1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_ALU64 0x07 /* alu mode in double word width */ 18 19 /* ld/ldx fields */ 20 #define BPF_DW 0x18 /* double word (64-bit) */ 21 #define BPF_XADD 0xc0 /* exclusive add */ 22 23 /* alu/jmp fields */ 24 #define BPF_MOV 0xb0 /* mov reg to reg */ 25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 26 27 /* change endianness of a register */ 28 #define BPF_END 0xd0 /* flags for endianness conversion: */ 29 #define BPF_TO_LE 0x00 /* convert to little-endian */ 30 #define BPF_TO_BE 0x08 /* convert to big-endian */ 31 #define BPF_FROM_LE BPF_TO_LE 32 #define BPF_FROM_BE BPF_TO_BE 33 34 /* jmp encodings */ 35 #define BPF_JNE 0x50 /* jump != */ 36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 42 #define BPF_CALL 0x80 /* function call */ 43 #define BPF_EXIT 0x90 /* function return */ 44 45 /* Register numbers */ 46 enum { 47 BPF_REG_0 = 0, 48 BPF_REG_1, 49 BPF_REG_2, 50 BPF_REG_3, 51 BPF_REG_4, 52 BPF_REG_5, 53 BPF_REG_6, 54 BPF_REG_7, 55 BPF_REG_8, 56 BPF_REG_9, 57 BPF_REG_10, 58 __MAX_BPF_REG, 59 }; 60 61 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 62 #define MAX_BPF_REG __MAX_BPF_REG 63 64 struct bpf_insn { 65 __u8 code; /* opcode */ 66 __u8 dst_reg:4; /* dest register */ 67 __u8 src_reg:4; /* source register */ 68 __s16 off; /* signed offset */ 69 __s32 imm; /* signed immediate constant */ 70 }; 71 72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 73 struct bpf_lpm_trie_key { 74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 75 __u8 data[0]; /* Arbitrary size */ 76 }; 77 78 /* BPF syscall commands, see bpf(2) man-page for details. */ 79 enum bpf_cmd { 80 BPF_MAP_CREATE, 81 BPF_MAP_LOOKUP_ELEM, 82 BPF_MAP_UPDATE_ELEM, 83 BPF_MAP_DELETE_ELEM, 84 BPF_MAP_GET_NEXT_KEY, 85 BPF_PROG_LOAD, 86 BPF_OBJ_PIN, 87 BPF_OBJ_GET, 88 BPF_PROG_ATTACH, 89 BPF_PROG_DETACH, 90 BPF_PROG_TEST_RUN, 91 BPF_PROG_GET_NEXT_ID, 92 BPF_MAP_GET_NEXT_ID, 93 BPF_PROG_GET_FD_BY_ID, 94 BPF_MAP_GET_FD_BY_ID, 95 BPF_OBJ_GET_INFO_BY_FD, 96 BPF_PROG_QUERY, 97 BPF_RAW_TRACEPOINT_OPEN, 98 BPF_BTF_LOAD, 99 BPF_BTF_GET_FD_BY_ID, 100 }; 101 102 enum bpf_map_type { 103 BPF_MAP_TYPE_UNSPEC, 104 BPF_MAP_TYPE_HASH, 105 BPF_MAP_TYPE_ARRAY, 106 BPF_MAP_TYPE_PROG_ARRAY, 107 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 108 BPF_MAP_TYPE_PERCPU_HASH, 109 BPF_MAP_TYPE_PERCPU_ARRAY, 110 BPF_MAP_TYPE_STACK_TRACE, 111 BPF_MAP_TYPE_CGROUP_ARRAY, 112 BPF_MAP_TYPE_LRU_HASH, 113 BPF_MAP_TYPE_LRU_PERCPU_HASH, 114 BPF_MAP_TYPE_LPM_TRIE, 115 BPF_MAP_TYPE_ARRAY_OF_MAPS, 116 BPF_MAP_TYPE_HASH_OF_MAPS, 117 BPF_MAP_TYPE_DEVMAP, 118 BPF_MAP_TYPE_SOCKMAP, 119 BPF_MAP_TYPE_CPUMAP, 120 BPF_MAP_TYPE_XSKMAP, 121 }; 122 123 enum bpf_prog_type { 124 BPF_PROG_TYPE_UNSPEC, 125 BPF_PROG_TYPE_SOCKET_FILTER, 126 BPF_PROG_TYPE_KPROBE, 127 BPF_PROG_TYPE_SCHED_CLS, 128 BPF_PROG_TYPE_SCHED_ACT, 129 BPF_PROG_TYPE_TRACEPOINT, 130 BPF_PROG_TYPE_XDP, 131 BPF_PROG_TYPE_PERF_EVENT, 132 BPF_PROG_TYPE_CGROUP_SKB, 133 BPF_PROG_TYPE_CGROUP_SOCK, 134 BPF_PROG_TYPE_LWT_IN, 135 BPF_PROG_TYPE_LWT_OUT, 136 BPF_PROG_TYPE_LWT_XMIT, 137 BPF_PROG_TYPE_SOCK_OPS, 138 BPF_PROG_TYPE_SK_SKB, 139 BPF_PROG_TYPE_CGROUP_DEVICE, 140 BPF_PROG_TYPE_SK_MSG, 141 BPF_PROG_TYPE_RAW_TRACEPOINT, 142 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 143 }; 144 145 enum bpf_attach_type { 146 BPF_CGROUP_INET_INGRESS, 147 BPF_CGROUP_INET_EGRESS, 148 BPF_CGROUP_INET_SOCK_CREATE, 149 BPF_CGROUP_SOCK_OPS, 150 BPF_SK_SKB_STREAM_PARSER, 151 BPF_SK_SKB_STREAM_VERDICT, 152 BPF_CGROUP_DEVICE, 153 BPF_SK_MSG_VERDICT, 154 BPF_CGROUP_INET4_BIND, 155 BPF_CGROUP_INET6_BIND, 156 BPF_CGROUP_INET4_CONNECT, 157 BPF_CGROUP_INET6_CONNECT, 158 BPF_CGROUP_INET4_POST_BIND, 159 BPF_CGROUP_INET6_POST_BIND, 160 __MAX_BPF_ATTACH_TYPE 161 }; 162 163 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 164 165 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 166 * 167 * NONE(default): No further bpf programs allowed in the subtree. 168 * 169 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 170 * the program in this cgroup yields to sub-cgroup program. 171 * 172 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 173 * that cgroup program gets run in addition to the program in this cgroup. 174 * 175 * Only one program is allowed to be attached to a cgroup with 176 * NONE or BPF_F_ALLOW_OVERRIDE flag. 177 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 178 * release old program and attach the new one. Attach flags has to match. 179 * 180 * Multiple programs are allowed to be attached to a cgroup with 181 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 182 * (those that were attached first, run first) 183 * The programs of sub-cgroup are executed first, then programs of 184 * this cgroup and then programs of parent cgroup. 185 * When children program makes decision (like picking TCP CA or sock bind) 186 * parent program has a chance to override it. 187 * 188 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 189 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 190 * Ex1: 191 * cgrp1 (MULTI progs A, B) -> 192 * cgrp2 (OVERRIDE prog C) -> 193 * cgrp3 (MULTI prog D) -> 194 * cgrp4 (OVERRIDE prog E) -> 195 * cgrp5 (NONE prog F) 196 * the event in cgrp5 triggers execution of F,D,A,B in that order. 197 * if prog F is detached, the execution is E,D,A,B 198 * if prog F and D are detached, the execution is E,A,B 199 * if prog F, E and D are detached, the execution is C,A,B 200 * 201 * All eligible programs are executed regardless of return code from 202 * earlier programs. 203 */ 204 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 205 #define BPF_F_ALLOW_MULTI (1U << 1) 206 207 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 208 * verifier will perform strict alignment checking as if the kernel 209 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 210 * and NET_IP_ALIGN defined to 2. 211 */ 212 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 213 214 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */ 215 #define BPF_PSEUDO_MAP_FD 1 216 217 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 218 * offset to another bpf function 219 */ 220 #define BPF_PSEUDO_CALL 1 221 222 /* flags for BPF_MAP_UPDATE_ELEM command */ 223 #define BPF_ANY 0 /* create new element or update existing */ 224 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 225 #define BPF_EXIST 2 /* update existing element */ 226 227 /* flags for BPF_MAP_CREATE command */ 228 #define BPF_F_NO_PREALLOC (1U << 0) 229 /* Instead of having one common LRU list in the 230 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 231 * which can scale and perform better. 232 * Note, the LRU nodes (including free nodes) cannot be moved 233 * across different LRU lists. 234 */ 235 #define BPF_F_NO_COMMON_LRU (1U << 1) 236 /* Specify numa node during map creation */ 237 #define BPF_F_NUMA_NODE (1U << 2) 238 239 /* flags for BPF_PROG_QUERY */ 240 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 241 242 #define BPF_OBJ_NAME_LEN 16U 243 244 /* Flags for accessing BPF object */ 245 #define BPF_F_RDONLY (1U << 3) 246 #define BPF_F_WRONLY (1U << 4) 247 248 /* Flag for stack_map, store build_id+offset instead of pointer */ 249 #define BPF_F_STACK_BUILD_ID (1U << 5) 250 251 enum bpf_stack_build_id_status { 252 /* user space need an empty entry to identify end of a trace */ 253 BPF_STACK_BUILD_ID_EMPTY = 0, 254 /* with valid build_id and offset */ 255 BPF_STACK_BUILD_ID_VALID = 1, 256 /* couldn't get build_id, fallback to ip */ 257 BPF_STACK_BUILD_ID_IP = 2, 258 }; 259 260 #define BPF_BUILD_ID_SIZE 20 261 struct bpf_stack_build_id { 262 __s32 status; 263 unsigned char build_id[BPF_BUILD_ID_SIZE]; 264 union { 265 __u64 offset; 266 __u64 ip; 267 }; 268 }; 269 270 union bpf_attr { 271 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 272 __u32 map_type; /* one of enum bpf_map_type */ 273 __u32 key_size; /* size of key in bytes */ 274 __u32 value_size; /* size of value in bytes */ 275 __u32 max_entries; /* max number of entries in a map */ 276 __u32 map_flags; /* BPF_MAP_CREATE related 277 * flags defined above. 278 */ 279 __u32 inner_map_fd; /* fd pointing to the inner map */ 280 __u32 numa_node; /* numa node (effective only if 281 * BPF_F_NUMA_NODE is set). 282 */ 283 char map_name[BPF_OBJ_NAME_LEN]; 284 __u32 map_ifindex; /* ifindex of netdev to create on */ 285 __u32 btf_fd; /* fd pointing to a BTF type data */ 286 __u32 btf_key_id; /* BTF type_id of the key */ 287 __u32 btf_value_id; /* BTF type_id of the value */ 288 }; 289 290 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 291 __u32 map_fd; 292 __aligned_u64 key; 293 union { 294 __aligned_u64 value; 295 __aligned_u64 next_key; 296 }; 297 __u64 flags; 298 }; 299 300 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 301 __u32 prog_type; /* one of enum bpf_prog_type */ 302 __u32 insn_cnt; 303 __aligned_u64 insns; 304 __aligned_u64 license; 305 __u32 log_level; /* verbosity level of verifier */ 306 __u32 log_size; /* size of user buffer */ 307 __aligned_u64 log_buf; /* user supplied buffer */ 308 __u32 kern_version; /* checked when prog_type=kprobe */ 309 __u32 prog_flags; 310 char prog_name[BPF_OBJ_NAME_LEN]; 311 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 312 /* For some prog types expected attach type must be known at 313 * load time to verify attach type specific parts of prog 314 * (context accesses, allowed helpers, etc). 315 */ 316 __u32 expected_attach_type; 317 }; 318 319 struct { /* anonymous struct used by BPF_OBJ_* commands */ 320 __aligned_u64 pathname; 321 __u32 bpf_fd; 322 __u32 file_flags; 323 }; 324 325 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 326 __u32 target_fd; /* container object to attach to */ 327 __u32 attach_bpf_fd; /* eBPF program to attach */ 328 __u32 attach_type; 329 __u32 attach_flags; 330 }; 331 332 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 333 __u32 prog_fd; 334 __u32 retval; 335 __u32 data_size_in; 336 __u32 data_size_out; 337 __aligned_u64 data_in; 338 __aligned_u64 data_out; 339 __u32 repeat; 340 __u32 duration; 341 } test; 342 343 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 344 union { 345 __u32 start_id; 346 __u32 prog_id; 347 __u32 map_id; 348 __u32 btf_id; 349 }; 350 __u32 next_id; 351 __u32 open_flags; 352 }; 353 354 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 355 __u32 bpf_fd; 356 __u32 info_len; 357 __aligned_u64 info; 358 } info; 359 360 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 361 __u32 target_fd; /* container object to query */ 362 __u32 attach_type; 363 __u32 query_flags; 364 __u32 attach_flags; 365 __aligned_u64 prog_ids; 366 __u32 prog_cnt; 367 } query; 368 369 struct { 370 __u64 name; 371 __u32 prog_fd; 372 } raw_tracepoint; 373 374 struct { /* anonymous struct for BPF_BTF_LOAD */ 375 __aligned_u64 btf; 376 __aligned_u64 btf_log_buf; 377 __u32 btf_size; 378 __u32 btf_log_size; 379 __u32 btf_log_level; 380 }; 381 } __attribute__((aligned(8))); 382 383 /* The description below is an attempt at providing documentation to eBPF 384 * developers about the multiple available eBPF helper functions. It can be 385 * parsed and used to produce a manual page. The workflow is the following, 386 * and requires the rst2man utility: 387 * 388 * $ ./scripts/bpf_helpers_doc.py \ 389 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 390 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 391 * $ man /tmp/bpf-helpers.7 392 * 393 * Note that in order to produce this external documentation, some RST 394 * formatting is used in the descriptions to get "bold" and "italics" in 395 * manual pages. Also note that the few trailing white spaces are 396 * intentional, removing them would break paragraphs for rst2man. 397 * 398 * Start of BPF helper function descriptions: 399 * 400 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 401 * Description 402 * Perform a lookup in *map* for an entry associated to *key*. 403 * Return 404 * Map value associated to *key*, or **NULL** if no entry was 405 * found. 406 * 407 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 408 * Description 409 * Add or update the value of the entry associated to *key* in 410 * *map* with *value*. *flags* is one of: 411 * 412 * **BPF_NOEXIST** 413 * The entry for *key* must not exist in the map. 414 * **BPF_EXIST** 415 * The entry for *key* must already exist in the map. 416 * **BPF_ANY** 417 * No condition on the existence of the entry for *key*. 418 * 419 * Flag value **BPF_NOEXIST** cannot be used for maps of types 420 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 421 * elements always exist), the helper would return an error. 422 * Return 423 * 0 on success, or a negative error in case of failure. 424 * 425 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 426 * Description 427 * Delete entry with *key* from *map*. 428 * Return 429 * 0 on success, or a negative error in case of failure. 430 * 431 * int bpf_probe_read(void *dst, u32 size, const void *src) 432 * Description 433 * For tracing programs, safely attempt to read *size* bytes from 434 * address *src* and store the data in *dst*. 435 * Return 436 * 0 on success, or a negative error in case of failure. 437 * 438 * u64 bpf_ktime_get_ns(void) 439 * Description 440 * Return the time elapsed since system boot, in nanoseconds. 441 * Return 442 * Current *ktime*. 443 * 444 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 445 * Description 446 * This helper is a "printk()-like" facility for debugging. It 447 * prints a message defined by format *fmt* (of size *fmt_size*) 448 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 449 * available. It can take up to three additional **u64** 450 * arguments (as an eBPF helpers, the total number of arguments is 451 * limited to five). 452 * 453 * Each time the helper is called, it appends a line to the trace. 454 * The format of the trace is customizable, and the exact output 455 * one will get depends on the options set in 456 * *\/sys/kernel/debug/tracing/trace_options* (see also the 457 * *README* file under the same directory). However, it usually 458 * defaults to something like: 459 * 460 * :: 461 * 462 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 463 * 464 * In the above: 465 * 466 * * ``telnet`` is the name of the current task. 467 * * ``470`` is the PID of the current task. 468 * * ``001`` is the CPU number on which the task is 469 * running. 470 * * In ``.N..``, each character refers to a set of 471 * options (whether irqs are enabled, scheduling 472 * options, whether hard/softirqs are running, level of 473 * preempt_disabled respectively). **N** means that 474 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 475 * are set. 476 * * ``419421.045894`` is a timestamp. 477 * * ``0x00000001`` is a fake value used by BPF for the 478 * instruction pointer register. 479 * * ``<formatted msg>`` is the message formatted with 480 * *fmt*. 481 * 482 * The conversion specifiers supported by *fmt* are similar, but 483 * more limited than for printk(). They are **%d**, **%i**, 484 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 485 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 486 * of field, padding with zeroes, etc.) is available, and the 487 * helper will return **-EINVAL** (but print nothing) if it 488 * encounters an unknown specifier. 489 * 490 * Also, note that **bpf_trace_printk**\ () is slow, and should 491 * only be used for debugging purposes. For this reason, a notice 492 * bloc (spanning several lines) is printed to kernel logs and 493 * states that the helper should not be used "for production use" 494 * the first time this helper is used (or more precisely, when 495 * **trace_printk**\ () buffers are allocated). For passing values 496 * to user space, perf events should be preferred. 497 * Return 498 * The number of bytes written to the buffer, or a negative error 499 * in case of failure. 500 * 501 * u32 bpf_get_prandom_u32(void) 502 * Description 503 * Get a pseudo-random number. 504 * 505 * From a security point of view, this helper uses its own 506 * pseudo-random internal state, and cannot be used to infer the 507 * seed of other random functions in the kernel. However, it is 508 * essential to note that the generator used by the helper is not 509 * cryptographically secure. 510 * Return 511 * A random 32-bit unsigned value. 512 * 513 * u32 bpf_get_smp_processor_id(void) 514 * Description 515 * Get the SMP (symmetric multiprocessing) processor id. Note that 516 * all programs run with preemption disabled, which means that the 517 * SMP processor id is stable during all the execution of the 518 * program. 519 * Return 520 * The SMP id of the processor running the program. 521 * 522 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 523 * Description 524 * Store *len* bytes from address *from* into the packet 525 * associated to *skb*, at *offset*. *flags* are a combination of 526 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 527 * checksum for the packet after storing the bytes) and 528 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 529 * **->swhash** and *skb*\ **->l4hash** to 0). 530 * 531 * A call to this helper is susceptible to change the underlaying 532 * packet buffer. Therefore, at load time, all checks on pointers 533 * previously done by the verifier are invalidated and must be 534 * performed again, if the helper is used in combination with 535 * direct packet access. 536 * Return 537 * 0 on success, or a negative error in case of failure. 538 * 539 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 540 * Description 541 * Recompute the layer 3 (e.g. IP) checksum for the packet 542 * associated to *skb*. Computation is incremental, so the helper 543 * must know the former value of the header field that was 544 * modified (*from*), the new value of this field (*to*), and the 545 * number of bytes (2 or 4) for this field, stored in *size*. 546 * Alternatively, it is possible to store the difference between 547 * the previous and the new values of the header field in *to*, by 548 * setting *from* and *size* to 0. For both methods, *offset* 549 * indicates the location of the IP checksum within the packet. 550 * 551 * This helper works in combination with **bpf_csum_diff**\ (), 552 * which does not update the checksum in-place, but offers more 553 * flexibility and can handle sizes larger than 2 or 4 for the 554 * checksum to update. 555 * 556 * A call to this helper is susceptible to change the underlaying 557 * packet buffer. Therefore, at load time, all checks on pointers 558 * previously done by the verifier are invalidated and must be 559 * performed again, if the helper is used in combination with 560 * direct packet access. 561 * Return 562 * 0 on success, or a negative error in case of failure. 563 * 564 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 565 * Description 566 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 567 * packet associated to *skb*. Computation is incremental, so the 568 * helper must know the former value of the header field that was 569 * modified (*from*), the new value of this field (*to*), and the 570 * number of bytes (2 or 4) for this field, stored on the lowest 571 * four bits of *flags*. Alternatively, it is possible to store 572 * the difference between the previous and the new values of the 573 * header field in *to*, by setting *from* and the four lowest 574 * bits of *flags* to 0. For both methods, *offset* indicates the 575 * location of the IP checksum within the packet. In addition to 576 * the size of the field, *flags* can be added (bitwise OR) actual 577 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 578 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 579 * for updates resulting in a null checksum the value is set to 580 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 581 * the checksum is to be computed against a pseudo-header. 582 * 583 * This helper works in combination with **bpf_csum_diff**\ (), 584 * which does not update the checksum in-place, but offers more 585 * flexibility and can handle sizes larger than 2 or 4 for the 586 * checksum to update. 587 * 588 * A call to this helper is susceptible to change the underlaying 589 * packet buffer. Therefore, at load time, all checks on pointers 590 * previously done by the verifier are invalidated and must be 591 * performed again, if the helper is used in combination with 592 * direct packet access. 593 * Return 594 * 0 on success, or a negative error in case of failure. 595 * 596 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 597 * Description 598 * This special helper is used to trigger a "tail call", or in 599 * other words, to jump into another eBPF program. The same stack 600 * frame is used (but values on stack and in registers for the 601 * caller are not accessible to the callee). This mechanism allows 602 * for program chaining, either for raising the maximum number of 603 * available eBPF instructions, or to execute given programs in 604 * conditional blocks. For security reasons, there is an upper 605 * limit to the number of successive tail calls that can be 606 * performed. 607 * 608 * Upon call of this helper, the program attempts to jump into a 609 * program referenced at index *index* in *prog_array_map*, a 610 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 611 * *ctx*, a pointer to the context. 612 * 613 * If the call succeeds, the kernel immediately runs the first 614 * instruction of the new program. This is not a function call, 615 * and it never returns to the previous program. If the call 616 * fails, then the helper has no effect, and the caller continues 617 * to run its subsequent instructions. A call can fail if the 618 * destination program for the jump does not exist (i.e. *index* 619 * is superior to the number of entries in *prog_array_map*), or 620 * if the maximum number of tail calls has been reached for this 621 * chain of programs. This limit is defined in the kernel by the 622 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 623 * which is currently set to 32. 624 * Return 625 * 0 on success, or a negative error in case of failure. 626 * 627 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 628 * Description 629 * Clone and redirect the packet associated to *skb* to another 630 * net device of index *ifindex*. Both ingress and egress 631 * interfaces can be used for redirection. The **BPF_F_INGRESS** 632 * value in *flags* is used to make the distinction (ingress path 633 * is selected if the flag is present, egress path otherwise). 634 * This is the only flag supported for now. 635 * 636 * In comparison with **bpf_redirect**\ () helper, 637 * **bpf_clone_redirect**\ () has the associated cost of 638 * duplicating the packet buffer, but this can be executed out of 639 * the eBPF program. Conversely, **bpf_redirect**\ () is more 640 * efficient, but it is handled through an action code where the 641 * redirection happens only after the eBPF program has returned. 642 * 643 * A call to this helper is susceptible to change the underlaying 644 * packet buffer. Therefore, at load time, all checks on pointers 645 * previously done by the verifier are invalidated and must be 646 * performed again, if the helper is used in combination with 647 * direct packet access. 648 * Return 649 * 0 on success, or a negative error in case of failure. 650 * 651 * u64 bpf_get_current_pid_tgid(void) 652 * Return 653 * A 64-bit integer containing the current tgid and pid, and 654 * created as such: 655 * *current_task*\ **->tgid << 32 \|** 656 * *current_task*\ **->pid**. 657 * 658 * u64 bpf_get_current_uid_gid(void) 659 * Return 660 * A 64-bit integer containing the current GID and UID, and 661 * created as such: *current_gid* **<< 32 \|** *current_uid*. 662 * 663 * int bpf_get_current_comm(char *buf, u32 size_of_buf) 664 * Description 665 * Copy the **comm** attribute of the current task into *buf* of 666 * *size_of_buf*. The **comm** attribute contains the name of 667 * the executable (excluding the path) for the current task. The 668 * *size_of_buf* must be strictly positive. On success, the 669 * helper makes sure that the *buf* is NUL-terminated. On failure, 670 * it is filled with zeroes. 671 * Return 672 * 0 on success, or a negative error in case of failure. 673 * 674 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 675 * Description 676 * Retrieve the classid for the current task, i.e. for the net_cls 677 * cgroup to which *skb* belongs. 678 * 679 * This helper can be used on TC egress path, but not on ingress. 680 * 681 * The net_cls cgroup provides an interface to tag network packets 682 * based on a user-provided identifier for all traffic coming from 683 * the tasks belonging to the related cgroup. See also the related 684 * kernel documentation, available from the Linux sources in file 685 * *Documentation/cgroup-v1/net_cls.txt*. 686 * 687 * The Linux kernel has two versions for cgroups: there are 688 * cgroups v1 and cgroups v2. Both are available to users, who can 689 * use a mixture of them, but note that the net_cls cgroup is for 690 * cgroup v1 only. This makes it incompatible with BPF programs 691 * run on cgroups, which is a cgroup-v2-only feature (a socket can 692 * only hold data for one version of cgroups at a time). 693 * 694 * This helper is only available is the kernel was compiled with 695 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 696 * "**y**" or to "**m**". 697 * Return 698 * The classid, or 0 for the default unconfigured classid. 699 * 700 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 701 * Description 702 * Push a *vlan_tci* (VLAN tag control information) of protocol 703 * *vlan_proto* to the packet associated to *skb*, then update 704 * the checksum. Note that if *vlan_proto* is different from 705 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 706 * be **ETH_P_8021Q**. 707 * 708 * A call to this helper is susceptible to change the underlaying 709 * packet buffer. Therefore, at load time, all checks on pointers 710 * previously done by the verifier are invalidated and must be 711 * performed again, if the helper is used in combination with 712 * direct packet access. 713 * Return 714 * 0 on success, or a negative error in case of failure. 715 * 716 * int bpf_skb_vlan_pop(struct sk_buff *skb) 717 * Description 718 * Pop a VLAN header from the packet associated to *skb*. 719 * 720 * A call to this helper is susceptible to change the underlaying 721 * packet buffer. Therefore, at load time, all checks on pointers 722 * previously done by the verifier are invalidated and must be 723 * performed again, if the helper is used in combination with 724 * direct packet access. 725 * Return 726 * 0 on success, or a negative error in case of failure. 727 * 728 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 729 * Description 730 * Get tunnel metadata. This helper takes a pointer *key* to an 731 * empty **struct bpf_tunnel_key** of **size**, that will be 732 * filled with tunnel metadata for the packet associated to *skb*. 733 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 734 * indicates that the tunnel is based on IPv6 protocol instead of 735 * IPv4. 736 * 737 * The **struct bpf_tunnel_key** is an object that generalizes the 738 * principal parameters used by various tunneling protocols into a 739 * single struct. This way, it can be used to easily make a 740 * decision based on the contents of the encapsulation header, 741 * "summarized" in this struct. In particular, it holds the IP 742 * address of the remote end (IPv4 or IPv6, depending on the case) 743 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 744 * this struct exposes the *key*\ **->tunnel_id**, which is 745 * generally mapped to a VNI (Virtual Network Identifier), making 746 * it programmable together with the **bpf_skb_set_tunnel_key**\ 747 * () helper. 748 * 749 * Let's imagine that the following code is part of a program 750 * attached to the TC ingress interface, on one end of a GRE 751 * tunnel, and is supposed to filter out all messages coming from 752 * remote ends with IPv4 address other than 10.0.0.1: 753 * 754 * :: 755 * 756 * int ret; 757 * struct bpf_tunnel_key key = {}; 758 * 759 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 760 * if (ret < 0) 761 * return TC_ACT_SHOT; // drop packet 762 * 763 * if (key.remote_ipv4 != 0x0a000001) 764 * return TC_ACT_SHOT; // drop packet 765 * 766 * return TC_ACT_OK; // accept packet 767 * 768 * This interface can also be used with all encapsulation devices 769 * that can operate in "collect metadata" mode: instead of having 770 * one network device per specific configuration, the "collect 771 * metadata" mode only requires a single device where the 772 * configuration can be extracted from this helper. 773 * 774 * This can be used together with various tunnels such as VXLan, 775 * Geneve, GRE or IP in IP (IPIP). 776 * Return 777 * 0 on success, or a negative error in case of failure. 778 * 779 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 780 * Description 781 * Populate tunnel metadata for packet associated to *skb.* The 782 * tunnel metadata is set to the contents of *key*, of *size*. The 783 * *flags* can be set to a combination of the following values: 784 * 785 * **BPF_F_TUNINFO_IPV6** 786 * Indicate that the tunnel is based on IPv6 protocol 787 * instead of IPv4. 788 * **BPF_F_ZERO_CSUM_TX** 789 * For IPv4 packets, add a flag to tunnel metadata 790 * indicating that checksum computation should be skipped 791 * and checksum set to zeroes. 792 * **BPF_F_DONT_FRAGMENT** 793 * Add a flag to tunnel metadata indicating that the 794 * packet should not be fragmented. 795 * **BPF_F_SEQ_NUMBER** 796 * Add a flag to tunnel metadata indicating that a 797 * sequence number should be added to tunnel header before 798 * sending the packet. This flag was added for GRE 799 * encapsulation, but might be used with other protocols 800 * as well in the future. 801 * 802 * Here is a typical usage on the transmit path: 803 * 804 * :: 805 * 806 * struct bpf_tunnel_key key; 807 * populate key ... 808 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 809 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 810 * 811 * See also the description of the **bpf_skb_get_tunnel_key**\ () 812 * helper for additional information. 813 * Return 814 * 0 on success, or a negative error in case of failure. 815 * 816 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 817 * Description 818 * Read the value of a perf event counter. This helper relies on a 819 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 820 * the perf event counter is selected when *map* is updated with 821 * perf event file descriptors. The *map* is an array whose size 822 * is the number of available CPUs, and each cell contains a value 823 * relative to one CPU. The value to retrieve is indicated by 824 * *flags*, that contains the index of the CPU to look up, masked 825 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 826 * **BPF_F_CURRENT_CPU** to indicate that the value for the 827 * current CPU should be retrieved. 828 * 829 * Note that before Linux 4.13, only hardware perf event can be 830 * retrieved. 831 * 832 * Also, be aware that the newer helper 833 * **bpf_perf_event_read_value**\ () is recommended over 834 * **bpf_perf_event_read**\ () in general. The latter has some ABI 835 * quirks where error and counter value are used as a return code 836 * (which is wrong to do since ranges may overlap). This issue is 837 * fixed with **bpf_perf_event_read_value**\ (), which at the same 838 * time provides more features over the **bpf_perf_event_read**\ 839 * () interface. Please refer to the description of 840 * **bpf_perf_event_read_value**\ () for details. 841 * Return 842 * The value of the perf event counter read from the map, or a 843 * negative error code in case of failure. 844 * 845 * int bpf_redirect(u32 ifindex, u64 flags) 846 * Description 847 * Redirect the packet to another net device of index *ifindex*. 848 * This helper is somewhat similar to **bpf_clone_redirect**\ 849 * (), except that the packet is not cloned, which provides 850 * increased performance. 851 * 852 * Except for XDP, both ingress and egress interfaces can be used 853 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 854 * to make the distinction (ingress path is selected if the flag 855 * is present, egress path otherwise). Currently, XDP only 856 * supports redirection to the egress interface, and accepts no 857 * flag at all. 858 * 859 * The same effect can be attained with the more generic 860 * **bpf_redirect_map**\ (), which requires specific maps to be 861 * used but offers better performance. 862 * Return 863 * For XDP, the helper returns **XDP_REDIRECT** on success or 864 * **XDP_ABORTED** on error. For other program types, the values 865 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 866 * error. 867 * 868 * u32 bpf_get_route_realm(struct sk_buff *skb) 869 * Description 870 * Retrieve the realm or the route, that is to say the 871 * **tclassid** field of the destination for the *skb*. The 872 * indentifier retrieved is a user-provided tag, similar to the 873 * one used with the net_cls cgroup (see description for 874 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 875 * held by a route (a destination entry), not by a task. 876 * 877 * Retrieving this identifier works with the clsact TC egress hook 878 * (see also **tc-bpf(8)**), or alternatively on conventional 879 * classful egress qdiscs, but not on TC ingress path. In case of 880 * clsact TC egress hook, this has the advantage that, internally, 881 * the destination entry has not been dropped yet in the transmit 882 * path. Therefore, the destination entry does not need to be 883 * artificially held via **netif_keep_dst**\ () for a classful 884 * qdisc until the *skb* is freed. 885 * 886 * This helper is available only if the kernel was compiled with 887 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 888 * Return 889 * The realm of the route for the packet associated to *skb*, or 0 890 * if none was found. 891 * 892 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 893 * Description 894 * Write raw *data* blob into a special BPF perf event held by 895 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 896 * event must have the following attributes: **PERF_SAMPLE_RAW** 897 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 898 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 899 * 900 * The *flags* are used to indicate the index in *map* for which 901 * the value must be put, masked with **BPF_F_INDEX_MASK**. 902 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 903 * to indicate that the index of the current CPU core should be 904 * used. 905 * 906 * The value to write, of *size*, is passed through eBPF stack and 907 * pointed by *data*. 908 * 909 * The context of the program *ctx* needs also be passed to the 910 * helper. 911 * 912 * On user space, a program willing to read the values needs to 913 * call **perf_event_open**\ () on the perf event (either for 914 * one or for all CPUs) and to store the file descriptor into the 915 * *map*. This must be done before the eBPF program can send data 916 * into it. An example is available in file 917 * *samples/bpf/trace_output_user.c* in the Linux kernel source 918 * tree (the eBPF program counterpart is in 919 * *samples/bpf/trace_output_kern.c*). 920 * 921 * **bpf_perf_event_output**\ () achieves better performance 922 * than **bpf_trace_printk**\ () for sharing data with user 923 * space, and is much better suitable for streaming data from eBPF 924 * programs. 925 * 926 * Note that this helper is not restricted to tracing use cases 927 * and can be used with programs attached to TC or XDP as well, 928 * where it allows for passing data to user space listeners. Data 929 * can be: 930 * 931 * * Only custom structs, 932 * * Only the packet payload, or 933 * * A combination of both. 934 * Return 935 * 0 on success, or a negative error in case of failure. 936 * 937 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 938 * Description 939 * This helper was provided as an easy way to load data from a 940 * packet. It can be used to load *len* bytes from *offset* from 941 * the packet associated to *skb*, into the buffer pointed by 942 * *to*. 943 * 944 * Since Linux 4.7, usage of this helper has mostly been replaced 945 * by "direct packet access", enabling packet data to be 946 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 947 * pointing respectively to the first byte of packet data and to 948 * the byte after the last byte of packet data. However, it 949 * remains useful if one wishes to read large quantities of data 950 * at once from a packet into the eBPF stack. 951 * Return 952 * 0 on success, or a negative error in case of failure. 953 * 954 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 955 * Description 956 * Walk a user or a kernel stack and return its id. To achieve 957 * this, the helper needs *ctx*, which is a pointer to the context 958 * on which the tracing program is executed, and a pointer to a 959 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 960 * 961 * The last argument, *flags*, holds the number of stack frames to 962 * skip (from 0 to 255), masked with 963 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 964 * a combination of the following flags: 965 * 966 * **BPF_F_USER_STACK** 967 * Collect a user space stack instead of a kernel stack. 968 * **BPF_F_FAST_STACK_CMP** 969 * Compare stacks by hash only. 970 * **BPF_F_REUSE_STACKID** 971 * If two different stacks hash into the same *stackid*, 972 * discard the old one. 973 * 974 * The stack id retrieved is a 32 bit long integer handle which 975 * can be further combined with other data (including other stack 976 * ids) and used as a key into maps. This can be useful for 977 * generating a variety of graphs (such as flame graphs or off-cpu 978 * graphs). 979 * 980 * For walking a stack, this helper is an improvement over 981 * **bpf_probe_read**\ (), which can be used with unrolled loops 982 * but is not efficient and consumes a lot of eBPF instructions. 983 * Instead, **bpf_get_stackid**\ () can collect up to 984 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 985 * this limit can be controlled with the **sysctl** program, and 986 * that it should be manually increased in order to profile long 987 * user stacks (such as stacks for Java programs). To do so, use: 988 * 989 * :: 990 * 991 * # sysctl kernel.perf_event_max_stack=<new value> 992 * 993 * Return 994 * The positive or null stack id on success, or a negative error 995 * in case of failure. 996 * 997 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 998 * Description 999 * Compute a checksum difference, from the raw buffer pointed by 1000 * *from*, of length *from_size* (that must be a multiple of 4), 1001 * towards the raw buffer pointed by *to*, of size *to_size* 1002 * (same remark). An optional *seed* can be added to the value 1003 * (this can be cascaded, the seed may come from a previous call 1004 * to the helper). 1005 * 1006 * This is flexible enough to be used in several ways: 1007 * 1008 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1009 * checksum, it can be used when pushing new data. 1010 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1011 * checksum, it can be used when removing data from a packet. 1012 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1013 * can be used to compute a diff. Note that *from_size* and 1014 * *to_size* do not need to be equal. 1015 * 1016 * This helper can be used in combination with 1017 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1018 * which one can feed in the difference computed with 1019 * **bpf_csum_diff**\ (). 1020 * Return 1021 * The checksum result, or a negative error code in case of 1022 * failure. 1023 * 1024 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1025 * Description 1026 * Retrieve tunnel options metadata for the packet associated to 1027 * *skb*, and store the raw tunnel option data to the buffer *opt* 1028 * of *size*. 1029 * 1030 * This helper can be used with encapsulation devices that can 1031 * operate in "collect metadata" mode (please refer to the related 1032 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1033 * more details). A particular example where this can be used is 1034 * in combination with the Geneve encapsulation protocol, where it 1035 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1036 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1037 * the eBPF program. This allows for full customization of these 1038 * headers. 1039 * Return 1040 * The size of the option data retrieved. 1041 * 1042 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 1043 * Description 1044 * Set tunnel options metadata for the packet associated to *skb* 1045 * to the option data contained in the raw buffer *opt* of *size*. 1046 * 1047 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1048 * helper for additional information. 1049 * Return 1050 * 0 on success, or a negative error in case of failure. 1051 * 1052 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1053 * Description 1054 * Change the protocol of the *skb* to *proto*. Currently 1055 * supported are transition from IPv4 to IPv6, and from IPv6 to 1056 * IPv4. The helper takes care of the groundwork for the 1057 * transition, including resizing the socket buffer. The eBPF 1058 * program is expected to fill the new headers, if any, via 1059 * **skb_store_bytes**\ () and to recompute the checksums with 1060 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1061 * (). The main case for this helper is to perform NAT64 1062 * operations out of an eBPF program. 1063 * 1064 * Internally, the GSO type is marked as dodgy so that headers are 1065 * checked and segments are recalculated by the GSO/GRO engine. 1066 * The size for GSO target is adapted as well. 1067 * 1068 * All values for *flags* are reserved for future usage, and must 1069 * be left at zero. 1070 * 1071 * A call to this helper is susceptible to change the underlaying 1072 * packet buffer. Therefore, at load time, all checks on pointers 1073 * previously done by the verifier are invalidated and must be 1074 * performed again, if the helper is used in combination with 1075 * direct packet access. 1076 * Return 1077 * 0 on success, or a negative error in case of failure. 1078 * 1079 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1080 * Description 1081 * Change the packet type for the packet associated to *skb*. This 1082 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1083 * the eBPF program does not have a write access to *skb*\ 1084 * **->pkt_type** beside this helper. Using a helper here allows 1085 * for graceful handling of errors. 1086 * 1087 * The major use case is to change incoming *skb*s to 1088 * **PACKET_HOST** in a programmatic way instead of having to 1089 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1090 * example. 1091 * 1092 * Note that *type* only allows certain values. At this time, they 1093 * are: 1094 * 1095 * **PACKET_HOST** 1096 * Packet is for us. 1097 * **PACKET_BROADCAST** 1098 * Send packet to all. 1099 * **PACKET_MULTICAST** 1100 * Send packet to group. 1101 * **PACKET_OTHERHOST** 1102 * Send packet to someone else. 1103 * Return 1104 * 0 on success, or a negative error in case of failure. 1105 * 1106 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1107 * Description 1108 * Check whether *skb* is a descendant of the cgroup2 held by 1109 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1110 * Return 1111 * The return value depends on the result of the test, and can be: 1112 * 1113 * * 0, if the *skb* failed the cgroup2 descendant test. 1114 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1115 * * A negative error code, if an error occurred. 1116 * 1117 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1118 * Description 1119 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1120 * not set, in particular if the hash was cleared due to mangling, 1121 * recompute this hash. Later accesses to the hash can be done 1122 * directly with *skb*\ **->hash**. 1123 * 1124 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1125 * prototype with **bpf_skb_change_proto**\ (), or calling 1126 * **bpf_skb_store_bytes**\ () with the 1127 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1128 * the hash and to trigger a new computation for the next call to 1129 * **bpf_get_hash_recalc**\ (). 1130 * Return 1131 * The 32-bit hash. 1132 * 1133 * u64 bpf_get_current_task(void) 1134 * Return 1135 * A pointer to the current task struct. 1136 * 1137 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1138 * Description 1139 * Attempt in a safe way to write *len* bytes from the buffer 1140 * *src* to *dst* in memory. It only works for threads that are in 1141 * user context, and *dst* must be a valid user space address. 1142 * 1143 * This helper should not be used to implement any kind of 1144 * security mechanism because of TOC-TOU attacks, but rather to 1145 * debug, divert, and manipulate execution of semi-cooperative 1146 * processes. 1147 * 1148 * Keep in mind that this feature is meant for experiments, and it 1149 * has a risk of crashing the system and running programs. 1150 * Therefore, when an eBPF program using this helper is attached, 1151 * a warning including PID and process name is printed to kernel 1152 * logs. 1153 * Return 1154 * 0 on success, or a negative error in case of failure. 1155 * 1156 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1157 * Description 1158 * Check whether the probe is being run is the context of a given 1159 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1160 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1161 * Return 1162 * The return value depends on the result of the test, and can be: 1163 * 1164 * * 0, if the *skb* task belongs to the cgroup2. 1165 * * 1, if the *skb* task does not belong to the cgroup2. 1166 * * A negative error code, if an error occurred. 1167 * 1168 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1169 * Description 1170 * Resize (trim or grow) the packet associated to *skb* to the 1171 * new *len*. The *flags* are reserved for future usage, and must 1172 * be left at zero. 1173 * 1174 * The basic idea is that the helper performs the needed work to 1175 * change the size of the packet, then the eBPF program rewrites 1176 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1177 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1178 * and others. This helper is a slow path utility intended for 1179 * replies with control messages. And because it is targeted for 1180 * slow path, the helper itself can afford to be slow: it 1181 * implicitly linearizes, unclones and drops offloads from the 1182 * *skb*. 1183 * 1184 * A call to this helper is susceptible to change the underlaying 1185 * packet buffer. Therefore, at load time, all checks on pointers 1186 * previously done by the verifier are invalidated and must be 1187 * performed again, if the helper is used in combination with 1188 * direct packet access. 1189 * Return 1190 * 0 on success, or a negative error in case of failure. 1191 * 1192 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1193 * Description 1194 * Pull in non-linear data in case the *skb* is non-linear and not 1195 * all of *len* are part of the linear section. Make *len* bytes 1196 * from *skb* readable and writable. If a zero value is passed for 1197 * *len*, then the whole length of the *skb* is pulled. 1198 * 1199 * This helper is only needed for reading and writing with direct 1200 * packet access. 1201 * 1202 * For direct packet access, testing that offsets to access 1203 * are within packet boundaries (test on *skb*\ **->data_end**) is 1204 * susceptible to fail if offsets are invalid, or if the requested 1205 * data is in non-linear parts of the *skb*. On failure the 1206 * program can just bail out, or in the case of a non-linear 1207 * buffer, use a helper to make the data available. The 1208 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1209 * the data. Another one consists in using **bpf_skb_pull_data** 1210 * to pull in once the non-linear parts, then retesting and 1211 * eventually access the data. 1212 * 1213 * At the same time, this also makes sure the *skb* is uncloned, 1214 * which is a necessary condition for direct write. As this needs 1215 * to be an invariant for the write part only, the verifier 1216 * detects writes and adds a prologue that is calling 1217 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1218 * the very beginning in case it is indeed cloned. 1219 * 1220 * A call to this helper is susceptible to change the underlaying 1221 * packet buffer. Therefore, at load time, all checks on pointers 1222 * previously done by the verifier are invalidated and must be 1223 * performed again, if the helper is used in combination with 1224 * direct packet access. 1225 * Return 1226 * 0 on success, or a negative error in case of failure. 1227 * 1228 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1229 * Description 1230 * Add the checksum *csum* into *skb*\ **->csum** in case the 1231 * driver has supplied a checksum for the entire packet into that 1232 * field. Return an error otherwise. This helper is intended to be 1233 * used in combination with **bpf_csum_diff**\ (), in particular 1234 * when the checksum needs to be updated after data has been 1235 * written into the packet through direct packet access. 1236 * Return 1237 * The checksum on success, or a negative error code in case of 1238 * failure. 1239 * 1240 * void bpf_set_hash_invalid(struct sk_buff *skb) 1241 * Description 1242 * Invalidate the current *skb*\ **->hash**. It can be used after 1243 * mangling on headers through direct packet access, in order to 1244 * indicate that the hash is outdated and to trigger a 1245 * recalculation the next time the kernel tries to access this 1246 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1247 * 1248 * int bpf_get_numa_node_id(void) 1249 * Description 1250 * Return the id of the current NUMA node. The primary use case 1251 * for this helper is the selection of sockets for the local NUMA 1252 * node, when the program is attached to sockets using the 1253 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1254 * but the helper is also available to other eBPF program types, 1255 * similarly to **bpf_get_smp_processor_id**\ (). 1256 * Return 1257 * The id of current NUMA node. 1258 * 1259 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1260 * Description 1261 * Grows headroom of packet associated to *skb* and adjusts the 1262 * offset of the MAC header accordingly, adding *len* bytes of 1263 * space. It automatically extends and reallocates memory as 1264 * required. 1265 * 1266 * This helper can be used on a layer 3 *skb* to push a MAC header 1267 * for redirection into a layer 2 device. 1268 * 1269 * All values for *flags* are reserved for future usage, and must 1270 * be left at zero. 1271 * 1272 * A call to this helper is susceptible to change the underlaying 1273 * packet buffer. Therefore, at load time, all checks on pointers 1274 * previously done by the verifier are invalidated and must be 1275 * performed again, if the helper is used in combination with 1276 * direct packet access. 1277 * Return 1278 * 0 on success, or a negative error in case of failure. 1279 * 1280 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1281 * Description 1282 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1283 * it is possible to use a negative value for *delta*. This helper 1284 * can be used to prepare the packet for pushing or popping 1285 * headers. 1286 * 1287 * A call to this helper is susceptible to change the underlaying 1288 * packet buffer. Therefore, at load time, all checks on pointers 1289 * previously done by the verifier are invalidated and must be 1290 * performed again, if the helper is used in combination with 1291 * direct packet access. 1292 * Return 1293 * 0 on success, or a negative error in case of failure. 1294 * 1295 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1296 * Description 1297 * Copy a NUL terminated string from an unsafe address 1298 * *unsafe_ptr* to *dst*. The *size* should include the 1299 * terminating NUL byte. In case the string length is smaller than 1300 * *size*, the target is not padded with further NUL bytes. If the 1301 * string length is larger than *size*, just *size*-1 bytes are 1302 * copied and the last byte is set to NUL. 1303 * 1304 * On success, the length of the copied string is returned. This 1305 * makes this helper useful in tracing programs for reading 1306 * strings, and more importantly to get its length at runtime. See 1307 * the following snippet: 1308 * 1309 * :: 1310 * 1311 * SEC("kprobe/sys_open") 1312 * void bpf_sys_open(struct pt_regs *ctx) 1313 * { 1314 * char buf[PATHLEN]; // PATHLEN is defined to 256 1315 * int res = bpf_probe_read_str(buf, sizeof(buf), 1316 * ctx->di); 1317 * 1318 * // Consume buf, for example push it to 1319 * // userspace via bpf_perf_event_output(); we 1320 * // can use res (the string length) as event 1321 * // size, after checking its boundaries. 1322 * } 1323 * 1324 * In comparison, using **bpf_probe_read()** helper here instead 1325 * to read the string would require to estimate the length at 1326 * compile time, and would often result in copying more memory 1327 * than necessary. 1328 * 1329 * Another useful use case is when parsing individual process 1330 * arguments or individual environment variables navigating 1331 * *current*\ **->mm->arg_start** and *current*\ 1332 * **->mm->env_start**: using this helper and the return value, 1333 * one can quickly iterate at the right offset of the memory area. 1334 * Return 1335 * On success, the strictly positive length of the string, 1336 * including the trailing NUL character. On error, a negative 1337 * value. 1338 * 1339 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1340 * Description 1341 * If the **struct sk_buff** pointed by *skb* has a known socket, 1342 * retrieve the cookie (generated by the kernel) of this socket. 1343 * If no cookie has been set yet, generate a new cookie. Once 1344 * generated, the socket cookie remains stable for the life of the 1345 * socket. This helper can be useful for monitoring per socket 1346 * networking traffic statistics as it provides a unique socket 1347 * identifier per namespace. 1348 * Return 1349 * A 8-byte long non-decreasing number on success, or 0 if the 1350 * socket field is missing inside *skb*. 1351 * 1352 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1353 * Return 1354 * The owner UID of the socket associated to *skb*. If the socket 1355 * is **NULL**, or if it is not a full socket (i.e. if it is a 1356 * time-wait or a request socket instead), **overflowuid** value 1357 * is returned (note that **overflowuid** might also be the actual 1358 * UID value for the socket). 1359 * 1360 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1361 * Description 1362 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1363 * to value *hash*. 1364 * Return 1365 * 0 1366 * 1367 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1368 * Description 1369 * Emulate a call to **setsockopt()** on the socket associated to 1370 * *bpf_socket*, which must be a full socket. The *level* at 1371 * which the option resides and the name *optname* of the option 1372 * must be specified, see **setsockopt(2)** for more information. 1373 * The option value of length *optlen* is pointed by *optval*. 1374 * 1375 * This helper actually implements a subset of **setsockopt()**. 1376 * It supports the following *level*\ s: 1377 * 1378 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1379 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1380 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1381 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1382 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1383 * **TCP_BPF_SNDCWND_CLAMP**. 1384 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1385 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1386 * Return 1387 * 0 on success, or a negative error in case of failure. 1388 * 1389 * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags) 1390 * Description 1391 * Grow or shrink the room for data in the packet associated to 1392 * *skb* by *len_diff*, and according to the selected *mode*. 1393 * 1394 * There is a single supported mode at this time: 1395 * 1396 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1397 * (room space is added or removed below the layer 3 header). 1398 * 1399 * All values for *flags* are reserved for future usage, and must 1400 * be left at zero. 1401 * 1402 * A call to this helper is susceptible to change the underlaying 1403 * packet buffer. Therefore, at load time, all checks on pointers 1404 * previously done by the verifier are invalidated and must be 1405 * performed again, if the helper is used in combination with 1406 * direct packet access. 1407 * Return 1408 * 0 on success, or a negative error in case of failure. 1409 * 1410 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1411 * Description 1412 * Redirect the packet to the endpoint referenced by *map* at 1413 * index *key*. Depending on its type, this *map* can contain 1414 * references to net devices (for forwarding packets through other 1415 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1416 * but this is only implemented for native XDP (with driver 1417 * support) as of this writing). 1418 * 1419 * All values for *flags* are reserved for future usage, and must 1420 * be left at zero. 1421 * 1422 * When used to redirect packets to net devices, this helper 1423 * provides a high performance increase over **bpf_redirect**\ (). 1424 * This is due to various implementation details of the underlying 1425 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1426 * () tries to send packet as a "bulk" to the device. 1427 * Return 1428 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1429 * 1430 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1431 * Description 1432 * Redirect the packet to the socket referenced by *map* (of type 1433 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1434 * egress interfaces can be used for redirection. The 1435 * **BPF_F_INGRESS** value in *flags* is used to make the 1436 * distinction (ingress path is selected if the flag is present, 1437 * egress path otherwise). This is the only flag supported for now. 1438 * Return 1439 * **SK_PASS** on success, or **SK_DROP** on error. 1440 * 1441 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1442 * Description 1443 * Add an entry to, or update a *map* referencing sockets. The 1444 * *skops* is used as a new value for the entry associated to 1445 * *key*. *flags* is one of: 1446 * 1447 * **BPF_NOEXIST** 1448 * The entry for *key* must not exist in the map. 1449 * **BPF_EXIST** 1450 * The entry for *key* must already exist in the map. 1451 * **BPF_ANY** 1452 * No condition on the existence of the entry for *key*. 1453 * 1454 * If the *map* has eBPF programs (parser and verdict), those will 1455 * be inherited by the socket being added. If the socket is 1456 * already attached to eBPF programs, this results in an error. 1457 * Return 1458 * 0 on success, or a negative error in case of failure. 1459 * 1460 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1461 * Description 1462 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1463 * *delta* (which can be positive or negative). Note that this 1464 * operation modifies the address stored in *xdp_md*\ **->data**, 1465 * so the latter must be loaded only after the helper has been 1466 * called. 1467 * 1468 * The use of *xdp_md*\ **->data_meta** is optional and programs 1469 * are not required to use it. The rationale is that when the 1470 * packet is processed with XDP (e.g. as DoS filter), it is 1471 * possible to push further meta data along with it before passing 1472 * to the stack, and to give the guarantee that an ingress eBPF 1473 * program attached as a TC classifier on the same device can pick 1474 * this up for further post-processing. Since TC works with socket 1475 * buffers, it remains possible to set from XDP the **mark** or 1476 * **priority** pointers, or other pointers for the socket buffer. 1477 * Having this scratch space generic and programmable allows for 1478 * more flexibility as the user is free to store whatever meta 1479 * data they need. 1480 * 1481 * A call to this helper is susceptible to change the underlaying 1482 * packet buffer. Therefore, at load time, all checks on pointers 1483 * previously done by the verifier are invalidated and must be 1484 * performed again, if the helper is used in combination with 1485 * direct packet access. 1486 * Return 1487 * 0 on success, or a negative error in case of failure. 1488 * 1489 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1490 * Description 1491 * Read the value of a perf event counter, and store it into *buf* 1492 * of size *buf_size*. This helper relies on a *map* of type 1493 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1494 * counter is selected when *map* is updated with perf event file 1495 * descriptors. The *map* is an array whose size is the number of 1496 * available CPUs, and each cell contains a value relative to one 1497 * CPU. The value to retrieve is indicated by *flags*, that 1498 * contains the index of the CPU to look up, masked with 1499 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1500 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1501 * current CPU should be retrieved. 1502 * 1503 * This helper behaves in a way close to 1504 * **bpf_perf_event_read**\ () helper, save that instead of 1505 * just returning the value observed, it fills the *buf* 1506 * structure. This allows for additional data to be retrieved: in 1507 * particular, the enabled and running times (in *buf*\ 1508 * **->enabled** and *buf*\ **->running**, respectively) are 1509 * copied. In general, **bpf_perf_event_read_value**\ () is 1510 * recommended over **bpf_perf_event_read**\ (), which has some 1511 * ABI issues and provides fewer functionalities. 1512 * 1513 * These values are interesting, because hardware PMU (Performance 1514 * Monitoring Unit) counters are limited resources. When there are 1515 * more PMU based perf events opened than available counters, 1516 * kernel will multiplex these events so each event gets certain 1517 * percentage (but not all) of the PMU time. In case that 1518 * multiplexing happens, the number of samples or counter value 1519 * will not reflect the case compared to when no multiplexing 1520 * occurs. This makes comparison between different runs difficult. 1521 * Typically, the counter value should be normalized before 1522 * comparing to other experiments. The usual normalization is done 1523 * as follows. 1524 * 1525 * :: 1526 * 1527 * normalized_counter = counter * t_enabled / t_running 1528 * 1529 * Where t_enabled is the time enabled for event and t_running is 1530 * the time running for event since last normalization. The 1531 * enabled and running times are accumulated since the perf event 1532 * open. To achieve scaling factor between two invocations of an 1533 * eBPF program, users can can use CPU id as the key (which is 1534 * typical for perf array usage model) to remember the previous 1535 * value and do the calculation inside the eBPF program. 1536 * Return 1537 * 0 on success, or a negative error in case of failure. 1538 * 1539 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1540 * Description 1541 * For en eBPF program attached to a perf event, retrieve the 1542 * value of the event counter associated to *ctx* and store it in 1543 * the structure pointed by *buf* and of size *buf_size*. Enabled 1544 * and running times are also stored in the structure (see 1545 * description of helper **bpf_perf_event_read_value**\ () for 1546 * more details). 1547 * Return 1548 * 0 on success, or a negative error in case of failure. 1549 * 1550 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 1551 * Description 1552 * Emulate a call to **getsockopt()** on the socket associated to 1553 * *bpf_socket*, which must be a full socket. The *level* at 1554 * which the option resides and the name *optname* of the option 1555 * must be specified, see **getsockopt(2)** for more information. 1556 * The retrieved value is stored in the structure pointed by 1557 * *opval* and of length *optlen*. 1558 * 1559 * This helper actually implements a subset of **getsockopt()**. 1560 * It supports the following *level*\ s: 1561 * 1562 * * **IPPROTO_TCP**, which supports *optname* 1563 * **TCP_CONGESTION**. 1564 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1565 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1566 * Return 1567 * 0 on success, or a negative error in case of failure. 1568 * 1569 * int bpf_override_return(struct pt_reg *regs, u64 rc) 1570 * Description 1571 * Used for error injection, this helper uses kprobes to override 1572 * the return value of the probed function, and to set it to *rc*. 1573 * The first argument is the context *regs* on which the kprobe 1574 * works. 1575 * 1576 * This helper works by setting setting the PC (program counter) 1577 * to an override function which is run in place of the original 1578 * probed function. This means the probed function is not run at 1579 * all. The replacement function just returns with the required 1580 * value. 1581 * 1582 * This helper has security implications, and thus is subject to 1583 * restrictions. It is only available if the kernel was compiled 1584 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1585 * option, and in this case it only works on functions tagged with 1586 * **ALLOW_ERROR_INJECTION** in the kernel code. 1587 * 1588 * Also, the helper is only available for the architectures having 1589 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1590 * x86 architecture is the only one to support this feature. 1591 * Return 1592 * 0 1593 * 1594 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1595 * Description 1596 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1597 * for the full TCP socket associated to *bpf_sock_ops* to 1598 * *argval*. 1599 * 1600 * The primary use of this field is to determine if there should 1601 * be calls to eBPF programs of type 1602 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1603 * code. A program of the same type can change its value, per 1604 * connection and as necessary, when the connection is 1605 * established. This field is directly accessible for reading, but 1606 * this helper must be used for updates in order to return an 1607 * error if an eBPF program tries to set a callback that is not 1608 * supported in the current kernel. 1609 * 1610 * The supported callback values that *argval* can combine are: 1611 * 1612 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1613 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1614 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1615 * 1616 * Here are some examples of where one could call such eBPF 1617 * program: 1618 * 1619 * * When RTO fires. 1620 * * When a packet is retransmitted. 1621 * * When the connection terminates. 1622 * * When a packet is sent. 1623 * * When a packet is received. 1624 * Return 1625 * Code **-EINVAL** if the socket is not a full TCP socket; 1626 * otherwise, a positive number containing the bits that could not 1627 * be set is returned (which comes down to 0 if all bits were set 1628 * as required). 1629 * 1630 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1631 * Description 1632 * This helper is used in programs implementing policies at the 1633 * socket level. If the message *msg* is allowed to pass (i.e. if 1634 * the verdict eBPF program returns **SK_PASS**), redirect it to 1635 * the socket referenced by *map* (of type 1636 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1637 * egress interfaces can be used for redirection. The 1638 * **BPF_F_INGRESS** value in *flags* is used to make the 1639 * distinction (ingress path is selected if the flag is present, 1640 * egress path otherwise). This is the only flag supported for now. 1641 * Return 1642 * **SK_PASS** on success, or **SK_DROP** on error. 1643 * 1644 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1645 * Description 1646 * For socket policies, apply the verdict of the eBPF program to 1647 * the next *bytes* (number of bytes) of message *msg*. 1648 * 1649 * For example, this helper can be used in the following cases: 1650 * 1651 * * A single **sendmsg**\ () or **sendfile**\ () system call 1652 * contains multiple logical messages that the eBPF program is 1653 * supposed to read and for which it should apply a verdict. 1654 * * An eBPF program only cares to read the first *bytes* of a 1655 * *msg*. If the message has a large payload, then setting up 1656 * and calling the eBPF program repeatedly for all bytes, even 1657 * though the verdict is already known, would create unnecessary 1658 * overhead. 1659 * 1660 * When called from within an eBPF program, the helper sets a 1661 * counter internal to the BPF infrastructure, that is used to 1662 * apply the last verdict to the next *bytes*. If *bytes* is 1663 * smaller than the current data being processed from a 1664 * **sendmsg**\ () or **sendfile**\ () system call, the first 1665 * *bytes* will be sent and the eBPF program will be re-run with 1666 * the pointer for start of data pointing to byte number *bytes* 1667 * **+ 1**. If *bytes* is larger than the current data being 1668 * processed, then the eBPF verdict will be applied to multiple 1669 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1670 * consumed. 1671 * 1672 * Note that if a socket closes with the internal counter holding 1673 * a non-zero value, this is not a problem because data is not 1674 * being buffered for *bytes* and is sent as it is received. 1675 * Return 1676 * 0 1677 * 1678 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1679 * Description 1680 * For socket policies, prevent the execution of the verdict eBPF 1681 * program for message *msg* until *bytes* (byte number) have been 1682 * accumulated. 1683 * 1684 * This can be used when one needs a specific number of bytes 1685 * before a verdict can be assigned, even if the data spans 1686 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1687 * case would be a user calling **sendmsg**\ () repeatedly with 1688 * 1-byte long message segments. Obviously, this is bad for 1689 * performance, but it is still valid. If the eBPF program needs 1690 * *bytes* bytes to validate a header, this helper can be used to 1691 * prevent the eBPF program to be called again until *bytes* have 1692 * been accumulated. 1693 * Return 1694 * 0 1695 * 1696 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1697 * Description 1698 * For socket policies, pull in non-linear data from user space 1699 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1700 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1701 * respectively. 1702 * 1703 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1704 * *msg* it can only parse data that the (**data**, **data_end**) 1705 * pointers have already consumed. For **sendmsg**\ () hooks this 1706 * is likely the first scatterlist element. But for calls relying 1707 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1708 * be the range (**0**, **0**) because the data is shared with 1709 * user space and by default the objective is to avoid allowing 1710 * user space to modify data while (or after) eBPF verdict is 1711 * being decided. This helper can be used to pull in data and to 1712 * set the start and end pointer to given values. Data will be 1713 * copied if necessary (i.e. if data was not linear and if start 1714 * and end pointers do not point to the same chunk). 1715 * 1716 * A call to this helper is susceptible to change the underlaying 1717 * packet buffer. Therefore, at load time, all checks on pointers 1718 * previously done by the verifier are invalidated and must be 1719 * performed again, if the helper is used in combination with 1720 * direct packet access. 1721 * 1722 * All values for *flags* are reserved for future usage, and must 1723 * be left at zero. 1724 * Return 1725 * 0 on success, or a negative error in case of failure. 1726 * 1727 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1728 * Description 1729 * Bind the socket associated to *ctx* to the address pointed by 1730 * *addr*, of length *addr_len*. This allows for making outgoing 1731 * connection from the desired IP address, which can be useful for 1732 * example when all processes inside a cgroup should use one 1733 * single IP address on a host that has multiple IP configured. 1734 * 1735 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1736 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1737 * **AF_INET6**). Looking for a free port to bind to can be 1738 * expensive, therefore binding to port is not permitted by the 1739 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1740 * must be set to zero. 1741 * Return 1742 * 0 on success, or a negative error in case of failure. 1743 * 1744 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1745 * Description 1746 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1747 * only possible to shrink the packet as of this writing, 1748 * therefore *delta* must be a negative integer. 1749 * 1750 * A call to this helper is susceptible to change the underlaying 1751 * packet buffer. Therefore, at load time, all checks on pointers 1752 * previously done by the verifier are invalidated and must be 1753 * performed again, if the helper is used in combination with 1754 * direct packet access. 1755 * Return 1756 * 0 on success, or a negative error in case of failure. 1757 * 1758 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1759 * Description 1760 * Retrieve the XFRM state (IP transform framework, see also 1761 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1762 * 1763 * The retrieved value is stored in the **struct bpf_xfrm_state** 1764 * pointed by *xfrm_state* and of length *size*. 1765 * 1766 * All values for *flags* are reserved for future usage, and must 1767 * be left at zero. 1768 * 1769 * This helper is available only if the kernel was compiled with 1770 * **CONFIG_XFRM** configuration option. 1771 * Return 1772 * 0 on success, or a negative error in case of failure. 1773 * 1774 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1775 * Description 1776 * Return a user or a kernel stack in bpf program provided buffer. 1777 * To achieve this, the helper needs *ctx*, which is a pointer 1778 * to the context on which the tracing program is executed. 1779 * To store the stacktrace, the bpf program provides *buf* with 1780 * a nonnegative *size*. 1781 * 1782 * The last argument, *flags*, holds the number of stack frames to 1783 * skip (from 0 to 255), masked with 1784 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1785 * the following flags: 1786 * 1787 * **BPF_F_USER_STACK** 1788 * Collect a user space stack instead of a kernel stack. 1789 * **BPF_F_USER_BUILD_ID** 1790 * Collect buildid+offset instead of ips for user stack, 1791 * only valid if **BPF_F_USER_STACK** is also specified. 1792 * 1793 * **bpf_get_stack**\ () can collect up to 1794 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1795 * to sufficient large buffer size. Note that 1796 * this limit can be controlled with the **sysctl** program, and 1797 * that it should be manually increased in order to profile long 1798 * user stacks (such as stacks for Java programs). To do so, use: 1799 * 1800 * :: 1801 * 1802 * # sysctl kernel.perf_event_max_stack=<new value> 1803 * 1804 * Return 1805 * a non-negative value equal to or less than size on success, or 1806 * a negative error in case of failure. 1807 * 1808 * int skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 1809 * Description 1810 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1811 * it provides an easy way to load *len* bytes from *offset* 1812 * from the packet associated to *skb*, into the buffer pointed 1813 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1814 * a fifth argument *start_header* exists in order to select a 1815 * base offset to start from. *start_header* can be one of: 1816 * 1817 * **BPF_HDR_START_MAC** 1818 * Base offset to load data from is *skb*'s mac header. 1819 * **BPF_HDR_START_NET** 1820 * Base offset to load data from is *skb*'s network header. 1821 * 1822 * In general, "direct packet access" is the preferred method to 1823 * access packet data, however, this helper is in particular useful 1824 * in socket filters where *skb*\ **->data** does not always point 1825 * to the start of the mac header and where "direct packet access" 1826 * is not available. 1827 * 1828 * Return 1829 * 0 on success, or a negative error in case of failure. 1830 * 1831 */ 1832 #define __BPF_FUNC_MAPPER(FN) \ 1833 FN(unspec), \ 1834 FN(map_lookup_elem), \ 1835 FN(map_update_elem), \ 1836 FN(map_delete_elem), \ 1837 FN(probe_read), \ 1838 FN(ktime_get_ns), \ 1839 FN(trace_printk), \ 1840 FN(get_prandom_u32), \ 1841 FN(get_smp_processor_id), \ 1842 FN(skb_store_bytes), \ 1843 FN(l3_csum_replace), \ 1844 FN(l4_csum_replace), \ 1845 FN(tail_call), \ 1846 FN(clone_redirect), \ 1847 FN(get_current_pid_tgid), \ 1848 FN(get_current_uid_gid), \ 1849 FN(get_current_comm), \ 1850 FN(get_cgroup_classid), \ 1851 FN(skb_vlan_push), \ 1852 FN(skb_vlan_pop), \ 1853 FN(skb_get_tunnel_key), \ 1854 FN(skb_set_tunnel_key), \ 1855 FN(perf_event_read), \ 1856 FN(redirect), \ 1857 FN(get_route_realm), \ 1858 FN(perf_event_output), \ 1859 FN(skb_load_bytes), \ 1860 FN(get_stackid), \ 1861 FN(csum_diff), \ 1862 FN(skb_get_tunnel_opt), \ 1863 FN(skb_set_tunnel_opt), \ 1864 FN(skb_change_proto), \ 1865 FN(skb_change_type), \ 1866 FN(skb_under_cgroup), \ 1867 FN(get_hash_recalc), \ 1868 FN(get_current_task), \ 1869 FN(probe_write_user), \ 1870 FN(current_task_under_cgroup), \ 1871 FN(skb_change_tail), \ 1872 FN(skb_pull_data), \ 1873 FN(csum_update), \ 1874 FN(set_hash_invalid), \ 1875 FN(get_numa_node_id), \ 1876 FN(skb_change_head), \ 1877 FN(xdp_adjust_head), \ 1878 FN(probe_read_str), \ 1879 FN(get_socket_cookie), \ 1880 FN(get_socket_uid), \ 1881 FN(set_hash), \ 1882 FN(setsockopt), \ 1883 FN(skb_adjust_room), \ 1884 FN(redirect_map), \ 1885 FN(sk_redirect_map), \ 1886 FN(sock_map_update), \ 1887 FN(xdp_adjust_meta), \ 1888 FN(perf_event_read_value), \ 1889 FN(perf_prog_read_value), \ 1890 FN(getsockopt), \ 1891 FN(override_return), \ 1892 FN(sock_ops_cb_flags_set), \ 1893 FN(msg_redirect_map), \ 1894 FN(msg_apply_bytes), \ 1895 FN(msg_cork_bytes), \ 1896 FN(msg_pull_data), \ 1897 FN(bind), \ 1898 FN(xdp_adjust_tail), \ 1899 FN(skb_get_xfrm_state), \ 1900 FN(get_stack), \ 1901 FN(skb_load_bytes_relative), 1902 1903 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 1904 * function eBPF program intends to call 1905 */ 1906 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 1907 enum bpf_func_id { 1908 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 1909 __BPF_FUNC_MAX_ID, 1910 }; 1911 #undef __BPF_ENUM_FN 1912 1913 /* All flags used by eBPF helper functions, placed here. */ 1914 1915 /* BPF_FUNC_skb_store_bytes flags. */ 1916 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 1917 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 1918 1919 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 1920 * First 4 bits are for passing the header field size. 1921 */ 1922 #define BPF_F_HDR_FIELD_MASK 0xfULL 1923 1924 /* BPF_FUNC_l4_csum_replace flags. */ 1925 #define BPF_F_PSEUDO_HDR (1ULL << 4) 1926 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 1927 #define BPF_F_MARK_ENFORCE (1ULL << 6) 1928 1929 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 1930 #define BPF_F_INGRESS (1ULL << 0) 1931 1932 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 1933 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 1934 1935 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 1936 #define BPF_F_SKIP_FIELD_MASK 0xffULL 1937 #define BPF_F_USER_STACK (1ULL << 8) 1938 /* flags used by BPF_FUNC_get_stackid only. */ 1939 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 1940 #define BPF_F_REUSE_STACKID (1ULL << 10) 1941 /* flags used by BPF_FUNC_get_stack only. */ 1942 #define BPF_F_USER_BUILD_ID (1ULL << 11) 1943 1944 /* BPF_FUNC_skb_set_tunnel_key flags. */ 1945 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 1946 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 1947 #define BPF_F_SEQ_NUMBER (1ULL << 3) 1948 1949 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 1950 * BPF_FUNC_perf_event_read_value flags. 1951 */ 1952 #define BPF_F_INDEX_MASK 0xffffffffULL 1953 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 1954 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 1955 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 1956 1957 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 1958 enum bpf_adj_room_mode { 1959 BPF_ADJ_ROOM_NET, 1960 }; 1961 1962 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 1963 enum bpf_hdr_start_off { 1964 BPF_HDR_START_MAC, 1965 BPF_HDR_START_NET, 1966 }; 1967 1968 /* user accessible mirror of in-kernel sk_buff. 1969 * new fields can only be added to the end of this structure 1970 */ 1971 struct __sk_buff { 1972 __u32 len; 1973 __u32 pkt_type; 1974 __u32 mark; 1975 __u32 queue_mapping; 1976 __u32 protocol; 1977 __u32 vlan_present; 1978 __u32 vlan_tci; 1979 __u32 vlan_proto; 1980 __u32 priority; 1981 __u32 ingress_ifindex; 1982 __u32 ifindex; 1983 __u32 tc_index; 1984 __u32 cb[5]; 1985 __u32 hash; 1986 __u32 tc_classid; 1987 __u32 data; 1988 __u32 data_end; 1989 __u32 napi_id; 1990 1991 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 1992 __u32 family; 1993 __u32 remote_ip4; /* Stored in network byte order */ 1994 __u32 local_ip4; /* Stored in network byte order */ 1995 __u32 remote_ip6[4]; /* Stored in network byte order */ 1996 __u32 local_ip6[4]; /* Stored in network byte order */ 1997 __u32 remote_port; /* Stored in network byte order */ 1998 __u32 local_port; /* stored in host byte order */ 1999 /* ... here. */ 2000 2001 __u32 data_meta; 2002 }; 2003 2004 struct bpf_tunnel_key { 2005 __u32 tunnel_id; 2006 union { 2007 __u32 remote_ipv4; 2008 __u32 remote_ipv6[4]; 2009 }; 2010 __u8 tunnel_tos; 2011 __u8 tunnel_ttl; 2012 __u16 tunnel_ext; 2013 __u32 tunnel_label; 2014 }; 2015 2016 /* user accessible mirror of in-kernel xfrm_state. 2017 * new fields can only be added to the end of this structure 2018 */ 2019 struct bpf_xfrm_state { 2020 __u32 reqid; 2021 __u32 spi; /* Stored in network byte order */ 2022 __u16 family; 2023 union { 2024 __u32 remote_ipv4; /* Stored in network byte order */ 2025 __u32 remote_ipv6[4]; /* Stored in network byte order */ 2026 }; 2027 }; 2028 2029 /* Generic BPF return codes which all BPF program types may support. 2030 * The values are binary compatible with their TC_ACT_* counter-part to 2031 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 2032 * programs. 2033 * 2034 * XDP is handled seprately, see XDP_*. 2035 */ 2036 enum bpf_ret_code { 2037 BPF_OK = 0, 2038 /* 1 reserved */ 2039 BPF_DROP = 2, 2040 /* 3-6 reserved */ 2041 BPF_REDIRECT = 7, 2042 /* >127 are reserved for prog type specific return codes */ 2043 }; 2044 2045 struct bpf_sock { 2046 __u32 bound_dev_if; 2047 __u32 family; 2048 __u32 type; 2049 __u32 protocol; 2050 __u32 mark; 2051 __u32 priority; 2052 __u32 src_ip4; /* Allows 1,2,4-byte read. 2053 * Stored in network byte order. 2054 */ 2055 __u32 src_ip6[4]; /* Allows 1,2,4-byte read. 2056 * Stored in network byte order. 2057 */ 2058 __u32 src_port; /* Allows 4-byte read. 2059 * Stored in host byte order 2060 */ 2061 }; 2062 2063 #define XDP_PACKET_HEADROOM 256 2064 2065 /* User return codes for XDP prog type. 2066 * A valid XDP program must return one of these defined values. All other 2067 * return codes are reserved for future use. Unknown return codes will 2068 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 2069 */ 2070 enum xdp_action { 2071 XDP_ABORTED = 0, 2072 XDP_DROP, 2073 XDP_PASS, 2074 XDP_TX, 2075 XDP_REDIRECT, 2076 }; 2077 2078 /* user accessible metadata for XDP packet hook 2079 * new fields must be added to the end of this structure 2080 */ 2081 struct xdp_md { 2082 __u32 data; 2083 __u32 data_end; 2084 __u32 data_meta; 2085 /* Below access go through struct xdp_rxq_info */ 2086 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 2087 __u32 rx_queue_index; /* rxq->queue_index */ 2088 }; 2089 2090 enum sk_action { 2091 SK_DROP = 0, 2092 SK_PASS, 2093 }; 2094 2095 /* user accessible metadata for SK_MSG packet hook, new fields must 2096 * be added to the end of this structure 2097 */ 2098 struct sk_msg_md { 2099 void *data; 2100 void *data_end; 2101 }; 2102 2103 #define BPF_TAG_SIZE 8 2104 2105 struct bpf_prog_info { 2106 __u32 type; 2107 __u32 id; 2108 __u8 tag[BPF_TAG_SIZE]; 2109 __u32 jited_prog_len; 2110 __u32 xlated_prog_len; 2111 __aligned_u64 jited_prog_insns; 2112 __aligned_u64 xlated_prog_insns; 2113 __u64 load_time; /* ns since boottime */ 2114 __u32 created_by_uid; 2115 __u32 nr_map_ids; 2116 __aligned_u64 map_ids; 2117 char name[BPF_OBJ_NAME_LEN]; 2118 __u32 ifindex; 2119 __u32 gpl_compatible:1; 2120 __u64 netns_dev; 2121 __u64 netns_ino; 2122 } __attribute__((aligned(8))); 2123 2124 struct bpf_map_info { 2125 __u32 type; 2126 __u32 id; 2127 __u32 key_size; 2128 __u32 value_size; 2129 __u32 max_entries; 2130 __u32 map_flags; 2131 char name[BPF_OBJ_NAME_LEN]; 2132 __u32 ifindex; 2133 __u64 netns_dev; 2134 __u64 netns_ino; 2135 __u32 btf_id; 2136 __u32 btf_key_id; 2137 __u32 btf_value_id; 2138 } __attribute__((aligned(8))); 2139 2140 struct bpf_btf_info { 2141 __aligned_u64 btf; 2142 __u32 btf_size; 2143 __u32 id; 2144 } __attribute__((aligned(8))); 2145 2146 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 2147 * by user and intended to be used by socket (e.g. to bind to, depends on 2148 * attach attach type). 2149 */ 2150 struct bpf_sock_addr { 2151 __u32 user_family; /* Allows 4-byte read, but no write. */ 2152 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 2153 * Stored in network byte order. 2154 */ 2155 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 2156 * Stored in network byte order. 2157 */ 2158 __u32 user_port; /* Allows 4-byte read and write. 2159 * Stored in network byte order 2160 */ 2161 __u32 family; /* Allows 4-byte read, but no write */ 2162 __u32 type; /* Allows 4-byte read, but no write */ 2163 __u32 protocol; /* Allows 4-byte read, but no write */ 2164 }; 2165 2166 /* User bpf_sock_ops struct to access socket values and specify request ops 2167 * and their replies. 2168 * Some of this fields are in network (bigendian) byte order and may need 2169 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 2170 * New fields can only be added at the end of this structure 2171 */ 2172 struct bpf_sock_ops { 2173 __u32 op; 2174 union { 2175 __u32 args[4]; /* Optionally passed to bpf program */ 2176 __u32 reply; /* Returned by bpf program */ 2177 __u32 replylong[4]; /* Optionally returned by bpf prog */ 2178 }; 2179 __u32 family; 2180 __u32 remote_ip4; /* Stored in network byte order */ 2181 __u32 local_ip4; /* Stored in network byte order */ 2182 __u32 remote_ip6[4]; /* Stored in network byte order */ 2183 __u32 local_ip6[4]; /* Stored in network byte order */ 2184 __u32 remote_port; /* Stored in network byte order */ 2185 __u32 local_port; /* stored in host byte order */ 2186 __u32 is_fullsock; /* Some TCP fields are only valid if 2187 * there is a full socket. If not, the 2188 * fields read as zero. 2189 */ 2190 __u32 snd_cwnd; 2191 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 2192 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 2193 __u32 state; 2194 __u32 rtt_min; 2195 __u32 snd_ssthresh; 2196 __u32 rcv_nxt; 2197 __u32 snd_nxt; 2198 __u32 snd_una; 2199 __u32 mss_cache; 2200 __u32 ecn_flags; 2201 __u32 rate_delivered; 2202 __u32 rate_interval_us; 2203 __u32 packets_out; 2204 __u32 retrans_out; 2205 __u32 total_retrans; 2206 __u32 segs_in; 2207 __u32 data_segs_in; 2208 __u32 segs_out; 2209 __u32 data_segs_out; 2210 __u32 lost_out; 2211 __u32 sacked_out; 2212 __u32 sk_txhash; 2213 __u64 bytes_received; 2214 __u64 bytes_acked; 2215 }; 2216 2217 /* Definitions for bpf_sock_ops_cb_flags */ 2218 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 2219 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 2220 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 2221 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 2222 * supported cb flags 2223 */ 2224 2225 /* List of known BPF sock_ops operators. 2226 * New entries can only be added at the end 2227 */ 2228 enum { 2229 BPF_SOCK_OPS_VOID, 2230 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 2231 * -1 if default value should be used 2232 */ 2233 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 2234 * window (in packets) or -1 if default 2235 * value should be used 2236 */ 2237 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 2238 * active connection is initialized 2239 */ 2240 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 2241 * active connection is 2242 * established 2243 */ 2244 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 2245 * passive connection is 2246 * established 2247 */ 2248 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 2249 * needs ECN 2250 */ 2251 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 2252 * based on the path and may be 2253 * dependent on the congestion control 2254 * algorithm. In general it indicates 2255 * a congestion threshold. RTTs above 2256 * this indicate congestion 2257 */ 2258 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 2259 * Arg1: value of icsk_retransmits 2260 * Arg2: value of icsk_rto 2261 * Arg3: whether RTO has expired 2262 */ 2263 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 2264 * Arg1: sequence number of 1st byte 2265 * Arg2: # segments 2266 * Arg3: return value of 2267 * tcp_transmit_skb (0 => success) 2268 */ 2269 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 2270 * Arg1: old_state 2271 * Arg2: new_state 2272 */ 2273 }; 2274 2275 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 2276 * changes between the TCP and BPF versions. Ideally this should never happen. 2277 * If it does, we need to add code to convert them before calling 2278 * the BPF sock_ops function. 2279 */ 2280 enum { 2281 BPF_TCP_ESTABLISHED = 1, 2282 BPF_TCP_SYN_SENT, 2283 BPF_TCP_SYN_RECV, 2284 BPF_TCP_FIN_WAIT1, 2285 BPF_TCP_FIN_WAIT2, 2286 BPF_TCP_TIME_WAIT, 2287 BPF_TCP_CLOSE, 2288 BPF_TCP_CLOSE_WAIT, 2289 BPF_TCP_LAST_ACK, 2290 BPF_TCP_LISTEN, 2291 BPF_TCP_CLOSING, /* Now a valid state */ 2292 BPF_TCP_NEW_SYN_RECV, 2293 2294 BPF_TCP_MAX_STATES /* Leave at the end! */ 2295 }; 2296 2297 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 2298 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 2299 2300 struct bpf_perf_event_value { 2301 __u64 counter; 2302 __u64 enabled; 2303 __u64 running; 2304 }; 2305 2306 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 2307 #define BPF_DEVCG_ACC_READ (1ULL << 1) 2308 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 2309 2310 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 2311 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 2312 2313 struct bpf_cgroup_dev_ctx { 2314 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 2315 __u32 access_type; 2316 __u32 major; 2317 __u32 minor; 2318 }; 2319 2320 struct bpf_raw_tracepoint_args { 2321 __u64 args[0]; 2322 }; 2323 2324 #endif /* _UAPI__LINUX_BPF_H__ */ 2325