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