lens; 2635 } __attribute__((aligned(8))); 2636 2637 struct bpf_map_info { 2638 __u32 type; 2639 __u32 id; 2640 __u32 key_size; 2641 __u32 value_size; 2642 __u32 max_entries; 2643 __u32 map_flags; 2644 char name[BPF_OBJ_NAME_LEN]; 2645 __u32 ifindex; 2646 __u32 :32; 2647 __u64 netns_dev; 2648 __u64 netns_ino; 2649 __u32 btf_id; 2650 __u32 btf_key_type_id; 2651 __u32 btf_value_type_id; 2652 } __attribute__((aligned(8))); 2653 2654 struct bpf_btf_info { 2655 __aligned_u64 btf; 2656 __u32 btf_size; 2657 __u32 id; 2658 } __attribute__((aligned(8))); 2659 2660 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 2661 * by user and intended to be used by socket (e.g. to bind to, depends on 2662 * attach attach type). 2663 */ 2664 struct bpf_sock_addr { 2665 __u32 user_family; /* Allows 4-byte read, but no write. */ 2666 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 2667 * Stored in network byte order. 2668 */ 2669 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 2670 * Stored in network byte order. 2671 */ 2672 __u32 user_port; /* Allows 4-byte read and write. 2673 * Stored in network byte order 2674 */ 2675 __u32 family; /* Allows 4-byte read, but no write */ 2676 __u32 type; /* Allows 4-byte read, but no write */ 2677 __u32 protocol; /* Allows 4-byte read, but no write */ 2678 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write. 2679 * Stored in network byte order. 2680 */ 2681 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 2682 * Stored in network byte order. 2683 */ 2684 }; 2685 2686 /* User bpf_sock_ops struct to access socket values and specify request ops 2687 * and their replies. 2688 * Some of this fields are in network (bigendian) byte order and may need 2689 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 2690 * New fields can only be added at the end of this structure 2691 */ 2692 struct bpf_sock_ops { 2693 __u32 op; 2694 union { 2695 __u32 args[4]; /* Optionally passed to bpf program */ 2696 __u32 reply; /* Returned by bpf program */ 2697 __u32 replylong[4]; /* Optionally returned by bpf prog */ 2698 }; 2699 __u32 family; 2700 __u32 remote_ip4; /* Stored in network byte order */ 2701 __u32 local_ip4; /* Stored in network byte order */ 2702 __u32 remote_ip6[4]; /* Stored in network byte order */ 2703 __u32 local_ip6[4]; /* Stored in network byte order */ 2704 __u32 remote_port; /* Stored in network byte order */ 2705 __u32 local_port; /* stored in host byte order */ 2706 __u32 is_fullsock; /* Some TCP fields are only valid if 2707 * there is a full socket. If not, the 2708 * fields read as zero. 2709 */ 2710 __u32 snd_cwnd; 2711 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 2712 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 2713 __u32 state; 2714 __u32 rtt_min; 2715 __u32 snd_ssthresh; 2716 __u32 rcv_nxt; 2717 __u32 snd_nxt; 2718 __u32 snd_una; 2719 __u32 mss_cache; 2720 __u32 ecn_flags; 2721 __u32 rate_delivered; 2722 __u32 rate_interval_us; 2723 __u32 packets_out; 2724 __u32 retrans_out; 2725 __u32 total_retrans; 2726 __u32 segs_in; 2727 __u32 data_segs_in; 2728 __u32 segs_out; 2729 __u32 data_segs_out; 2730 __u32 lost_out; 2731 __u32 sacked_out; 2732 __u32 sk_txhash; 2733 __u64 bytes_received; 2734 __u64 bytes_acked; 2735 }; 2736 2737 /* Definitions for bpf_sock_ops_cb_flags */ 2738 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 2739 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 2740 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 2741 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 2742 * supported cb flags 2743 */ 2744 2745 /* List of known BPF sock_ops operators. 2746 * New entries can only be added at the end 2747 */ 2748 enum { 2749 BPF_SOCK_OPS_VOID, 2750 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 2751 * -1 if default value should be used 2752 */ 2753 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 2754 * window (in packets) or -1 if default 2755 * value should be used 2756 */ 2757 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 2758 * active connection is initialized 2759 */ 2760 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 2761 * active connection is 2762 * established 2763 */ 2764 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 2765 * passive connection is 2766 * established 2767 */ 2768 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 2769 * needs ECN 2770 */ 2771 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 2772 * based on the path and may be 2773 * dependent on the congestion control 2774 * algorithm. In general it indicates 2775 * a congestion threshold. RTTs above 2776 * this indicate congestion 2777 */ 2778 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 2779 * Arg1: value of icsk_retransmits 2780 * Arg2: value of icsk_rto 2781 * Arg3: whether RTO has expired 2782 */ 2783 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 2784 * Arg1: sequence number of 1st byte 2785 * Arg2: # segments 2786 * Arg3: return value of 2787 * tcp_transmit_skb (0 => success) 2788 */ 2789 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 2790 * Arg1: old_state 2791 * Arg2: new_state 2792 */ 2793 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 2794 * socket transition to LISTEN state. 2795 */ 2796 }; 2797 2798 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 2799 * changes between the TCP and BPF versions. Ideally this should never happen. 2800 * If it does, we need to add code to convert them before calling 2801 * the BPF sock_ops function. 2802 */ 2803 enum { 2804 BPF_TCP_ESTABLISHED = 1, 2805 BPF_TCP_SYN_SENT, 2806 BPF_TCP_SYN_RECV, 2807 BPF_TCP_FIN_WAIT1, 2808 BPF_TCP_FIN_WAIT2, 2809 BPF_TCP_TIME_WAIT, 2810 BPF_TCP_CLOSE, 2811 BPF_TCP_CLOSE_WAIT, 2812 BPF_TCP_LAST_ACK, 2813 BPF_TCP_LISTEN, 2814 BPF_TCP_CLOSING, /* Now a valid state */ 2815 BPF_TCP_NEW_SYN_RECV, 2816 2817 BPF_TCP_MAX_STATES /* Leave at the end! */ 2818 }; 2819 2820 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 2821 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 2822 2823 struct bpf_perf_event_value { 2824 __u64 counter; 2825 __u64 enabled; 2826 __u64 running; 2827 }; 2828 2829 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 2830 #define BPF_DEVCG_ACC_READ (1ULL << 1) 2831 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 2832 2833 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 2834 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 2835 2836 struct bpf_cgroup_dev_ctx { 2837 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 2838 __u32 access_type; 2839 __u32 major; 2840 __u32 minor; 2841 }; 2842 2843 struct bpf_raw_tracepoint_args { 2844 __u64 args[0]; 2845 }; 2846 2847 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 2848 * OUTPUT: Do lookup from egress perspective; default is ingress 2849 */ 2850 #define BPF_FIB_LOOKUP_DIRECT BIT(0) 2851 #define BPF_FIB_LOOKUP_OUTPUT BIT(1) 2852 2853 enum { 2854 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 2855 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 2856 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 2857 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 2858 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 2859 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 2860 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 2861 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 2862 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 2863 }; 2864 2865 struct bpf_fib_lookup { 2866 /* input: network family for lookup (AF_INET, AF_INET6) 2867 * output: network family of egress nexthop 2868 */ 2869 __u8 family; 2870 2871 /* set if lookup is to consider L4 data - e.g., FIB rules */ 2872 __u8 l4_protocol; 2873 __be16 sport; 2874 __be16 dport; 2875 2876 /* total length of packet from network header - used for MTU check */ 2877 __u16 tot_len; 2878 2879 /* input: L3 device index for lookup 2880 * output: device index from FIB lookup 2881 */ 2882 __u32 ifindex; 2883 2884 union { 2885 /* inputs to lookup */ 2886 __u8 tos; /* AF_INET */ 2887 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 2888 2889 /* output: metric of fib result (IPv4/IPv6 only) */ 2890 __u32 rt_metric; 2891 }; 2892 2893 union { 2894 __be32 ipv4_src; 2895 __u32 ipv6_src[4]; /* in6_addr; network order */ 2896 }; 2897 2898 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 2899 * network header. output: bpf_fib_lookup sets to gateway address 2900 * if FIB lookup returns gateway route 2901 */ 2902 union { 2903 __be32 ipv4_dst; 2904 __u32 ipv6_dst[4]; /* in6_addr; network order */ 2905 }; 2906 2907 /* output */ 2908 __be16 h_vlan_proto; 2909 __be16 h_vlan_TCI; 2910 __u8 smac[6]; /* ETH_ALEN */ 2911 __u8 dmac[6]; /* ETH_ALEN */ 2912 }; 2913 2914 enum bpf_task_fd_type { 2915 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 2916 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 2917 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 2918 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 2919 BPF_FD_TYPE_UPROBE, /* filename + offset */ 2920 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 2921 }; 2922 2923 struct bpf_flow_keys { 2924 __u16 nhoff; 2925 __u16 thoff; 2926 __u16 addr_proto; /* ETH_P_* of valid addrs */ 2927 __u8 is_frag; 2928 __u8 is_first_frag; 2929 __u8 is_encap; 2930 __u8 ip_proto; 2931 __be16 n_proto; 2932 __be16 sport; 2933 __be16 dport; 2934 union { 2935 struct { 2936 __be32 ipv4_src; 2937 __be32 ipv4_dst; 2938 }; 2939 struct { 2940 __u32 ipv6_src[4]; /* in6_addr; network order */ 2941 __u32 ipv6_dst[4]; /* in6_addr; network order */ 2942 }; 2943 }; 2944 }; 2945 2946 #endif /* _UAPI__LINUX_BPF_H__ */ 2947 )********" 2948