1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8#ifndef _UAPI__LINUX_BPF_H__ 9#define _UAPI__LINUX_BPF_H__ 10 11#include <linux/types.h> 12#include <linux/bpf_common.h> 13 14/* Extended instruction set based on top of classic BPF */ 15 16/* instruction classes */ 17#define BPF_JMP32 0x06 /* jmp mode in word width */ 18#define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20/* ld/ldx fields */ 21#define BPF_DW 0x18 /* double word (64-bit) */ 22#define BPF_XADD 0xc0 /* exclusive add */ 23 24/* alu/jmp fields */ 25#define BPF_MOV 0xb0 /* mov reg to reg */ 26#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28/* change endianness of a register */ 29#define BPF_END 0xd0 /* flags for endianness conversion: */ 30#define BPF_TO_LE 0x00 /* convert to little-endian */ 31#define BPF_TO_BE 0x08 /* convert to big-endian */ 32#define BPF_FROM_LE BPF_TO_LE 33#define BPF_FROM_BE BPF_TO_BE 34 35/* jmp encodings */ 36#define BPF_JNE 0x50 /* jump != */ 37#define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41#define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43#define BPF_CALL 0x80 /* function call */ 44#define BPF_EXIT 0x90 /* function return */ 45 46/* Register numbers */ 47enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60}; 61 62/* BPF has 10 general purpose 64-bit registers and stack frame. */ 63#define MAX_BPF_REG __MAX_BPF_REG 64 65struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71}; 72 73/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77}; 78 79struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82}; 83 84/* BPF syscall commands, see bpf(2) man-page for details. */ 85enum bpf_cmd { 86 BPF_MAP_CREATE, 87 BPF_MAP_LOOKUP_ELEM, 88 BPF_MAP_UPDATE_ELEM, 89 BPF_MAP_DELETE_ELEM, 90 BPF_MAP_GET_NEXT_KEY, 91 BPF_PROG_LOAD, 92 BPF_OBJ_PIN, 93 BPF_OBJ_GET, 94 BPF_PROG_ATTACH, 95 BPF_PROG_DETACH, 96 BPF_PROG_TEST_RUN, 97 BPF_PROG_GET_NEXT_ID, 98 BPF_MAP_GET_NEXT_ID, 99 BPF_PROG_GET_FD_BY_ID, 100 BPF_MAP_GET_FD_BY_ID, 101 BPF_OBJ_GET_INFO_BY_FD, 102 BPF_PROG_QUERY, 103 BPF_RAW_TRACEPOINT_OPEN, 104 BPF_BTF_LOAD, 105 BPF_BTF_GET_FD_BY_ID, 106 BPF_TASK_FD_QUERY, 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 108 BPF_MAP_FREEZE, 109 BPF_BTF_GET_NEXT_ID, 110 BPF_MAP_LOOKUP_BATCH, 111 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 112 BPF_MAP_UPDATE_BATCH, 113 BPF_MAP_DELETE_BATCH, 114}; 115 116enum bpf_map_type { 117 BPF_MAP_TYPE_UNSPEC, 118 BPF_MAP_TYPE_HASH, 119 BPF_MAP_TYPE_ARRAY, 120 BPF_MAP_TYPE_PROG_ARRAY, 121 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 122 BPF_MAP_TYPE_PERCPU_HASH, 123 BPF_MAP_TYPE_PERCPU_ARRAY, 124 BPF_MAP_TYPE_STACK_TRACE, 125 BPF_MAP_TYPE_CGROUP_ARRAY, 126 BPF_MAP_TYPE_LRU_HASH, 127 BPF_MAP_TYPE_LRU_PERCPU_HASH, 128 BPF_MAP_TYPE_LPM_TRIE, 129 BPF_MAP_TYPE_ARRAY_OF_MAPS, 130 BPF_MAP_TYPE_HASH_OF_MAPS, 131 BPF_MAP_TYPE_DEVMAP, 132 BPF_MAP_TYPE_SOCKMAP, 133 BPF_MAP_TYPE_CPUMAP, 134 BPF_MAP_TYPE_XSKMAP, 135 BPF_MAP_TYPE_SOCKHASH, 136 BPF_MAP_TYPE_CGROUP_STORAGE, 137 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 138 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 139 BPF_MAP_TYPE_QUEUE, 140 BPF_MAP_TYPE_STACK, 141 BPF_MAP_TYPE_SK_STORAGE, 142 BPF_MAP_TYPE_DEVMAP_HASH, 143 BPF_MAP_TYPE_STRUCT_OPS, 144}; 145 146/* Note that tracing related programs such as 147 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 148 * are not subject to a stable API since kernel internal data 149 * structures can change from release to release and may 150 * therefore break existing tracing BPF programs. Tracing BPF 151 * programs correspond to /a/ specific kernel which is to be 152 * analyzed, and not /a/ specific kernel /and/ all future ones. 153 */ 154enum bpf_prog_type { 155 BPF_PROG_TYPE_UNSPEC, 156 BPF_PROG_TYPE_SOCKET_FILTER, 157 BPF_PROG_TYPE_KPROBE, 158 BPF_PROG_TYPE_SCHED_CLS, 159 BPF_PROG_TYPE_SCHED_ACT, 160 BPF_PROG_TYPE_TRACEPOINT, 161 BPF_PROG_TYPE_XDP, 162 BPF_PROG_TYPE_PERF_EVENT, 163 BPF_PROG_TYPE_CGROUP_SKB, 164 BPF_PROG_TYPE_CGROUP_SOCK, 165 BPF_PROG_TYPE_LWT_IN, 166 BPF_PROG_TYPE_LWT_OUT, 167 BPF_PROG_TYPE_LWT_XMIT, 168 BPF_PROG_TYPE_SOCK_OPS, 169 BPF_PROG_TYPE_SK_SKB, 170 BPF_PROG_TYPE_CGROUP_DEVICE, 171 BPF_PROG_TYPE_SK_MSG, 172 BPF_PROG_TYPE_RAW_TRACEPOINT, 173 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 174 BPF_PROG_TYPE_LWT_SEG6LOCAL, 175 BPF_PROG_TYPE_LIRC_MODE2, 176 BPF_PROG_TYPE_SK_REUSEPORT, 177 BPF_PROG_TYPE_FLOW_DISSECTOR, 178 BPF_PROG_TYPE_CGROUP_SYSCTL, 179 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 180 BPF_PROG_TYPE_CGROUP_SOCKOPT, 181 BPF_PROG_TYPE_TRACING, 182 BPF_PROG_TYPE_STRUCT_OPS, 183 BPF_PROG_TYPE_EXT, 184}; 185 186enum bpf_attach_type { 187 BPF_CGROUP_INET_INGRESS, 188 BPF_CGROUP_INET_EGRESS, 189 BPF_CGROUP_INET_SOCK_CREATE, 190 BPF_CGROUP_SOCK_OPS, 191 BPF_SK_SKB_STREAM_PARSER, 192 BPF_SK_SKB_STREAM_VERDICT, 193 BPF_CGROUP_DEVICE, 194 BPF_SK_MSG_VERDICT, 195 BPF_CGROUP_INET4_BIND, 196 BPF_CGROUP_INET6_BIND, 197 BPF_CGROUP_INET4_CONNECT, 198 BPF_CGROUP_INET6_CONNECT, 199 BPF_CGROUP_INET4_POST_BIND, 200 BPF_CGROUP_INET6_POST_BIND, 201 BPF_CGROUP_UDP4_SENDMSG, 202 BPF_CGROUP_UDP6_SENDMSG, 203 BPF_LIRC_MODE2, 204 BPF_FLOW_DISSECTOR, 205 BPF_CGROUP_SYSCTL, 206 BPF_CGROUP_UDP4_RECVMSG, 207 BPF_CGROUP_UDP6_RECVMSG, 208 BPF_CGROUP_GETSOCKOPT, 209 BPF_CGROUP_SETSOCKOPT, 210 BPF_TRACE_RAW_TP, 211 BPF_TRACE_FENTRY, 212 BPF_TRACE_FEXIT, 213 __MAX_BPF_ATTACH_TYPE 214}; 215 216#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 217 218/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 219 * 220 * NONE(default): No further bpf programs allowed in the subtree. 221 * 222 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 223 * the program in this cgroup yields to sub-cgroup program. 224 * 225 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 226 * that cgroup program gets run in addition to the program in this cgroup. 227 * 228 * Only one program is allowed to be attached to a cgroup with 229 * NONE or BPF_F_ALLOW_OVERRIDE flag. 230 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 231 * release old program and attach the new one. Attach flags has to match. 232 * 233 * Multiple programs are allowed to be attached to a cgroup with 234 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 235 * (those that were attached first, run first) 236 * The programs of sub-cgroup are executed first, then programs of 237 * this cgroup and then programs of parent cgroup. 238 * When children program makes decision (like picking TCP CA or sock bind) 239 * parent program has a chance to override it. 240 * 241 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 242 * programs for a cgroup. Though it's possible to replace an old program at 243 * any position by also specifying BPF_F_REPLACE flag and position itself in 244 * replace_bpf_fd attribute. Old program at this position will be released. 245 * 246 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 247 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 248 * Ex1: 249 * cgrp1 (MULTI progs A, B) -> 250 * cgrp2 (OVERRIDE prog C) -> 251 * cgrp3 (MULTI prog D) -> 252 * cgrp4 (OVERRIDE prog E) -> 253 * cgrp5 (NONE prog F) 254 * the event in cgrp5 triggers execution of F,D,A,B in that order. 255 * if prog F is detached, the execution is E,D,A,B 256 * if prog F and D are detached, the execution is E,A,B 257 * if prog F, E and D are detached, the execution is C,A,B 258 * 259 * All eligible programs are executed regardless of return code from 260 * earlier programs. 261 */ 262#define BPF_F_ALLOW_OVERRIDE (1U << 0) 263#define BPF_F_ALLOW_MULTI (1U << 1) 264#define BPF_F_REPLACE (1U << 2) 265 266/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 267 * verifier will perform strict alignment checking as if the kernel 268 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 269 * and NET_IP_ALIGN defined to 2. 270 */ 271#define BPF_F_STRICT_ALIGNMENT (1U << 0) 272 273/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 274 * verifier will allow any alignment whatsoever. On platforms 275 * with strict alignment requirements for loads ands stores (such 276 * as sparc and mips) the verifier validates that all loads and 277 * stores provably follow this requirement. This flag turns that 278 * checking and enforcement off. 279 * 280 * It is mostly used for testing when we want to validate the 281 * context and memory access aspects of the verifier, but because 282 * of an unaligned access the alignment check would trigger before 283 * the one we are interested in. 284 */ 285#define BPF_F_ANY_ALIGNMENT (1U << 1) 286 287/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 288 * Verifier does sub-register def/use analysis and identifies instructions whose 289 * def only matters for low 32-bit, high 32-bit is never referenced later 290 * through implicit zero extension. Therefore verifier notifies JIT back-ends 291 * that it is safe to ignore clearing high 32-bit for these instructions. This 292 * saves some back-ends a lot of code-gen. However such optimization is not 293 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 294 * hence hasn't used verifier's analysis result. But, we really want to have a 295 * way to be able to verify the correctness of the described optimization on 296 * x86_64 on which testsuites are frequently exercised. 297 * 298 * So, this flag is introduced. Once it is set, verifier will randomize high 299 * 32-bit for those instructions who has been identified as safe to ignore them. 300 * Then, if verifier is not doing correct analysis, such randomization will 301 * regress tests to expose bugs. 302 */ 303#define BPF_F_TEST_RND_HI32 (1U << 2) 304 305/* The verifier internal test flag. Behavior is undefined */ 306#define BPF_F_TEST_STATE_FREQ (1U << 3) 307 308/* When BPF ldimm64's insn[0].src_reg != 0 then this can have 309 * two extensions: 310 * 311 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 312 * insn[0].imm: map fd map fd 313 * insn[1].imm: 0 offset into value 314 * insn[0].off: 0 0 315 * insn[1].off: 0 0 316 * ldimm64 rewrite: address of map address of map[0]+offset 317 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 318 */ 319#define BPF_PSEUDO_MAP_FD 1 320#define BPF_PSEUDO_MAP_VALUE 2 321 322/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 323 * offset to another bpf function 324 */ 325#define BPF_PSEUDO_CALL 1 326 327/* flags for BPF_MAP_UPDATE_ELEM command */ 328#define BPF_ANY 0 /* create new element or update existing */ 329#define BPF_NOEXIST 1 /* create new element if it didn't exist */ 330#define BPF_EXIST 2 /* update existing element */ 331#define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 332 333/* flags for BPF_MAP_CREATE command */ 334#define BPF_F_NO_PREALLOC (1U << 0) 335/* Instead of having one common LRU list in the 336 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 337 * which can scale and perform better. 338 * Note, the LRU nodes (including free nodes) cannot be moved 339 * across different LRU lists. 340 */ 341#define BPF_F_NO_COMMON_LRU (1U << 1) 342/* Specify numa node during map creation */ 343#define BPF_F_NUMA_NODE (1U << 2) 344 345#define BPF_OBJ_NAME_LEN 16U 346 347/* Flags for accessing BPF object from syscall side. */ 348#define BPF_F_RDONLY (1U << 3) 349#define BPF_F_WRONLY (1U << 4) 350 351/* Flag for stack_map, store build_id+offset instead of pointer */ 352#define BPF_F_STACK_BUILD_ID (1U << 5) 353 354/* Zero-initialize hash function seed. This should only be used for testing. */ 355#define BPF_F_ZERO_SEED (1U << 6) 356 357/* Flags for accessing BPF object from program side. */ 358#define BPF_F_RDONLY_PROG (1U << 7) 359#define BPF_F_WRONLY_PROG (1U << 8) 360 361/* Clone map from listener for newly accepted socket */ 362#define BPF_F_CLONE (1U << 9) 363 364/* Enable memory-mapping BPF map */ 365#define BPF_F_MMAPABLE (1U << 10) 366 367/* Flags for BPF_PROG_QUERY. */ 368 369/* Query effective (directly attached + inherited from ancestor cgroups) 370 * programs that will be executed for events within a cgroup. 371 * attach_flags with this flag are returned only for directly attached programs. 372 */ 373#define BPF_F_QUERY_EFFECTIVE (1U << 0) 374 375enum bpf_stack_build_id_status { 376 /* user space need an empty entry to identify end of a trace */ 377 BPF_STACK_BUILD_ID_EMPTY = 0, 378 /* with valid build_id and offset */ 379 BPF_STACK_BUILD_ID_VALID = 1, 380 /* couldn't get build_id, fallback to ip */ 381 BPF_STACK_BUILD_ID_IP = 2, 382}; 383 384#define BPF_BUILD_ID_SIZE 20 385struct bpf_stack_build_id { 386 __s32 status; 387 unsigned char build_id[BPF_BUILD_ID_SIZE]; 388 union { 389 __u64 offset; 390 __u64 ip; 391 }; 392}; 393 394union bpf_attr { 395 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 396 __u32 map_type; /* one of enum bpf_map_type */ 397 __u32 key_size; /* size of key in bytes */ 398 __u32 value_size; /* size of value in bytes */ 399 __u32 max_entries; /* max number of entries in a map */ 400 __u32 map_flags; /* BPF_MAP_CREATE related 401 * flags defined above. 402 */ 403 __u32 inner_map_fd; /* fd pointing to the inner map */ 404 __u32 numa_node; /* numa node (effective only if 405 * BPF_F_NUMA_NODE is set). 406 */ 407 char map_name[BPF_OBJ_NAME_LEN]; 408 __u32 map_ifindex; /* ifindex of netdev to create on */ 409 __u32 btf_fd; /* fd pointing to a BTF type data */ 410 __u32 btf_key_type_id; /* BTF type_id of the key */ 411 __u32 btf_value_type_id; /* BTF type_id of the value */ 412 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 413 * struct stored as the 414 * map value 415 */ 416 }; 417 418 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 419 __u32 map_fd; 420 __aligned_u64 key; 421 union { 422 __aligned_u64 value; 423 __aligned_u64 next_key; 424 }; 425 __u64 flags; 426 }; 427 428 struct { /* struct used by BPF_MAP_*_BATCH commands */ 429 __aligned_u64 in_batch; /* start batch, 430 * NULL to start from beginning 431 */ 432 __aligned_u64 out_batch; /* output: next start batch */ 433 __aligned_u64 keys; 434 __aligned_u64 values; 435 __u32 count; /* input/output: 436 * input: # of key/value 437 * elements 438 * output: # of filled elements 439 */ 440 __u32 map_fd; 441 __u64 elem_flags; 442 __u64 flags; 443 } batch; 444 445 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 446 __u32 prog_type; /* one of enum bpf_prog_type */ 447 __u32 insn_cnt; 448 __aligned_u64 insns; 449 __aligned_u64 license; 450 __u32 log_level; /* verbosity level of verifier */ 451 __u32 log_size; /* size of user buffer */ 452 __aligned_u64 log_buf; /* user supplied buffer */ 453 __u32 kern_version; /* not used */ 454 __u32 prog_flags; 455 char prog_name[BPF_OBJ_NAME_LEN]; 456 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 457 /* For some prog types expected attach type must be known at 458 * load time to verify attach type specific parts of prog 459 * (context accesses, allowed helpers, etc). 460 */ 461 __u32 expected_attach_type; 462 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 463 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 464 __aligned_u64 func_info; /* func info */ 465 __u32 func_info_cnt; /* number of bpf_func_info records */ 466 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 467 __aligned_u64 line_info; /* line info */ 468 __u32 line_info_cnt; /* number of bpf_line_info records */ 469 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 470 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 471 }; 472 473 struct { /* anonymous struct used by BPF_OBJ_* commands */ 474 __aligned_u64 pathname; 475 __u32 bpf_fd; 476 __u32 file_flags; 477 }; 478 479 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 480 __u32 target_fd; /* container object to attach to */ 481 __u32 attach_bpf_fd; /* eBPF program to attach */ 482 __u32 attach_type; 483 __u32 attach_flags; 484 __u32 replace_bpf_fd; /* previously attached eBPF 485 * program to replace if 486 * BPF_F_REPLACE is used 487 */ 488 }; 489 490 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 491 __u32 prog_fd; 492 __u32 retval; 493 __u32 data_size_in; /* input: len of data_in */ 494 __u32 data_size_out; /* input/output: len of data_out 495 * returns ENOSPC if data_out 496 * is too small. 497 */ 498 __aligned_u64 data_in; 499 __aligned_u64 data_out; 500 __u32 repeat; 501 __u32 duration; 502 __u32 ctx_size_in; /* input: len of ctx_in */ 503 __u32 ctx_size_out; /* input/output: len of ctx_out 504 * returns ENOSPC if ctx_out 505 * is too small. 506 */ 507 __aligned_u64 ctx_in; 508 __aligned_u64 ctx_out; 509 } test; 510 511 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 512 union { 513 __u32 start_id; 514 __u32 prog_id; 515 __u32 map_id; 516 __u32 btf_id; 517 }; 518 __u32 next_id; 519 __u32 open_flags; 520 }; 521 522 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 523 __u32 bpf_fd; 524 __u32 info_len; 525 __aligned_u64 info; 526 } info; 527 528 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 529 __u32 target_fd; /* container object to query */ 530 __u32 attach_type; 531 __u32 query_flags; 532 __u32 attach_flags; 533 __aligned_u64 prog_ids; 534 __u32 prog_cnt; 535 } query; 536 537 struct { 538 __u64 name; 539 __u32 prog_fd; 540 } raw_tracepoint; 541 542 struct { /* anonymous struct for BPF_BTF_LOAD */ 543 __aligned_u64 btf; 544 __aligned_u64 btf_log_buf; 545 __u32 btf_size; 546 __u32 btf_log_size; 547 __u32 btf_log_level; 548 }; 549 550 struct { 551 __u32 pid; /* input: pid */ 552 __u32 fd; /* input: fd */ 553 __u32 flags; /* input: flags */ 554 __u32 buf_len; /* input/output: buf len */ 555 __aligned_u64 buf; /* input/output: 556 * tp_name for tracepoint 557 * symbol for kprobe 558 * filename for uprobe 559 */ 560 __u32 prog_id; /* output: prod_id */ 561 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 562 __u64 probe_offset; /* output: probe_offset */ 563 __u64 probe_addr; /* output: probe_addr */ 564 } task_fd_query; 565} __attribute__((aligned(8))); 566 567/* The description below is an attempt at providing documentation to eBPF 568 * developers about the multiple available eBPF helper functions. It can be 569 * parsed and used to produce a manual page. The workflow is the following, 570 * and requires the rst2man utility: 571 * 572 * $ ./scripts/bpf_helpers_doc.py \ 573 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 574 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 575 * $ man /tmp/bpf-helpers.7 576 * 577 * Note that in order to produce this external documentation, some RST 578 * formatting is used in the descriptions to get "bold" and "italics" in 579 * manual pages. Also note that the few trailing white spaces are 580 * intentional, removing them would break paragraphs for rst2man. 581 * 582 * Start of BPF helper function descriptions: 583 * 584 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 585 * Description 586 * Perform a lookup in *map* for an entry associated to *key*. 587 * Return 588 * Map value associated to *key*, or **NULL** if no entry was 589 * found. 590 * 591 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 592 * Description 593 * Add or update the value of the entry associated to *key* in 594 * *map* with *value*. *flags* is one of: 595 * 596 * **BPF_NOEXIST** 597 * The entry for *key* must not exist in the map. 598 * **BPF_EXIST** 599 * The entry for *key* must already exist in the map. 600 * **BPF_ANY** 601 * No condition on the existence of the entry for *key*. 602 * 603 * Flag value **BPF_NOEXIST** cannot be used for maps of types 604 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 605 * elements always exist), the helper would return an error. 606 * Return 607 * 0 on success, or a negative error in case of failure. 608 * 609 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 610 * Description 611 * Delete entry with *key* from *map*. 612 * Return 613 * 0 on success, or a negative error in case of failure. 614 * 615 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 616 * Description 617 * For tracing programs, safely attempt to read *size* bytes from 618 * kernel space address *unsafe_ptr* and store the data in *dst*. 619 * 620 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel() 621 * instead. 622 * Return 623 * 0 on success, or a negative error in case of failure. 624 * 625 * u64 bpf_ktime_get_ns(void) 626 * Description 627 * Return the time elapsed since system boot, in nanoseconds. 628 * Return 629 * Current *ktime*. 630 * 631 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 632 * Description 633 * This helper is a "printk()-like" facility for debugging. It 634 * prints a message defined by format *fmt* (of size *fmt_size*) 635 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 636 * available. It can take up to three additional **u64** 637 * arguments (as an eBPF helpers, the total number of arguments is 638 * limited to five). 639 * 640 * Each time the helper is called, it appends a line to the trace. 641 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 642 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 643 * The format of the trace is customizable, and the exact output 644 * one will get depends on the options set in 645 * *\/sys/kernel/debug/tracing/trace_options* (see also the 646 * *README* file under the same directory). However, it usually 647 * defaults to something like: 648 * 649 * :: 650 * 651 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 652 * 653 * In the above: 654 * 655 * * ``telnet`` is the name of the current task. 656 * * ``470`` is the PID of the current task. 657 * * ``001`` is the CPU number on which the task is 658 * running. 659 * * In ``.N..``, each character refers to a set of 660 * options (whether irqs are enabled, scheduling 661 * options, whether hard/softirqs are running, level of 662 * preempt_disabled respectively). **N** means that 663 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 664 * are set. 665 * * ``419421.045894`` is a timestamp. 666 * * ``0x00000001`` is a fake value used by BPF for the 667 * instruction pointer register. 668 * * ``<formatted msg>`` is the message formatted with 669 * *fmt*. 670 * 671 * The conversion specifiers supported by *fmt* are similar, but 672 * more limited than for printk(). They are **%d**, **%i**, 673 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 674 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 675 * of field, padding with zeroes, etc.) is available, and the 676 * helper will return **-EINVAL** (but print nothing) if it 677 * encounters an unknown specifier. 678 * 679 * Also, note that **bpf_trace_printk**\ () is slow, and should 680 * only be used for debugging purposes. For this reason, a notice 681 * bloc (spanning several lines) is printed to kernel logs and 682 * states that the helper should not be used "for production use" 683 * the first time this helper is used (or more precisely, when 684 * **trace_printk**\ () buffers are allocated). For passing values 685 * to user space, perf events should be preferred. 686 * Return 687 * The number of bytes written to the buffer, or a negative error 688 * in case of failure. 689 * 690 * u32 bpf_get_prandom_u32(void) 691 * Description 692 * Get a pseudo-random number. 693 * 694 * From a security point of view, this helper uses its own 695 * pseudo-random internal state, and cannot be used to infer the 696 * seed of other random functions in the kernel. However, it is 697 * essential to note that the generator used by the helper is not 698 * cryptographically secure. 699 * Return 700 * A random 32-bit unsigned value. 701 * 702 * u32 bpf_get_smp_processor_id(void) 703 * Description 704 * Get the SMP (symmetric multiprocessing) processor id. Note that 705 * all programs run with preemption disabled, which means that the 706 * SMP processor id is stable during all the execution of the 707 * program. 708 * Return 709 * The SMP id of the processor running the program. 710 * 711 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 712 * Description 713 * Store *len* bytes from address *from* into the packet 714 * associated to *skb*, at *offset*. *flags* are a combination of 715 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 716 * checksum for the packet after storing the bytes) and 717 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 718 * **->swhash** and *skb*\ **->l4hash** to 0). 719 * 720 * A call to this helper is susceptible to change the underlying 721 * packet buffer. Therefore, at load time, all checks on pointers 722 * previously done by the verifier are invalidated and must be 723 * performed again, if the helper is used in combination with 724 * direct packet access. 725 * Return 726 * 0 on success, or a negative error in case of failure. 727 * 728 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 729 * Description 730 * Recompute the layer 3 (e.g. IP) checksum for the packet 731 * associated to *skb*. Computation is incremental, so the helper 732 * must know the former value of the header field that was 733 * modified (*from*), the new value of this field (*to*), and the 734 * number of bytes (2 or 4) for this field, stored in *size*. 735 * Alternatively, it is possible to store the difference between 736 * the previous and the new values of the header field in *to*, by 737 * setting *from* and *size* to 0. For both methods, *offset* 738 * indicates the location of the IP checksum within the packet. 739 * 740 * This helper works in combination with **bpf_csum_diff**\ (), 741 * which does not update the checksum in-place, but offers more 742 * flexibility and can handle sizes larger than 2 or 4 for the 743 * checksum to update. 744 * 745 * A call to this helper is susceptible to change the underlying 746 * packet buffer. Therefore, at load time, all checks on pointers 747 * previously done by the verifier are invalidated and must be 748 * performed again, if the helper is used in combination with 749 * direct packet access. 750 * Return 751 * 0 on success, or a negative error in case of failure. 752 * 753 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 754 * Description 755 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 756 * packet associated to *skb*. Computation is incremental, so the 757 * helper must know the former value of the header field that was 758 * modified (*from*), the new value of this field (*to*), and the 759 * number of bytes (2 or 4) for this field, stored on the lowest 760 * four bits of *flags*. Alternatively, it is possible to store 761 * the difference between the previous and the new values of the 762 * header field in *to*, by setting *from* and the four lowest 763 * bits of *flags* to 0. For both methods, *offset* indicates the 764 * location of the IP checksum within the packet. In addition to 765 * the size of the field, *flags* can be added (bitwise OR) actual 766 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 767 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 768 * for updates resulting in a null checksum the value is set to 769 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 770 * the checksum is to be computed against a pseudo-header. 771 * 772 * This helper works in combination with **bpf_csum_diff**\ (), 773 * which does not update the checksum in-place, but offers more 774 * flexibility and can handle sizes larger than 2 or 4 for the 775 * checksum to update. 776 * 777 * A call to this helper is susceptible to change the underlying 778 * packet buffer. Therefore, at load time, all checks on pointers 779 * previously done by the verifier are invalidated and must be 780 * performed again, if the helper is used in combination with 781 * direct packet access. 782 * Return 783 * 0 on success, or a negative error in case of failure. 784 * 785 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 786 * Description 787 * This special helper is used to trigger a "tail call", or in 788 * other words, to jump into another eBPF program. The same stack 789 * frame is used (but values on stack and in registers for the 790 * caller are not accessible to the callee). This mechanism allows 791 * for program chaining, either for raising the maximum number of 792 * available eBPF instructions, or to execute given programs in 793 * conditional blocks. For security reasons, there is an upper 794 * limit to the number of successive tail calls that can be 795 * performed. 796 * 797 * Upon call of this helper, the program attempts to jump into a 798 * program referenced at index *index* in *prog_array_map*, a 799 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 800 * *ctx*, a pointer to the context. 801 * 802 * If the call succeeds, the kernel immediately runs the first 803 * instruction of the new program. This is not a function call, 804 * and it never returns to the previous program. If the call 805 * fails, then the helper has no effect, and the caller continues 806 * to run its subsequent instructions. A call can fail if the 807 * destination program for the jump does not exist (i.e. *index* 808 * is superior to the number of entries in *prog_array_map*), or 809 * if the maximum number of tail calls has been reached for this 810 * chain of programs. This limit is defined in the kernel by the 811 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 812 * which is currently set to 32. 813 * Return 814 * 0 on success, or a negative error in case of failure. 815 * 816 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 817 * Description 818 * Clone and redirect the packet associated to *skb* to another 819 * net device of index *ifindex*. Both ingress and egress 820 * interfaces can be used for redirection. The **BPF_F_INGRESS** 821 * value in *flags* is used to make the distinction (ingress path 822 * is selected if the flag is present, egress path otherwise). 823 * This is the only flag supported for now. 824 * 825 * In comparison with **bpf_redirect**\ () helper, 826 * **bpf_clone_redirect**\ () has the associated cost of 827 * duplicating the packet buffer, but this can be executed out of 828 * the eBPF program. Conversely, **bpf_redirect**\ () is more 829 * efficient, but it is handled through an action code where the 830 * redirection happens only after the eBPF program has returned. 831 * 832 * A call to this helper is susceptible to change the underlying 833 * packet buffer. Therefore, at load time, all checks on pointers 834 * previously done by the verifier are invalidated and must be 835 * performed again, if the helper is used in combination with 836 * direct packet access. 837 * Return 838 * 0 on success, or a negative error in case of failure. 839 * 840 * u64 bpf_get_current_pid_tgid(void) 841 * Return 842 * A 64-bit integer containing the current tgid and pid, and 843 * created as such: 844 * *current_task*\ **->tgid << 32 \|** 845 * *current_task*\ **->pid**. 846 * 847 * u64 bpf_get_current_uid_gid(void) 848 * Return 849 * A 64-bit integer containing the current GID and UID, and 850 * created as such: *current_gid* **<< 32 \|** *current_uid*. 851 * 852 * int bpf_get_current_comm(void *buf, u32 size_of_buf) 853 * Description 854 * Copy the **comm** attribute of the current task into *buf* of 855 * *size_of_buf*. The **comm** attribute contains the name of 856 * the executable (excluding the path) for the current task. The 857 * *size_of_buf* must be strictly positive. On success, the 858 * helper makes sure that the *buf* is NUL-terminated. On failure, 859 * it is filled with zeroes. 860 * Return 861 * 0 on success, or a negative error in case of failure. 862 * 863 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 864 * Description 865 * Retrieve the classid for the current task, i.e. for the net_cls 866 * cgroup to which *skb* belongs. 867 * 868 * This helper can be used on TC egress path, but not on ingress. 869 * 870 * The net_cls cgroup provides an interface to tag network packets 871 * based on a user-provided identifier for all traffic coming from 872 * the tasks belonging to the related cgroup. See also the related 873 * kernel documentation, available from the Linux sources in file 874 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 875 * 876 * The Linux kernel has two versions for cgroups: there are 877 * cgroups v1 and cgroups v2. Both are available to users, who can 878 * use a mixture of them, but note that the net_cls cgroup is for 879 * cgroup v1 only. This makes it incompatible with BPF programs 880 * run on cgroups, which is a cgroup-v2-only feature (a socket can 881 * only hold data for one version of cgroups at a time). 882 * 883 * This helper is only available is the kernel was compiled with 884 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 885 * "**y**" or to "**m**". 886 * Return 887 * The classid, or 0 for the default unconfigured classid. 888 * 889 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 890 * Description 891 * Push a *vlan_tci* (VLAN tag control information) of protocol 892 * *vlan_proto* to the packet associated to *skb*, then update 893 * the checksum. Note that if *vlan_proto* is different from 894 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 895 * be **ETH_P_8021Q**. 896 * 897 * A call to this helper is susceptible to change the underlying 898 * packet buffer. Therefore, at load time, all checks on pointers 899 * previously done by the verifier are invalidated and must be 900 * performed again, if the helper is used in combination with 901 * direct packet access. 902 * Return 903 * 0 on success, or a negative error in case of failure. 904 * 905 * int bpf_skb_vlan_pop(struct sk_buff *skb) 906 * Description 907 * Pop a VLAN header from the packet associated to *skb*. 908 * 909 * A call to this helper is susceptible to change the underlying 910 * packet buffer. Therefore, at load time, all checks on pointers 911 * previously done by the verifier are invalidated and must be 912 * performed again, if the helper is used in combination with 913 * direct packet access. 914 * Return 915 * 0 on success, or a negative error in case of failure. 916 * 917 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 918 * Description 919 * Get tunnel metadata. This helper takes a pointer *key* to an 920 * empty **struct bpf_tunnel_key** of **size**, that will be 921 * filled with tunnel metadata for the packet associated to *skb*. 922 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 923 * indicates that the tunnel is based on IPv6 protocol instead of 924 * IPv4. 925 * 926 * The **struct bpf_tunnel_key** is an object that generalizes the 927 * principal parameters used by various tunneling protocols into a 928 * single struct. This way, it can be used to easily make a 929 * decision based on the contents of the encapsulation header, 930 * "summarized" in this struct. In particular, it holds the IP 931 * address of the remote end (IPv4 or IPv6, depending on the case) 932 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 933 * this struct exposes the *key*\ **->tunnel_id**, which is 934 * generally mapped to a VNI (Virtual Network Identifier), making 935 * it programmable together with the **bpf_skb_set_tunnel_key**\ 936 * () helper. 937 * 938 * Let's imagine that the following code is part of a program 939 * attached to the TC ingress interface, on one end of a GRE 940 * tunnel, and is supposed to filter out all messages coming from 941 * remote ends with IPv4 address other than 10.0.0.1: 942 * 943 * :: 944 * 945 * int ret; 946 * struct bpf_tunnel_key key = {}; 947 * 948 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 949 * if (ret < 0) 950 * return TC_ACT_SHOT; // drop packet 951 * 952 * if (key.remote_ipv4 != 0x0a000001) 953 * return TC_ACT_SHOT; // drop packet 954 * 955 * return TC_ACT_OK; // accept packet 956 * 957 * This interface can also be used with all encapsulation devices 958 * that can operate in "collect metadata" mode: instead of having 959 * one network device per specific configuration, the "collect 960 * metadata" mode only requires a single device where the 961 * configuration can be extracted from this helper. 962 * 963 * This can be used together with various tunnels such as VXLan, 964 * Geneve, GRE or IP in IP (IPIP). 965 * Return 966 * 0 on success, or a negative error in case of failure. 967 * 968 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 969 * Description 970 * Populate tunnel metadata for packet associated to *skb.* The 971 * tunnel metadata is set to the contents of *key*, of *size*. The 972 * *flags* can be set to a combination of the following values: 973 * 974 * **BPF_F_TUNINFO_IPV6** 975 * Indicate that the tunnel is based on IPv6 protocol 976 * instead of IPv4. 977 * **BPF_F_ZERO_CSUM_TX** 978 * For IPv4 packets, add a flag to tunnel metadata 979 * indicating that checksum computation should be skipped 980 * and checksum set to zeroes. 981 * **BPF_F_DONT_FRAGMENT** 982 * Add a flag to tunnel metadata indicating that the 983 * packet should not be fragmented. 984 * **BPF_F_SEQ_NUMBER** 985 * Add a flag to tunnel metadata indicating that a 986 * sequence number should be added to tunnel header before 987 * sending the packet. This flag was added for GRE 988 * encapsulation, but might be used with other protocols 989 * as well in the future. 990 * 991 * Here is a typical usage on the transmit path: 992 * 993 * :: 994 * 995 * struct bpf_tunnel_key key; 996 * populate key ... 997 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 998 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 999 * 1000 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1001 * helper for additional information. 1002 * Return 1003 * 0 on success, or a negative error in case of failure. 1004 * 1005 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1006 * Description 1007 * Read the value of a perf event counter. This helper relies on a 1008 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1009 * the perf event counter is selected when *map* is updated with 1010 * perf event file descriptors. The *map* is an array whose size 1011 * is the number of available CPUs, and each cell contains a value 1012 * relative to one CPU. The value to retrieve is indicated by 1013 * *flags*, that contains the index of the CPU to look up, masked 1014 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1015 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1016 * current CPU should be retrieved. 1017 * 1018 * Note that before Linux 4.13, only hardware perf event can be 1019 * retrieved. 1020 * 1021 * Also, be aware that the newer helper 1022 * **bpf_perf_event_read_value**\ () is recommended over 1023 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1024 * quirks where error and counter value are used as a return code 1025 * (which is wrong to do since ranges may overlap). This issue is 1026 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1027 * time provides more features over the **bpf_perf_event_read**\ 1028 * () interface. Please refer to the description of 1029 * **bpf_perf_event_read_value**\ () for details. 1030 * Return 1031 * The value of the perf event counter read from the map, or a 1032 * negative error code in case of failure. 1033 * 1034 * int bpf_redirect(u32 ifindex, u64 flags) 1035 * Description 1036 * Redirect the packet to another net device of index *ifindex*. 1037 * This helper is somewhat similar to **bpf_clone_redirect**\ 1038 * (), except that the packet is not cloned, which provides 1039 * increased performance. 1040 * 1041 * Except for XDP, both ingress and egress interfaces can be used 1042 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1043 * to make the distinction (ingress path is selected if the flag 1044 * is present, egress path otherwise). Currently, XDP only 1045 * supports redirection to the egress interface, and accepts no 1046 * flag at all. 1047 * 1048 * The same effect can also be attained with the more generic 1049 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1050 * redirect target instead of providing it directly to the helper. 1051 * Return 1052 * For XDP, the helper returns **XDP_REDIRECT** on success or 1053 * **XDP_ABORTED** on error. For other program types, the values 1054 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1055 * error. 1056 * 1057 * u32 bpf_get_route_realm(struct sk_buff *skb) 1058 * Description 1059 * Retrieve the realm or the route, that is to say the 1060 * **tclassid** field of the destination for the *skb*. The 1061 * indentifier retrieved is a user-provided tag, similar to the 1062 * one used with the net_cls cgroup (see description for 1063 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1064 * held by a route (a destination entry), not by a task. 1065 * 1066 * Retrieving this identifier works with the clsact TC egress hook 1067 * (see also **tc-bpf(8)**), or alternatively on conventional 1068 * classful egress qdiscs, but not on TC ingress path. In case of 1069 * clsact TC egress hook, this has the advantage that, internally, 1070 * the destination entry has not been dropped yet in the transmit 1071 * path. Therefore, the destination entry does not need to be 1072 * artificially held via **netif_keep_dst**\ () for a classful 1073 * qdisc until the *skb* is freed. 1074 * 1075 * This helper is available only if the kernel was compiled with 1076 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1077 * Return 1078 * The realm of the route for the packet associated to *skb*, or 0 1079 * if none was found. 1080 * 1081 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1082 * Description 1083 * Write raw *data* blob into a special BPF perf event held by 1084 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1085 * event must have the following attributes: **PERF_SAMPLE_RAW** 1086 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1087 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1088 * 1089 * The *flags* are used to indicate the index in *map* for which 1090 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1091 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1092 * to indicate that the index of the current CPU core should be 1093 * used. 1094 * 1095 * The value to write, of *size*, is passed through eBPF stack and 1096 * pointed by *data*. 1097 * 1098 * The context of the program *ctx* needs also be passed to the 1099 * helper. 1100 * 1101 * On user space, a program willing to read the values needs to 1102 * call **perf_event_open**\ () on the perf event (either for 1103 * one or for all CPUs) and to store the file descriptor into the 1104 * *map*. This must be done before the eBPF program can send data 1105 * into it. An example is available in file 1106 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1107 * tree (the eBPF program counterpart is in 1108 * *samples/bpf/trace_output_kern.c*). 1109 * 1110 * **bpf_perf_event_output**\ () achieves better performance 1111 * than **bpf_trace_printk**\ () for sharing data with user 1112 * space, and is much better suitable for streaming data from eBPF 1113 * programs. 1114 * 1115 * Note that this helper is not restricted to tracing use cases 1116 * and can be used with programs attached to TC or XDP as well, 1117 * where it allows for passing data to user space listeners. Data 1118 * can be: 1119 * 1120 * * Only custom structs, 1121 * * Only the packet payload, or 1122 * * A combination of both. 1123 * Return 1124 * 0 on success, or a negative error in case of failure. 1125 * 1126 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1127 * Description 1128 * This helper was provided as an easy way to load data from a 1129 * packet. It can be used to load *len* bytes from *offset* from 1130 * the packet associated to *skb*, into the buffer pointed by 1131 * *to*. 1132 * 1133 * Since Linux 4.7, usage of this helper has mostly been replaced 1134 * by "direct packet access", enabling packet data to be 1135 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1136 * pointing respectively to the first byte of packet data and to 1137 * the byte after the last byte of packet data. However, it 1138 * remains useful if one wishes to read large quantities of data 1139 * at once from a packet into the eBPF stack. 1140 * Return 1141 * 0 on success, or a negative error in case of failure. 1142 * 1143 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1144 * Description 1145 * Walk a user or a kernel stack and return its id. To achieve 1146 * this, the helper needs *ctx*, which is a pointer to the context 1147 * on which the tracing program is executed, and a pointer to a 1148 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1149 * 1150 * The last argument, *flags*, holds the number of stack frames to 1151 * skip (from 0 to 255), masked with 1152 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1153 * a combination of the following flags: 1154 * 1155 * **BPF_F_USER_STACK** 1156 * Collect a user space stack instead of a kernel stack. 1157 * **BPF_F_FAST_STACK_CMP** 1158 * Compare stacks by hash only. 1159 * **BPF_F_REUSE_STACKID** 1160 * If two different stacks hash into the same *stackid*, 1161 * discard the old one. 1162 * 1163 * The stack id retrieved is a 32 bit long integer handle which 1164 * can be further combined with other data (including other stack 1165 * ids) and used as a key into maps. This can be useful for 1166 * generating a variety of graphs (such as flame graphs or off-cpu 1167 * graphs). 1168 * 1169 * For walking a stack, this helper is an improvement over 1170 * **bpf_probe_read**\ (), which can be used with unrolled loops 1171 * but is not efficient and consumes a lot of eBPF instructions. 1172 * Instead, **bpf_get_stackid**\ () can collect up to 1173 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1174 * this limit can be controlled with the **sysctl** program, and 1175 * that it should be manually increased in order to profile long 1176 * user stacks (such as stacks for Java programs). To do so, use: 1177 * 1178 * :: 1179 * 1180 * # sysctl kernel.perf_event_max_stack=<new value> 1181 * Return 1182 * The positive or null stack id on success, or a negative error 1183 * in case of failure. 1184 * 1185 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1186 * Description 1187 * Compute a checksum difference, from the raw buffer pointed by 1188 * *from*, of length *from_size* (that must be a multiple of 4), 1189 * towards the raw buffer pointed by *to*, of size *to_size* 1190 * (same remark). An optional *seed* can be added to the value 1191 * (this can be cascaded, the seed may come from a previous call 1192 * to the helper). 1193 * 1194 * This is flexible enough to be used in several ways: 1195 * 1196 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1197 * checksum, it can be used when pushing new data. 1198 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1199 * checksum, it can be used when removing data from a packet. 1200 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1201 * can be used to compute a diff. Note that *from_size* and 1202 * *to_size* do not need to be equal. 1203 * 1204 * This helper can be used in combination with 1205 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1206 * which one can feed in the difference computed with 1207 * **bpf_csum_diff**\ (). 1208 * Return 1209 * The checksum result, or a negative error code in case of 1210 * failure. 1211 * 1212 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1213 * Description 1214 * Retrieve tunnel options metadata for the packet associated to 1215 * *skb*, and store the raw tunnel option data to the buffer *opt* 1216 * of *size*. 1217 * 1218 * This helper can be used with encapsulation devices that can 1219 * operate in "collect metadata" mode (please refer to the related 1220 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1221 * more details). A particular example where this can be used is 1222 * in combination with the Geneve encapsulation protocol, where it 1223 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1224 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1225 * the eBPF program. This allows for full customization of these 1226 * headers. 1227 * Return 1228 * The size of the option data retrieved. 1229 * 1230 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1231 * Description 1232 * Set tunnel options metadata for the packet associated to *skb* 1233 * to the option data contained in the raw buffer *opt* of *size*. 1234 * 1235 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1236 * helper for additional information. 1237 * Return 1238 * 0 on success, or a negative error in case of failure. 1239 * 1240 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1241 * Description 1242 * Change the protocol of the *skb* to *proto*. Currently 1243 * supported are transition from IPv4 to IPv6, and from IPv6 to 1244 * IPv4. The helper takes care of the groundwork for the 1245 * transition, including resizing the socket buffer. The eBPF 1246 * program is expected to fill the new headers, if any, via 1247 * **skb_store_bytes**\ () and to recompute the checksums with 1248 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1249 * (). The main case for this helper is to perform NAT64 1250 * operations out of an eBPF program. 1251 * 1252 * Internally, the GSO type is marked as dodgy so that headers are 1253 * checked and segments are recalculated by the GSO/GRO engine. 1254 * The size for GSO target is adapted as well. 1255 * 1256 * All values for *flags* are reserved for future usage, and must 1257 * be left at zero. 1258 * 1259 * A call to this helper is susceptible to change the underlying 1260 * packet buffer. Therefore, at load time, all checks on pointers 1261 * previously done by the verifier are invalidated and must be 1262 * performed again, if the helper is used in combination with 1263 * direct packet access. 1264 * Return 1265 * 0 on success, or a negative error in case of failure. 1266 * 1267 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1268 * Description 1269 * Change the packet type for the packet associated to *skb*. This 1270 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1271 * the eBPF program does not have a write access to *skb*\ 1272 * **->pkt_type** beside this helper. Using a helper here allows 1273 * for graceful handling of errors. 1274 * 1275 * The major use case is to change incoming *skb*s to 1276 * **PACKET_HOST** in a programmatic way instead of having to 1277 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1278 * example. 1279 * 1280 * Note that *type* only allows certain values. At this time, they 1281 * are: 1282 * 1283 * **PACKET_HOST** 1284 * Packet is for us. 1285 * **PACKET_BROADCAST** 1286 * Send packet to all. 1287 * **PACKET_MULTICAST** 1288 * Send packet to group. 1289 * **PACKET_OTHERHOST** 1290 * Send packet to someone else. 1291 * Return 1292 * 0 on success, or a negative error in case of failure. 1293 * 1294 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1295 * Description 1296 * Check whether *skb* is a descendant of the cgroup2 held by 1297 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1298 * Return 1299 * The return value depends on the result of the test, and can be: 1300 * 1301 * * 0, if the *skb* failed the cgroup2 descendant test. 1302 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1303 * * A negative error code, if an error occurred. 1304 * 1305 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1306 * Description 1307 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1308 * not set, in particular if the hash was cleared due to mangling, 1309 * recompute this hash. Later accesses to the hash can be done 1310 * directly with *skb*\ **->hash**. 1311 * 1312 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1313 * prototype with **bpf_skb_change_proto**\ (), or calling 1314 * **bpf_skb_store_bytes**\ () with the 1315 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1316 * the hash and to trigger a new computation for the next call to 1317 * **bpf_get_hash_recalc**\ (). 1318 * Return 1319 * The 32-bit hash. 1320 * 1321 * u64 bpf_get_current_task(void) 1322 * Return 1323 * A pointer to the current task struct. 1324 * 1325 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1326 * Description 1327 * Attempt in a safe way to write *len* bytes from the buffer 1328 * *src* to *dst* in memory. It only works for threads that are in 1329 * user context, and *dst* must be a valid user space address. 1330 * 1331 * This helper should not be used to implement any kind of 1332 * security mechanism because of TOC-TOU attacks, but rather to 1333 * debug, divert, and manipulate execution of semi-cooperative 1334 * processes. 1335 * 1336 * Keep in mind that this feature is meant for experiments, and it 1337 * has a risk of crashing the system and running programs. 1338 * Therefore, when an eBPF program using this helper is attached, 1339 * a warning including PID and process name is printed to kernel 1340 * logs. 1341 * Return 1342 * 0 on success, or a negative error in case of failure. 1343 * 1344 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1345 * Description 1346 * Check whether the probe is being run is the context of a given 1347 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1348 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1349 * Return 1350 * The return value depends on the result of the test, and can be: 1351 * 1352 * * 0, if the *skb* task belongs to the cgroup2. 1353 * * 1, if the *skb* task does not belong to the cgroup2. 1354 * * A negative error code, if an error occurred. 1355 * 1356 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1357 * Description 1358 * Resize (trim or grow) the packet associated to *skb* to the 1359 * new *len*. The *flags* are reserved for future usage, and must 1360 * be left at zero. 1361 * 1362 * The basic idea is that the helper performs the needed work to 1363 * change the size of the packet, then the eBPF program rewrites 1364 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1365 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1366 * and others. This helper is a slow path utility intended for 1367 * replies with control messages. And because it is targeted for 1368 * slow path, the helper itself can afford to be slow: it 1369 * implicitly linearizes, unclones and drops offloads from the 1370 * *skb*. 1371 * 1372 * A call to this helper is susceptible to change the underlying 1373 * packet buffer. Therefore, at load time, all checks on pointers 1374 * previously done by the verifier are invalidated and must be 1375 * performed again, if the helper is used in combination with 1376 * direct packet access. 1377 * Return 1378 * 0 on success, or a negative error in case of failure. 1379 * 1380 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1381 * Description 1382 * Pull in non-linear data in case the *skb* is non-linear and not 1383 * all of *len* are part of the linear section. Make *len* bytes 1384 * from *skb* readable and writable. If a zero value is passed for 1385 * *len*, then the whole length of the *skb* is pulled. 1386 * 1387 * This helper is only needed for reading and writing with direct 1388 * packet access. 1389 * 1390 * For direct packet access, testing that offsets to access 1391 * are within packet boundaries (test on *skb*\ **->data_end**) is 1392 * susceptible to fail if offsets are invalid, or if the requested 1393 * data is in non-linear parts of the *skb*. On failure the 1394 * program can just bail out, or in the case of a non-linear 1395 * buffer, use a helper to make the data available. The 1396 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1397 * the data. Another one consists in using **bpf_skb_pull_data** 1398 * to pull in once the non-linear parts, then retesting and 1399 * eventually access the data. 1400 * 1401 * At the same time, this also makes sure the *skb* is uncloned, 1402 * which is a necessary condition for direct write. As this needs 1403 * to be an invariant for the write part only, the verifier 1404 * detects writes and adds a prologue that is calling 1405 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1406 * the very beginning in case it is indeed cloned. 1407 * 1408 * A call to this helper is susceptible to change the underlying 1409 * packet buffer. Therefore, at load time, all checks on pointers 1410 * previously done by the verifier are invalidated and must be 1411 * performed again, if the helper is used in combination with 1412 * direct packet access. 1413 * Return 1414 * 0 on success, or a negative error in case of failure. 1415 * 1416 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1417 * Description 1418 * Add the checksum *csum* into *skb*\ **->csum** in case the 1419 * driver has supplied a checksum for the entire packet into that 1420 * field. Return an error otherwise. This helper is intended to be 1421 * used in combination with **bpf_csum_diff**\ (), in particular 1422 * when the checksum needs to be updated after data has been 1423 * written into the packet through direct packet access. 1424 * Return 1425 * The checksum on success, or a negative error code in case of 1426 * failure. 1427 * 1428 * void bpf_set_hash_invalid(struct sk_buff *skb) 1429 * Description 1430 * Invalidate the current *skb*\ **->hash**. It can be used after 1431 * mangling on headers through direct packet access, in order to 1432 * indicate that the hash is outdated and to trigger a 1433 * recalculation the next time the kernel tries to access this 1434 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1435 * 1436 * int bpf_get_numa_node_id(void) 1437 * Description 1438 * Return the id of the current NUMA node. The primary use case 1439 * for this helper is the selection of sockets for the local NUMA 1440 * node, when the program is attached to sockets using the 1441 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1442 * but the helper is also available to other eBPF program types, 1443 * similarly to **bpf_get_smp_processor_id**\ (). 1444 * Return 1445 * The id of current NUMA node. 1446 * 1447 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1448 * Description 1449 * Grows headroom of packet associated to *skb* and adjusts the 1450 * offset of the MAC header accordingly, adding *len* bytes of 1451 * space. It automatically extends and reallocates memory as 1452 * required. 1453 * 1454 * This helper can be used on a layer 3 *skb* to push a MAC header 1455 * for redirection into a layer 2 device. 1456 * 1457 * All values for *flags* are reserved for future usage, and must 1458 * be left at zero. 1459 * 1460 * A call to this helper is susceptible to change the underlying 1461 * packet buffer. Therefore, at load time, all checks on pointers 1462 * previously done by the verifier are invalidated and must be 1463 * performed again, if the helper is used in combination with 1464 * direct packet access. 1465 * Return 1466 * 0 on success, or a negative error in case of failure. 1467 * 1468 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1469 * Description 1470 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1471 * it is possible to use a negative value for *delta*. This helper 1472 * can be used to prepare the packet for pushing or popping 1473 * headers. 1474 * 1475 * A call to this helper is susceptible to change the underlying 1476 * packet buffer. Therefore, at load time, all checks on pointers 1477 * previously done by the verifier are invalidated and must be 1478 * performed again, if the helper is used in combination with 1479 * direct packet access. 1480 * Return 1481 * 0 on success, or a negative error in case of failure. 1482 * 1483 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1484 * Description 1485 * Copy a NUL terminated string from an unsafe kernel address 1486 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for 1487 * more details. 1488 * 1489 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str() 1490 * instead. 1491 * Return 1492 * On success, the strictly positive length of the string, 1493 * including the trailing NUL character. On error, a negative 1494 * value. 1495 * 1496 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1497 * Description 1498 * If the **struct sk_buff** pointed by *skb* has a known socket, 1499 * retrieve the cookie (generated by the kernel) of this socket. 1500 * If no cookie has been set yet, generate a new cookie. Once 1501 * generated, the socket cookie remains stable for the life of the 1502 * socket. This helper can be useful for monitoring per socket 1503 * networking traffic statistics as it provides a global socket 1504 * identifier that can be assumed unique. 1505 * Return 1506 * A 8-byte long non-decreasing number on success, or 0 if the 1507 * socket field is missing inside *skb*. 1508 * 1509 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1510 * Description 1511 * Equivalent to bpf_get_socket_cookie() helper that accepts 1512 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1513 * Return 1514 * A 8-byte long non-decreasing number. 1515 * 1516 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1517 * Description 1518 * Equivalent to bpf_get_socket_cookie() helper that accepts 1519 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1520 * Return 1521 * A 8-byte long non-decreasing number. 1522 * 1523 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1524 * Return 1525 * The owner UID of the socket associated to *skb*. If the socket 1526 * is **NULL**, or if it is not a full socket (i.e. if it is a 1527 * time-wait or a request socket instead), **overflowuid** value 1528 * is returned (note that **overflowuid** might also be the actual 1529 * UID value for the socket). 1530 * 1531 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1532 * Description 1533 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1534 * to value *hash*. 1535 * Return 1536 * 0 1537 * 1538 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1539 * Description 1540 * Emulate a call to **setsockopt()** on the socket associated to 1541 * *bpf_socket*, which must be a full socket. The *level* at 1542 * which the option resides and the name *optname* of the option 1543 * must be specified, see **setsockopt(2)** for more information. 1544 * The option value of length *optlen* is pointed by *optval*. 1545 * 1546 * This helper actually implements a subset of **setsockopt()**. 1547 * It supports the following *level*\ s: 1548 * 1549 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1550 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1551 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1552 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1553 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1554 * **TCP_BPF_SNDCWND_CLAMP**. 1555 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1556 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1557 * Return 1558 * 0 on success, or a negative error in case of failure. 1559 * 1560 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1561 * Description 1562 * Grow or shrink the room for data in the packet associated to 1563 * *skb* by *len_diff*, and according to the selected *mode*. 1564 * 1565 * There are two supported modes at this time: 1566 * 1567 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1568 * (room space is added or removed below the layer 2 header). 1569 * 1570 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1571 * (room space is added or removed below the layer 3 header). 1572 * 1573 * The following flags are supported at this time: 1574 * 1575 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1576 * Adjusting mss in this way is not allowed for datagrams. 1577 * 1578 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1579 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1580 * Any new space is reserved to hold a tunnel header. 1581 * Configure skb offsets and other fields accordingly. 1582 * 1583 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1584 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1585 * Use with ENCAP_L3 flags to further specify the tunnel type. 1586 * 1587 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1588 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1589 * type; *len* is the length of the inner MAC header. 1590 * 1591 * A call to this helper is susceptible to change the underlying 1592 * packet buffer. Therefore, at load time, all checks on pointers 1593 * previously done by the verifier are invalidated and must be 1594 * performed again, if the helper is used in combination with 1595 * direct packet access. 1596 * Return 1597 * 0 on success, or a negative error in case of failure. 1598 * 1599 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1600 * Description 1601 * Redirect the packet to the endpoint referenced by *map* at 1602 * index *key*. Depending on its type, this *map* can contain 1603 * references to net devices (for forwarding packets through other 1604 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1605 * but this is only implemented for native XDP (with driver 1606 * support) as of this writing). 1607 * 1608 * The lower two bits of *flags* are used as the return code if 1609 * the map lookup fails. This is so that the return value can be 1610 * one of the XDP program return codes up to XDP_TX, as chosen by 1611 * the caller. Any higher bits in the *flags* argument must be 1612 * unset. 1613 * 1614 * See also bpf_redirect(), which only supports redirecting to an 1615 * ifindex, but doesn't require a map to do so. 1616 * Return 1617 * **XDP_REDIRECT** on success, or the value of the two lower bits 1618 * of the **flags* argument on error. 1619 * 1620 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1621 * Description 1622 * Redirect the packet to the socket referenced by *map* (of type 1623 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1624 * egress interfaces can be used for redirection. The 1625 * **BPF_F_INGRESS** value in *flags* is used to make the 1626 * distinction (ingress path is selected if the flag is present, 1627 * egress path otherwise). This is the only flag supported for now. 1628 * Return 1629 * **SK_PASS** on success, or **SK_DROP** on error. 1630 * 1631 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1632 * Description 1633 * Add an entry to, or update a *map* referencing sockets. The 1634 * *skops* is used as a new value for the entry associated to 1635 * *key*. *flags* is one of: 1636 * 1637 * **BPF_NOEXIST** 1638 * The entry for *key* must not exist in the map. 1639 * **BPF_EXIST** 1640 * The entry for *key* must already exist in the map. 1641 * **BPF_ANY** 1642 * No condition on the existence of the entry for *key*. 1643 * 1644 * If the *map* has eBPF programs (parser and verdict), those will 1645 * be inherited by the socket being added. If the socket is 1646 * already attached to eBPF programs, this results in an error. 1647 * Return 1648 * 0 on success, or a negative error in case of failure. 1649 * 1650 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1651 * Description 1652 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1653 * *delta* (which can be positive or negative). Note that this 1654 * operation modifies the address stored in *xdp_md*\ **->data**, 1655 * so the latter must be loaded only after the helper has been 1656 * called. 1657 * 1658 * The use of *xdp_md*\ **->data_meta** is optional and programs 1659 * are not required to use it. The rationale is that when the 1660 * packet is processed with XDP (e.g. as DoS filter), it is 1661 * possible to push further meta data along with it before passing 1662 * to the stack, and to give the guarantee that an ingress eBPF 1663 * program attached as a TC classifier on the same device can pick 1664 * this up for further post-processing. Since TC works with socket 1665 * buffers, it remains possible to set from XDP the **mark** or 1666 * **priority** pointers, or other pointers for the socket buffer. 1667 * Having this scratch space generic and programmable allows for 1668 * more flexibility as the user is free to store whatever meta 1669 * data they need. 1670 * 1671 * A call to this helper is susceptible to change the underlying 1672 * packet buffer. Therefore, at load time, all checks on pointers 1673 * previously done by the verifier are invalidated and must be 1674 * performed again, if the helper is used in combination with 1675 * direct packet access. 1676 * Return 1677 * 0 on success, or a negative error in case of failure. 1678 * 1679 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1680 * Description 1681 * Read the value of a perf event counter, and store it into *buf* 1682 * of size *buf_size*. This helper relies on a *map* of type 1683 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1684 * counter is selected when *map* is updated with perf event file 1685 * descriptors. The *map* is an array whose size is the number of 1686 * available CPUs, and each cell contains a value relative to one 1687 * CPU. The value to retrieve is indicated by *flags*, that 1688 * contains the index of the CPU to look up, masked with 1689 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1690 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1691 * current CPU should be retrieved. 1692 * 1693 * This helper behaves in a way close to 1694 * **bpf_perf_event_read**\ () helper, save that instead of 1695 * just returning the value observed, it fills the *buf* 1696 * structure. This allows for additional data to be retrieved: in 1697 * particular, the enabled and running times (in *buf*\ 1698 * **->enabled** and *buf*\ **->running**, respectively) are 1699 * copied. In general, **bpf_perf_event_read_value**\ () is 1700 * recommended over **bpf_perf_event_read**\ (), which has some 1701 * ABI issues and provides fewer functionalities. 1702 * 1703 * These values are interesting, because hardware PMU (Performance 1704 * Monitoring Unit) counters are limited resources. When there are 1705 * more PMU based perf events opened than available counters, 1706 * kernel will multiplex these events so each event gets certain 1707 * percentage (but not all) of the PMU time. In case that 1708 * multiplexing happens, the number of samples or counter value 1709 * will not reflect the case compared to when no multiplexing 1710 * occurs. This makes comparison between different runs difficult. 1711 * Typically, the counter value should be normalized before 1712 * comparing to other experiments. The usual normalization is done 1713 * as follows. 1714 * 1715 * :: 1716 * 1717 * normalized_counter = counter * t_enabled / t_running 1718 * 1719 * Where t_enabled is the time enabled for event and t_running is 1720 * the time running for event since last normalization. The 1721 * enabled and running times are accumulated since the perf event 1722 * open. To achieve scaling factor between two invocations of an 1723 * eBPF program, users can can use CPU id as the key (which is 1724 * typical for perf array usage model) to remember the previous 1725 * value and do the calculation inside the eBPF program. 1726 * Return 1727 * 0 on success, or a negative error in case of failure. 1728 * 1729 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1730 * Description 1731 * For en eBPF program attached to a perf event, retrieve the 1732 * value of the event counter associated to *ctx* and store it in 1733 * the structure pointed by *buf* and of size *buf_size*. Enabled 1734 * and running times are also stored in the structure (see 1735 * description of helper **bpf_perf_event_read_value**\ () for 1736 * more details). 1737 * Return 1738 * 0 on success, or a negative error in case of failure. 1739 * 1740 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1741 * Description 1742 * Emulate a call to **getsockopt()** on the socket associated to 1743 * *bpf_socket*, which must be a full socket. The *level* at 1744 * which the option resides and the name *optname* of the option 1745 * must be specified, see **getsockopt(2)** for more information. 1746 * The retrieved value is stored in the structure pointed by 1747 * *opval* and of length *optlen*. 1748 * 1749 * This helper actually implements a subset of **getsockopt()**. 1750 * It supports the following *level*\ s: 1751 * 1752 * * **IPPROTO_TCP**, which supports *optname* 1753 * **TCP_CONGESTION**. 1754 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1755 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1756 * Return 1757 * 0 on success, or a negative error in case of failure. 1758 * 1759 * int bpf_override_return(struct pt_regs *regs, u64 rc) 1760 * Description 1761 * Used for error injection, this helper uses kprobes to override 1762 * the return value of the probed function, and to set it to *rc*. 1763 * The first argument is the context *regs* on which the kprobe 1764 * works. 1765 * 1766 * This helper works by setting setting the PC (program counter) 1767 * to an override function which is run in place of the original 1768 * probed function. This means the probed function is not run at 1769 * all. The replacement function just returns with the required 1770 * value. 1771 * 1772 * This helper has security implications, and thus is subject to 1773 * restrictions. It is only available if the kernel was compiled 1774 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1775 * option, and in this case it only works on functions tagged with 1776 * **ALLOW_ERROR_INJECTION** in the kernel code. 1777 * 1778 * Also, the helper is only available for the architectures having 1779 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1780 * x86 architecture is the only one to support this feature. 1781 * Return 1782 * 0 1783 * 1784 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1785 * Description 1786 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1787 * for the full TCP socket associated to *bpf_sock_ops* to 1788 * *argval*. 1789 * 1790 * The primary use of this field is to determine if there should 1791 * be calls to eBPF programs of type 1792 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1793 * code. A program of the same type can change its value, per 1794 * connection and as necessary, when the connection is 1795 * established. This field is directly accessible for reading, but 1796 * this helper must be used for updates in order to return an 1797 * error if an eBPF program tries to set a callback that is not 1798 * supported in the current kernel. 1799 * 1800 * *argval* is a flag array which can combine these flags: 1801 * 1802 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1803 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1804 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1805 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1806 * 1807 * Therefore, this function can be used to clear a callback flag by 1808 * setting the appropriate bit to zero. e.g. to disable the RTO 1809 * callback: 1810 * 1811 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1812 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1813 * 1814 * Here are some examples of where one could call such eBPF 1815 * program: 1816 * 1817 * * When RTO fires. 1818 * * When a packet is retransmitted. 1819 * * When the connection terminates. 1820 * * When a packet is sent. 1821 * * When a packet is received. 1822 * Return 1823 * Code **-EINVAL** if the socket is not a full TCP socket; 1824 * otherwise, a positive number containing the bits that could not 1825 * be set is returned (which comes down to 0 if all bits were set 1826 * as required). 1827 * 1828 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1829 * Description 1830 * This helper is used in programs implementing policies at the 1831 * socket level. If the message *msg* is allowed to pass (i.e. if 1832 * the verdict eBPF program returns **SK_PASS**), redirect it to 1833 * the socket referenced by *map* (of type 1834 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1835 * egress interfaces can be used for redirection. The 1836 * **BPF_F_INGRESS** value in *flags* is used to make the 1837 * distinction (ingress path is selected if the flag is present, 1838 * egress path otherwise). This is the only flag supported for now. 1839 * Return 1840 * **SK_PASS** on success, or **SK_DROP** on error. 1841 * 1842 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1843 * Description 1844 * For socket policies, apply the verdict of the eBPF program to 1845 * the next *bytes* (number of bytes) of message *msg*. 1846 * 1847 * For example, this helper can be used in the following cases: 1848 * 1849 * * A single **sendmsg**\ () or **sendfile**\ () system call 1850 * contains multiple logical messages that the eBPF program is 1851 * supposed to read and for which it should apply a verdict. 1852 * * An eBPF program only cares to read the first *bytes* of a 1853 * *msg*. If the message has a large payload, then setting up 1854 * and calling the eBPF program repeatedly for all bytes, even 1855 * though the verdict is already known, would create unnecessary 1856 * overhead. 1857 * 1858 * When called from within an eBPF program, the helper sets a 1859 * counter internal to the BPF infrastructure, that is used to 1860 * apply the last verdict to the next *bytes*. If *bytes* is 1861 * smaller than the current data being processed from a 1862 * **sendmsg**\ () or **sendfile**\ () system call, the first 1863 * *bytes* will be sent and the eBPF program will be re-run with 1864 * the pointer for start of data pointing to byte number *bytes* 1865 * **+ 1**. If *bytes* is larger than the current data being 1866 * processed, then the eBPF verdict will be applied to multiple 1867 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1868 * consumed. 1869 * 1870 * Note that if a socket closes with the internal counter holding 1871 * a non-zero value, this is not a problem because data is not 1872 * being buffered for *bytes* and is sent as it is received. 1873 * Return 1874 * 0 1875 * 1876 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1877 * Description 1878 * For socket policies, prevent the execution of the verdict eBPF 1879 * program for message *msg* until *bytes* (byte number) have been 1880 * accumulated. 1881 * 1882 * This can be used when one needs a specific number of bytes 1883 * before a verdict can be assigned, even if the data spans 1884 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1885 * case would be a user calling **sendmsg**\ () repeatedly with 1886 * 1-byte long message segments. Obviously, this is bad for 1887 * performance, but it is still valid. If the eBPF program needs 1888 * *bytes* bytes to validate a header, this helper can be used to 1889 * prevent the eBPF program to be called again until *bytes* have 1890 * been accumulated. 1891 * Return 1892 * 0 1893 * 1894 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1895 * Description 1896 * For socket policies, pull in non-linear data from user space 1897 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1898 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1899 * respectively. 1900 * 1901 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1902 * *msg* it can only parse data that the (**data**, **data_end**) 1903 * pointers have already consumed. For **sendmsg**\ () hooks this 1904 * is likely the first scatterlist element. But for calls relying 1905 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1906 * be the range (**0**, **0**) because the data is shared with 1907 * user space and by default the objective is to avoid allowing 1908 * user space to modify data while (or after) eBPF verdict is 1909 * being decided. This helper can be used to pull in data and to 1910 * set the start and end pointer to given values. Data will be 1911 * copied if necessary (i.e. if data was not linear and if start 1912 * and end pointers do not point to the same chunk). 1913 * 1914 * A call to this helper is susceptible to change the underlying 1915 * packet buffer. Therefore, at load time, all checks on pointers 1916 * previously done by the verifier are invalidated and must be 1917 * performed again, if the helper is used in combination with 1918 * direct packet access. 1919 * 1920 * All values for *flags* are reserved for future usage, and must 1921 * be left at zero. 1922 * Return 1923 * 0 on success, or a negative error in case of failure. 1924 * 1925 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1926 * Description 1927 * Bind the socket associated to *ctx* to the address pointed by 1928 * *addr*, of length *addr_len*. This allows for making outgoing 1929 * connection from the desired IP address, which can be useful for 1930 * example when all processes inside a cgroup should use one 1931 * single IP address on a host that has multiple IP configured. 1932 * 1933 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1934 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1935 * **AF_INET6**). Looking for a free port to bind to can be 1936 * expensive, therefore binding to port is not permitted by the 1937 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1938 * must be set to zero. 1939 * Return 1940 * 0 on success, or a negative error in case of failure. 1941 * 1942 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1943 * Description 1944 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1945 * only possible to shrink the packet as of this writing, 1946 * therefore *delta* must be a negative integer. 1947 * 1948 * A call to this helper is susceptible to change the underlying 1949 * packet buffer. Therefore, at load time, all checks on pointers 1950 * previously done by the verifier are invalidated and must be 1951 * performed again, if the helper is used in combination with 1952 * direct packet access. 1953 * Return 1954 * 0 on success, or a negative error in case of failure. 1955 * 1956 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1957 * Description 1958 * Retrieve the XFRM state (IP transform framework, see also 1959 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1960 * 1961 * The retrieved value is stored in the **struct bpf_xfrm_state** 1962 * pointed by *xfrm_state* and of length *size*. 1963 * 1964 * All values for *flags* are reserved for future usage, and must 1965 * be left at zero. 1966 * 1967 * This helper is available only if the kernel was compiled with 1968 * **CONFIG_XFRM** configuration option. 1969 * Return 1970 * 0 on success, or a negative error in case of failure. 1971 * 1972 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 1973 * Description 1974 * Return a user or a kernel stack in bpf program provided buffer. 1975 * To achieve this, the helper needs *ctx*, which is a pointer 1976 * to the context on which the tracing program is executed. 1977 * To store the stacktrace, the bpf program provides *buf* with 1978 * a nonnegative *size*. 1979 * 1980 * The last argument, *flags*, holds the number of stack frames to 1981 * skip (from 0 to 255), masked with 1982 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1983 * the following flags: 1984 * 1985 * **BPF_F_USER_STACK** 1986 * Collect a user space stack instead of a kernel stack. 1987 * **BPF_F_USER_BUILD_ID** 1988 * Collect buildid+offset instead of ips for user stack, 1989 * only valid if **BPF_F_USER_STACK** is also specified. 1990 * 1991 * **bpf_get_stack**\ () can collect up to 1992 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1993 * to sufficient large buffer size. Note that 1994 * this limit can be controlled with the **sysctl** program, and 1995 * that it should be manually increased in order to profile long 1996 * user stacks (such as stacks for Java programs). To do so, use: 1997 * 1998 * :: 1999 * 2000 * # sysctl kernel.perf_event_max_stack=<new value>
2001 * Return 2002 * A non-negative value equal to or less than *size* on success, 2003 * or a negative error in case of failure. 2004 * 2005 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2006 * Description 2007 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2008 * it provides an easy way to load *len* bytes from *offset* 2009 * from the packet associated to *skb*, into the buffer pointed 2010 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2011 * a fifth argument *start_header* exists in order to select a 2012 * base offset to start from. *start_header* can be one of: 2013 * 2014 * **BPF_HDR_START_MAC** 2015 * Base offset to load data from is *skb*'s mac header. 2016 * **BPF_HDR_START_NET** 2017 * Base offset to load data from is *skb*'s network header. 2018 * 2019 * In general, "direct packet access" is the preferred method to 2020 * access packet data, however, this helper is in particular useful 2021 * in socket filters where *skb*\ **->data** does not always point 2022 * to the start of the mac header and where "direct packet access" 2023 * is not available. 2024 * Return 2025 * 0 on success, or a negative error in case of failure. 2026 * 2027 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2028 * Description 2029 * Do FIB lookup in kernel tables using parameters in *params*. 2030 * If lookup is successful and result shows packet is to be 2031 * forwarded, the neighbor tables are searched for the nexthop. 2032 * If successful (ie., FIB lookup shows forwarding and nexthop 2033 * is resolved), the nexthop address is returned in ipv4_dst 2034 * or ipv6_dst based on family, smac is set to mac address of 2035 * egress device, dmac is set to nexthop mac address, rt_metric 2036 * is set to metric from route (IPv4/IPv6 only), and ifindex 2037 * is set to the device index of the nexthop from the FIB lookup. 2038 * 2039 * *plen* argument is the size of the passed in struct. 2040 * *flags* argument can be a combination of one or more of the 2041 * following values: 2042 * 2043 * **BPF_FIB_LOOKUP_DIRECT** 2044 * Do a direct table lookup vs full lookup using FIB 2045 * rules. 2046 * **BPF_FIB_LOOKUP_OUTPUT** 2047 * Perform lookup from an egress perspective (default is 2048 * ingress). 2049 * 2050 * *ctx* is either **struct xdp_md** for XDP programs or 2051 * **struct sk_buff** tc cls_act programs. 2052 * Return 2053 * * < 0 if any input argument is invalid 2054 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2055 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2056 * packet is not forwarded or needs assist from full stack 2057 * 2058 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2059 * Description 2060 * Add an entry to, or update a sockhash *map* referencing sockets. 2061 * The *skops* is used as a new value for the entry associated to 2062 * *key*. *flags* is one of: 2063 * 2064 * **BPF_NOEXIST** 2065 * The entry for *key* must not exist in the map. 2066 * **BPF_EXIST** 2067 * The entry for *key* must already exist in the map. 2068 * **BPF_ANY** 2069 * No condition on the existence of the entry for *key*. 2070 * 2071 * If the *map* has eBPF programs (parser and verdict), those will 2072 * be inherited by the socket being added. If the socket is 2073 * already attached to eBPF programs, this results in an error. 2074 * Return 2075 * 0 on success, or a negative error in case of failure. 2076 * 2077 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2078 * Description 2079 * This helper is used in programs implementing policies at the 2080 * socket level. If the message *msg* is allowed to pass (i.e. if 2081 * the verdict eBPF program returns **SK_PASS**), redirect it to 2082 * the socket referenced by *map* (of type 2083 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2084 * egress interfaces can be used for redirection. The 2085 * **BPF_F_INGRESS** value in *flags* is used to make the 2086 * distinction (ingress path is selected if the flag is present, 2087 * egress path otherwise). This is the only flag supported for now. 2088 * Return 2089 * **SK_PASS** on success, or **SK_DROP** on error. 2090 * 2091 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2092 * Description 2093 * This helper is used in programs implementing policies at the 2094 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2095 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2096 * to the socket referenced by *map* (of type 2097 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2098 * egress interfaces can be used for redirection. The 2099 * **BPF_F_INGRESS** value in *flags* is used to make the 2100 * distinction (ingress path is selected if the flag is present, 2101 * egress otherwise). This is the only flag supported for now. 2102 * Return 2103 * **SK_PASS** on success, or **SK_DROP** on error. 2104 * 2105 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2106 * Description 2107 * Encapsulate the packet associated to *skb* within a Layer 3 2108 * protocol header. This header is provided in the buffer at 2109 * address *hdr*, with *len* its size in bytes. *type* indicates 2110 * the protocol of the header and can be one of: 2111 * 2112 * **BPF_LWT_ENCAP_SEG6** 2113 * IPv6 encapsulation with Segment Routing Header 2114 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2115 * the IPv6 header is computed by the kernel. 2116 * **BPF_LWT_ENCAP_SEG6_INLINE** 2117 * Only works if *skb* contains an IPv6 packet. Insert a 2118 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2119 * the IPv6 header. 2120 * **BPF_LWT_ENCAP_IP** 2121 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2122 * must be IPv4 or IPv6, followed by zero or more 2123 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2124 * total bytes in all prepended headers. Please note that 2125 * if **skb_is_gso**\ (*skb*) is true, no more than two 2126 * headers can be prepended, and the inner header, if 2127 * present, should be either GRE or UDP/GUE. 2128 * 2129 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2130 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2131 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2132 * **BPF_PROG_TYPE_LWT_XMIT**. 2133 * 2134 * A call to this helper is susceptible to change the underlying 2135 * packet buffer. Therefore, at load time, all checks on pointers 2136 * previously done by the verifier are invalidated and must be 2137 * performed again, if the helper is used in combination with 2138 * direct packet access. 2139 * Return 2140 * 0 on success, or a negative error in case of failure. 2141 * 2142 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2143 * Description 2144 * Store *len* bytes from address *from* into the packet 2145 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2146 * inside the outermost IPv6 Segment Routing Header can be 2147 * modified through this helper. 2148 * 2149 * A call to this helper is susceptible to change the underlying 2150 * packet buffer. Therefore, at load time, all checks on pointers 2151 * previously done by the verifier are invalidated and must be 2152 * performed again, if the helper is used in combination with 2153 * direct packet access. 2154 * Return 2155 * 0 on success, or a negative error in case of failure. 2156 * 2157 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2158 * Description 2159 * Adjust the size allocated to TLVs in the outermost IPv6 2160 * Segment Routing Header contained in the packet associated to 2161 * *skb*, at position *offset* by *delta* bytes. Only offsets 2162 * after the segments are accepted. *delta* can be as well 2163 * positive (growing) as negative (shrinking). 2164 * 2165 * A call to this helper is susceptible to change the underlying 2166 * packet buffer. Therefore, at load time, all checks on pointers 2167 * previously done by the verifier are invalidated and must be 2168 * performed again, if the helper is used in combination with 2169 * direct packet access. 2170 * Return 2171 * 0 on success, or a negative error in case of failure. 2172 * 2173 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2174 * Description 2175 * Apply an IPv6 Segment Routing action of type *action* to the 2176 * packet associated to *skb*. Each action takes a parameter 2177 * contained at address *param*, and of length *param_len* bytes. 2178 * *action* can be one of: 2179 * 2180 * **SEG6_LOCAL_ACTION_END_X** 2181 * End.X action: Endpoint with Layer-3 cross-connect. 2182 * Type of *param*: **struct in6_addr**. 2183 * **SEG6_LOCAL_ACTION_END_T** 2184 * End.T action: Endpoint with specific IPv6 table lookup. 2185 * Type of *param*: **int**. 2186 * **SEG6_LOCAL_ACTION_END_B6** 2187 * End.B6 action: Endpoint bound to an SRv6 policy. 2188 * Type of *param*: **struct ipv6_sr_hdr**. 2189 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2190 * End.B6.Encap action: Endpoint bound to an SRv6 2191 * encapsulation policy. 2192 * Type of *param*: **struct ipv6_sr_hdr**. 2193 * 2194 * A call to this helper is susceptible to change the underlying 2195 * packet buffer. Therefore, at load time, all checks on pointers 2196 * previously done by the verifier are invalidated and must be 2197 * performed again, if the helper is used in combination with 2198 * direct packet access. 2199 * Return 2200 * 0 on success, or a negative error in case of failure. 2201 * 2202 * int bpf_rc_repeat(void *ctx) 2203 * Description 2204 * This helper is used in programs implementing IR decoding, to 2205 * report a successfully decoded repeat key message. This delays 2206 * the generation of a key up event for previously generated 2207 * key down event. 2208 * 2209 * Some IR protocols like NEC have a special IR message for 2210 * repeating last button, for when a button is held down. 2211 * 2212 * The *ctx* should point to the lirc sample as passed into 2213 * the program. 2214 * 2215 * This helper is only available is the kernel was compiled with 2216 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2217 * "**y**". 2218 * Return 2219 * 0 2220 * 2221 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2222 * Description 2223 * This helper is used in programs implementing IR decoding, to 2224 * report a successfully decoded key press with *scancode*, 2225 * *toggle* value in the given *protocol*. The scancode will be 2226 * translated to a keycode using the rc keymap, and reported as 2227 * an input key down event. After a period a key up event is 2228 * generated. This period can be extended by calling either 2229 * **bpf_rc_keydown**\ () again with the same values, or calling 2230 * **bpf_rc_repeat**\ (). 2231 * 2232 * Some protocols include a toggle bit, in case the button was 2233 * released and pressed again between consecutive scancodes. 2234 * 2235 * The *ctx* should point to the lirc sample as passed into 2236 * the program. 2237 * 2238 * The *protocol* is the decoded protocol number (see 2239 * **enum rc_proto** for some predefined values). 2240 * 2241 * This helper is only available is the kernel was compiled with 2242 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2243 * "**y**". 2244 * Return 2245 * 0 2246 * 2247 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2248 * Description 2249 * Return the cgroup v2 id of the socket associated with the *skb*. 2250 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2251 * helper for cgroup v1 by providing a tag resp. identifier that 2252 * can be matched on or used for map lookups e.g. to implement 2253 * policy. The cgroup v2 id of a given path in the hierarchy is 2254 * exposed in user space through the f_handle API in order to get 2255 * to the same 64-bit id. 2256 * 2257 * This helper can be used on TC egress path, but not on ingress, 2258 * and is available only if the kernel was compiled with the 2259 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2260 * Return 2261 * The id is returned or 0 in case the id could not be retrieved. 2262 * 2263 * u64 bpf_get_current_cgroup_id(void) 2264 * Return 2265 * A 64-bit integer containing the current cgroup id based 2266 * on the cgroup within which the current task is running. 2267 * 2268 * void *bpf_get_local_storage(void *map, u64 flags) 2269 * Description 2270 * Get the pointer to the local storage area. 2271 * The type and the size of the local storage is defined 2272 * by the *map* argument. 2273 * The *flags* meaning is specific for each map type, 2274 * and has to be 0 for cgroup local storage. 2275 * 2276 * Depending on the BPF program type, a local storage area 2277 * can be shared between multiple instances of the BPF program, 2278 * running simultaneously. 2279 * 2280 * A user should care about the synchronization by himself. 2281 * For example, by using the **BPF_STX_XADD** instruction to alter 2282 * the shared data. 2283 * Return 2284 * A pointer to the local storage area. 2285 * 2286 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2287 * Description 2288 * Select a **SO_REUSEPORT** socket from a 2289 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2290 * It checks the selected socket is matching the incoming 2291 * request in the socket buffer. 2292 * Return 2293 * 0 on success, or a negative error in case of failure. 2294 * 2295 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2296 * Description 2297 * Return id of cgroup v2 that is ancestor of cgroup associated 2298 * with the *skb* at the *ancestor_level*. The root cgroup is at 2299 * *ancestor_level* zero and each step down the hierarchy 2300 * increments the level. If *ancestor_level* == level of cgroup 2301 * associated with *skb*, then return value will be same as that 2302 * of **bpf_skb_cgroup_id**\ (). 2303 * 2304 * The helper is useful to implement policies based on cgroups 2305 * that are upper in hierarchy than immediate cgroup associated 2306 * with *skb*. 2307 * 2308 * The format of returned id and helper limitations are same as in 2309 * **bpf_skb_cgroup_id**\ (). 2310 * Return 2311 * The id is returned or 0 in case the id could not be retrieved. 2312 * 2313 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2314 * Description 2315 * Look for TCP socket matching *tuple*, optionally in a child 2316 * network namespace *netns*. The return value must be checked, 2317 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2318 * 2319 * The *ctx* should point to the context of the program, such as 2320 * the skb or socket (depending on the hook in use). This is used 2321 * to determine the base network namespace for the lookup. 2322 * 2323 * *tuple_size* must be one of: 2324 * 2325 * **sizeof**\ (*tuple*\ **->ipv4**) 2326 * Look for an IPv4 socket. 2327 * **sizeof**\ (*tuple*\ **->ipv6**) 2328 * Look for an IPv6 socket. 2329 * 2330 * If the *netns* is a negative signed 32-bit integer, then the 2331 * socket lookup table in the netns associated with the *ctx* will 2332 * will be used. For the TC hooks, this is the netns of the device 2333 * in the skb. For socket hooks, this is the netns of the socket. 2334 * If *netns* is any other signed 32-bit value greater than or 2335 * equal to zero then it specifies the ID of the netns relative to 2336 * the netns associated with the *ctx*. *netns* values beyond the 2337 * range of 32-bit integers are reserved for future use. 2338 * 2339 * All values for *flags* are reserved for future usage, and must 2340 * be left at zero. 2341 * 2342 * This helper is available only if the kernel was compiled with 2343 * **CONFIG_NET** configuration option. 2344 * Return 2345 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2346 * For sockets with reuseport option, the **struct bpf_sock** 2347 * result is from *reuse*\ **->socks**\ [] using the hash of the 2348 * tuple. 2349 * 2350 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2351 * Description 2352 * Look for UDP socket matching *tuple*, optionally in a child 2353 * network namespace *netns*. The return value must be checked, 2354 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2355 * 2356 * The *ctx* should point to the context of the program, such as 2357 * the skb or socket (depending on the hook in use). This is used 2358 * to determine the base network namespace for the lookup. 2359 * 2360 * *tuple_size* must be one of: 2361 * 2362 * **sizeof**\ (*tuple*\ **->ipv4**) 2363 * Look for an IPv4 socket. 2364 * **sizeof**\ (*tuple*\ **->ipv6**) 2365 * Look for an IPv6 socket. 2366 * 2367 * If the *netns* is a negative signed 32-bit integer, then the 2368 * socket lookup table in the netns associated with the *ctx* will 2369 * will be used. For the TC hooks, this is the netns of the device 2370 * in the skb. For socket hooks, this is the netns of the socket. 2371 * If *netns* is any other signed 32-bit value greater than or 2372 * equal to zero then it specifies the ID of the netns relative to 2373 * the netns associated with the *ctx*. *netns* values beyond the 2374 * range of 32-bit integers are reserved for future use. 2375 * 2376 * All values for *flags* are reserved for future usage, and must 2377 * be left at zero. 2378 * 2379 * This helper is available only if the kernel was compiled with 2380 * **CONFIG_NET** configuration option. 2381 * Return 2382 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2383 * For sockets with reuseport option, the **struct bpf_sock** 2384 * result is from *reuse*\ **->socks**\ [] using the hash of the 2385 * tuple. 2386 * 2387 * int bpf_sk_release(struct bpf_sock *sock) 2388 * Description 2389 * Release the reference held by *sock*. *sock* must be a 2390 * non-**NULL** pointer that was returned from 2391 * **bpf_sk_lookup_xxx**\ (). 2392 * Return 2393 * 0 on success, or a negative error in case of failure. 2394 * 2395 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2396 * Description 2397 * Push an element *value* in *map*. *flags* is one of: 2398 * 2399 * **BPF_EXIST** 2400 * If the queue/stack is full, the oldest element is 2401 * removed to make room for this. 2402 * Return 2403 * 0 on success, or a negative error in case of failure. 2404 * 2405 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2406 * Description 2407 * Pop an element from *map*. 2408 * Return 2409 * 0 on success, or a negative error in case of failure. 2410 * 2411 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2412 * Description 2413 * Get an element from *map* without removing it. 2414 * Return 2415 * 0 on success, or a negative error in case of failure. 2416 * 2417 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2418 * Description 2419 * For socket policies, insert *len* bytes into *msg* at offset 2420 * *start*. 2421 * 2422 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2423 * *msg* it may want to insert metadata or options into the *msg*. 2424 * This can later be read and used by any of the lower layer BPF 2425 * hooks. 2426 * 2427 * This helper may fail if under memory pressure (a malloc 2428 * fails) in these cases BPF programs will get an appropriate 2429 * error and BPF programs will need to handle them. 2430 * Return 2431 * 0 on success, or a negative error in case of failure. 2432 * 2433 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2434 * Description 2435 * Will remove *len* bytes from a *msg* starting at byte *start*. 2436 * This may result in **ENOMEM** errors under certain situations if 2437 * an allocation and copy are required due to a full ring buffer. 2438 * However, the helper will try to avoid doing the allocation 2439 * if possible. Other errors can occur if input parameters are 2440 * invalid either due to *start* byte not being valid part of *msg* 2441 * payload and/or *pop* value being to large. 2442 * Return 2443 * 0 on success, or a negative error in case of failure. 2444 * 2445 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2446 * Description 2447 * This helper is used in programs implementing IR decoding, to 2448 * report a successfully decoded pointer movement. 2449 * 2450 * The *ctx* should point to the lirc sample as passed into 2451 * the program. 2452 * 2453 * This helper is only available is the kernel was compiled with 2454 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2455 * "**y**". 2456 * Return 2457 * 0 2458 * 2459 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2460 * Description 2461 * Acquire a spinlock represented by the pointer *lock*, which is 2462 * stored as part of a value of a map. Taking the lock allows to 2463 * safely update the rest of the fields in that value. The 2464 * spinlock can (and must) later be released with a call to 2465 * **bpf_spin_unlock**\ (\ *lock*\ ). 2466 * 2467 * Spinlocks in BPF programs come with a number of restrictions 2468 * and constraints: 2469 * 2470 * * **bpf_spin_lock** objects are only allowed inside maps of 2471 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2472 * list could be extended in the future). 2473 * * BTF description of the map is mandatory. 2474 * * The BPF program can take ONE lock at a time, since taking two 2475 * or more could cause dead locks. 2476 * * Only one **struct bpf_spin_lock** is allowed per map element. 2477 * * When the lock is taken, calls (either BPF to BPF or helpers) 2478 * are not allowed. 2479 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2480 * allowed inside a spinlock-ed region. 2481 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2482 * the lock, on all execution paths, before it returns. 2483 * * The BPF program can access **struct bpf_spin_lock** only via 2484 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2485 * helpers. Loading or storing data into the **struct 2486 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2487 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2488 * of the map value must be a struct and have **struct 2489 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2490 * Nested lock inside another struct is not allowed. 2491 * * The **struct bpf_spin_lock** *lock* field in a map value must 2492 * be aligned on a multiple of 4 bytes in that value. 2493 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2494 * the **bpf_spin_lock** field to user space. 2495 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2496 * a BPF program, do not update the **bpf_spin_lock** field. 2497 * * **bpf_spin_lock** cannot be on the stack or inside a 2498 * networking packet (it can only be inside of a map values). 2499 * * **bpf_spin_lock** is available to root only. 2500 * * Tracing programs and socket filter programs cannot use 2501 * **bpf_spin_lock**\ () due to insufficient preemption checks 2502 * (but this may change in the future). 2503 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2504 * Return 2505 * 0 2506 * 2507 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2508 * Description 2509 * Release the *lock* previously locked by a call to 2510 * **bpf_spin_lock**\ (\ *lock*\ ). 2511 * Return 2512 * 0 2513 * 2514 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2515 * Description 2516 * This helper gets a **struct bpf_sock** pointer such 2517 * that all the fields in this **bpf_sock** can be accessed. 2518 * Return 2519 * A **struct bpf_sock** pointer on success, or **NULL** in 2520 * case of failure. 2521 * 2522 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2523 * Description 2524 * This helper gets a **struct bpf_tcp_sock** pointer from a 2525 * **struct bpf_sock** pointer. 2526 * Return 2527 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2528 * case of failure. 2529 * 2530 * int bpf_skb_ecn_set_ce(struct sk_buff *skb) 2531 * Description 2532 * Set ECN (Explicit Congestion Notification) field of IP header 2533 * to **CE** (Congestion Encountered) if current value is **ECT** 2534 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2535 * and IPv4. 2536 * Return 2537 * 1 if the **CE** flag is set (either by the current helper call 2538 * or because it was already present), 0 if it is not set. 2539 * 2540 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2541 * Description 2542 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2543 * **bpf_sk_release**\ () is unnecessary and not allowed. 2544 * Return 2545 * A **struct bpf_sock** pointer on success, or **NULL** in 2546 * case of failure. 2547 * 2548 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2549 * Description 2550 * Look for TCP socket matching *tuple*, optionally in a child 2551 * network namespace *netns*. The return value must be checked, 2552 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2553 * 2554 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2555 * that it also returns timewait or request sockets. Use 2556 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2557 * full structure. 2558 * 2559 * This helper is available only if the kernel was compiled with 2560 * **CONFIG_NET** configuration option. 2561 * Return 2562 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2563 * For sockets with reuseport option, the **struct bpf_sock** 2564 * result is from *reuse*\ **->socks**\ [] using the hash of the 2565 * tuple. 2566 * 2567 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2568 * Description 2569 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2570 * the listening socket in *sk*. 2571 * 2572 * *iph* points to the start of the IPv4 or IPv6 header, while 2573 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2574 * **sizeof**\ (**struct ip6hdr**). 2575 * 2576 * *th* points to the start of the TCP header, while *th_len* 2577 * contains **sizeof**\ (**struct tcphdr**). 2578 * 2579 * Return 2580 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2581 * error otherwise. 2582 * 2583 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2584 * Description 2585 * Get name of sysctl in /proc/sys/ and copy it into provided by 2586 * program buffer *buf* of size *buf_len*. 2587 * 2588 * The buffer is always NUL terminated, unless it's zero-sized. 2589 * 2590 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2591 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2592 * only (e.g. "tcp_mem"). 2593 * Return 2594 * Number of character copied (not including the trailing NUL). 2595 * 2596 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2597 * truncated name in this case). 2598 * 2599 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2600 * Description 2601 * Get current value of sysctl as it is presented in /proc/sys 2602 * (incl. newline, etc), and copy it as a string into provided 2603 * by program buffer *buf* of size *buf_len*. 2604 * 2605 * The whole value is copied, no matter what file position user 2606 * space issued e.g. sys_read at. 2607 * 2608 * The buffer is always NUL terminated, unless it's zero-sized. 2609 * Return 2610 * Number of character copied (not including the trailing NUL). 2611 * 2612 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2613 * truncated name in this case). 2614 * 2615 * **-EINVAL** if current value was unavailable, e.g. because 2616 * sysctl is uninitialized and read returns -EIO for it. 2617 * 2618 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2619 * Description 2620 * Get new value being written by user space to sysctl (before 2621 * the actual write happens) and copy it as a string into 2622 * provided by program buffer *buf* of size *buf_len*. 2623 * 2624 * User space may write new value at file position > 0. 2625 * 2626 * The buffer is always NUL terminated, unless it's zero-sized. 2627 * Return 2628 * Number of character copied (not including the trailing NUL). 2629 * 2630 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2631 * truncated name in this case). 2632 * 2633 * **-EINVAL** if sysctl is being read. 2634 * 2635 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2636 * Description 2637 * Override new value being written by user space to sysctl with 2638 * value provided by program in buffer *buf* of size *buf_len*. 2639 * 2640 * *buf* should contain a string in same form as provided by user 2641 * space on sysctl write. 2642 * 2643 * User space may write new value at file position > 0. To override 2644 * the whole sysctl value file position should be set to zero. 2645 * Return 2646 * 0 on success. 2647 * 2648 * **-E2BIG** if the *buf_len* is too big. 2649 * 2650 * **-EINVAL** if sysctl is being read. 2651 * 2652 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2653 * Description 2654 * Convert the initial part of the string from buffer *buf* of 2655 * size *buf_len* to a long integer according to the given base 2656 * and save the result in *res*. 2657 * 2658 * The string may begin with an arbitrary amount of white space 2659 * (as determined by **isspace**\ (3)) followed by a single 2660 * optional '**-**' sign. 2661 * 2662 * Five least significant bits of *flags* encode base, other bits 2663 * are currently unused. 2664 * 2665 * Base must be either 8, 10, 16 or 0 to detect it automatically 2666 * similar to user space **strtol**\ (3). 2667 * Return 2668 * Number of characters consumed on success. Must be positive but 2669 * no more than *buf_len*. 2670 * 2671 * **-EINVAL** if no valid digits were found or unsupported base 2672 * was provided. 2673 * 2674 * **-ERANGE** if resulting value was out of range. 2675 * 2676 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2677 * Description 2678 * Convert the initial part of the string from buffer *buf* of 2679 * size *buf_len* to an unsigned long integer according to the 2680 * given base and save the result in *res*. 2681 * 2682 * The string may begin with an arbitrary amount of white space 2683 * (as determined by **isspace**\ (3)). 2684 * 2685 * Five least significant bits of *flags* encode base, other bits 2686 * are currently unused. 2687 * 2688 * Base must be either 8, 10, 16 or 0 to detect it automatically 2689 * similar to user space **strtoul**\ (3). 2690 * Return 2691 * Number of characters consumed on success. Must be positive but 2692 * no more than *buf_len*. 2693 * 2694 * **-EINVAL** if no valid digits were found or unsupported base 2695 * was provided. 2696 * 2697 * **-ERANGE** if resulting value was out of range. 2698 * 2699 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2700 * Description 2701 * Get a bpf-local-storage from a *sk*. 2702 * 2703 * Logically, it could be thought of getting the value from 2704 * a *map* with *sk* as the **key**. From this 2705 * perspective, the usage is not much different from 2706 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2707 * helper enforces the key must be a full socket and the map must 2708 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2709 * 2710 * Underneath, the value is stored locally at *sk* instead of 2711 * the *map*. The *map* is used as the bpf-local-storage 2712 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2713 * searched against all bpf-local-storages residing at *sk*. 2714 * 2715 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2716 * used such that a new bpf-local-storage will be 2717 * created if one does not exist. *value* can be used 2718 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2719 * the initial value of a bpf-local-storage. If *value* is 2720 * **NULL**, the new bpf-local-storage will be zero initialized. 2721 * Return 2722 * A bpf-local-storage pointer is returned on success. 2723 * 2724 * **NULL** if not found or there was an error in adding 2725 * a new bpf-local-storage. 2726 * 2727 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2728 * Description 2729 * Delete a bpf-local-storage from a *sk*. 2730 * Return 2731 * 0 on success. 2732 * 2733 * **-ENOENT** if the bpf-local-storage cannot be found. 2734 * 2735 * int bpf_send_signal(u32 sig) 2736 * Description 2737 * Send signal *sig* to the process of the current task. 2738 * The signal may be delivered to any of this process's threads. 2739 * Return 2740 * 0 on success or successfully queued. 2741 * 2742 * **-EBUSY** if work queue under nmi is full. 2743 * 2744 * **-EINVAL** if *sig* is invalid. 2745 * 2746 * **-EPERM** if no permission to send the *sig*. 2747 * 2748 * **-EAGAIN** if bpf program can try again. 2749 * 2750 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2751 * Description 2752 * Try to issue a SYN cookie for the packet with corresponding 2753 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2754 * 2755 * *iph* points to the start of the IPv4 or IPv6 header, while 2756 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2757 * **sizeof**\ (**struct ip6hdr**). 2758 * 2759 * *th* points to the start of the TCP header, while *th_len* 2760 * contains the length of the TCP header. 2761 * 2762 * Return 2763 * On success, lower 32 bits hold the generated SYN cookie in 2764 * followed by 16 bits which hold the MSS value for that cookie, 2765 * and the top 16 bits are unused. 2766 * 2767 * On failure, the returned value is one of the following: 2768 * 2769 * **-EINVAL** SYN cookie cannot be issued due to error 2770 * 2771 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2772 * 2773 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2774 * 2775 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2776 * 2777 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2778 * Description 2779 * Write raw *data* blob into a special BPF perf event held by 2780 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2781 * event must have the following attributes: **PERF_SAMPLE_RAW** 2782 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2783 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2784 * 2785 * The *flags* are used to indicate the index in *map* for which 2786 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2787 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2788 * to indicate that the index of the current CPU core should be 2789 * used. 2790 * 2791 * The value to write, of *size*, is passed through eBPF stack and 2792 * pointed by *data*. 2793 * 2794 * *ctx* is a pointer to in-kernel struct sk_buff. 2795 * 2796 * This helper is similar to **bpf_perf_event_output**\ () but 2797 * restricted to raw_tracepoint bpf programs. 2798 * Return 2799 * 0 on success, or a negative error in case of failure. 2800 * 2801 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2802 * Description 2803 * Safely attempt to read *size* bytes from user space address 2804 * *unsafe_ptr* and store the data in *dst*. 2805 * Return 2806 * 0 on success, or a negative error in case of failure. 2807 * 2808 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2809 * Description 2810 * Safely attempt to read *size* bytes from kernel space address 2811 * *unsafe_ptr* and store the data in *dst*. 2812 * Return 2813 * 0 on success, or a negative error in case of failure. 2814 * 2815 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2816 * Description 2817 * Copy a NUL terminated string from an unsafe user address 2818 * *unsafe_ptr* to *dst*. The *size* should include the 2819 * terminating NUL byte. In case the string length is smaller than 2820 * *size*, the target is not padded with further NUL bytes. If the 2821 * string length is larger than *size*, just *size*-1 bytes are 2822 * copied and the last byte is set to NUL. 2823 * 2824 * On success, the length of the copied string is returned. This 2825 * makes this helper useful in tracing programs for reading 2826 * strings, and more importantly to get its length at runtime. See 2827 * the following snippet: 2828 * 2829 * :: 2830 * 2831 * SEC("kprobe/sys_open") 2832 * void bpf_sys_open(struct pt_regs *ctx) 2833 * { 2834 * char buf[PATHLEN]; // PATHLEN is defined to 256 2835 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2836 * ctx->di); 2837 * 2838 * // Consume buf, for example push it to 2839 * // userspace via bpf_perf_event_output(); we 2840 * // can use res (the string length) as event 2841 * // size, after checking its boundaries. 2842 * } 2843 * 2844 * In comparison, using **bpf_probe_read_user()** helper here 2845 * instead to read the string would require to estimate the length 2846 * at compile time, and would often result in copying more memory 2847 * than necessary. 2848 * 2849 * Another useful use case is when parsing individual process 2850 * arguments or individual environment variables navigating 2851 * *current*\ **->mm->arg_start** and *current*\ 2852 * **->mm->env_start**: using this helper and the return value, 2853 * one can quickly iterate at the right offset of the memory area. 2854 * Return 2855 * On success, the strictly positive length of the string, 2856 * including the trailing NUL character. On error, a negative 2857 * value. 2858 * 2859 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 2860 * Description 2861 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2862 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply. 2863 * Return 2864 * On success, the strictly positive length of the string, including 2865 * the trailing NUL character. On error, a negative value. 2866 * 2867 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 2868 * Description 2869 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock. 2870 * *rcv_nxt* is the ack_seq to be sent out. 2871 * Return 2872 * 0 on success, or a negative error in case of failure. 2873 * 2874 * int bpf_send_signal_thread(u32 sig) 2875 * Description 2876 * Send signal *sig* to the thread corresponding to the current task. 2877 * Return 2878 * 0 on success or successfully queued. 2879 * 2880 * **-EBUSY** if work queue under nmi is full. 2881 * 2882 * **-EINVAL** if *sig* is invalid. 2883 * 2884 * **-EPERM** if no permission to send the *sig*. 2885 * 2886 * **-EAGAIN** if bpf program can try again. 2887 * 2888 * u64 bpf_jiffies64(void) 2889 * Description 2890 * Obtain the 64bit jiffies 2891 * Return 2892 * The 64 bit jiffies 2893 */ 2894#define __BPF_FUNC_MAPPER(FN) \ 2895 FN(unspec), \ 2896 FN(map_lookup_elem), \ 2897 FN(map_update_elem), \ 2898 FN(map_delete_elem), \ 2899 FN(probe_read), \ 2900 FN(ktime_get_ns), \ 2901 FN(trace_printk), \ 2902 FN(get_prandom_u32), \ 2903 FN(get_smp_processor_id), \ 2904 FN(skb_store_bytes), \ 2905 FN(l3_csum_replace), \ 2906 FN(l4_csum_replace), \ 2907 FN(tail_call), \ 2908 FN(clone_redirect), \ 2909 FN(get_current_pid_tgid), \ 2910 FN(get_current_uid_gid), \ 2911 FN(get_current_comm), \ 2912 FN(get_cgroup_classid), \ 2913 FN(skb_vlan_push), \ 2914 FN(skb_vlan_pop), \ 2915 FN(skb_get_tunnel_key), \ 2916 FN(skb_set_tunnel_key), \ 2917 FN(perf_event_read), \ 2918 FN(redirect), \ 2919 FN(get_route_realm), \ 2920 FN(perf_event_output), \ 2921 FN(skb_load_bytes), \ 2922 FN(get_stackid), \ 2923 FN(csum_diff), \ 2924 FN(skb_get_tunnel_opt), \ 2925 FN(skb_set_tunnel_opt), \ 2926 FN(skb_change_proto), \ 2927 FN(skb_change_type), \ 2928 FN(skb_under_cgroup), \ 2929 FN(get_hash_recalc), \ 2930 FN(get_current_task), \ 2931 FN(probe_write_user), \ 2932 FN(current_task_under_cgroup), \ 2933 FN(skb_change_tail), \ 2934 FN(skb_pull_data), \ 2935 FN(csum_update), \ 2936 FN(set_hash_invalid), \ 2937 FN(get_numa_node_id), \ 2938 FN(skb_change_head), \ 2939 FN(xdp_adjust_head), \ 2940 FN(probe_read_str), \ 2941 FN(get_socket_cookie), \ 2942 FN(get_socket_uid), \ 2943 FN(set_hash), \ 2944 FN(setsockopt), \ 2945 FN(skb_adjust_room), \ 2946 FN(redirect_map), \ 2947 FN(sk_redirect_map), \ 2948 FN(sock_map_update), \ 2949 FN(xdp_adjust_meta), \ 2950 FN(perf_event_read_value), \ 2951 FN(perf_prog_read_value), \ 2952 FN(getsockopt), \ 2953 FN(override_return), \ 2954 FN(sock_ops_cb_flags_set), \ 2955 FN(msg_redirect_map), \ 2956 FN(msg_apply_bytes), \ 2957 FN(msg_cork_bytes), \ 2958 FN(msg_pull_data), \ 2959 FN(bind), \ 2960 FN(xdp_adjust_tail), \ 2961 FN(skb_get_xfrm_state), \ 2962 FN(get_stack), \ 2963 FN(skb_load_bytes_relative), \ 2964 FN(fib_lookup), \ 2965 FN(sock_hash_update), \ 2966 FN(msg_redirect_hash), \ 2967 FN(sk_redirect_hash), \ 2968 FN(lwt_push_encap), \ 2969 FN(lwt_seg6_store_bytes), \ 2970 FN(lwt_seg6_adjust_srh), \ 2971 FN(lwt_seg6_action), \ 2972 FN(rc_repeat), \ 2973 FN(rc_keydown), \ 2974 FN(skb_cgroup_id), \ 2975 FN(get_current_cgroup_id), \ 2976 FN(get_local_storage), \ 2977 FN(sk_select_reuseport), \ 2978 FN(skb_ancestor_cgroup_id), \ 2979 FN(sk_lookup_tcp), \ 2980 FN(sk_lookup_udp), \ 2981 FN(sk_release), \ 2982 FN(map_push_elem), \ 2983 FN(map_pop_elem), \ 2984 FN(map_peek_elem), \ 2985 FN(msg_push_data), \ 2986 FN(msg_pop_data), \ 2987 FN(rc_pointer_rel), \ 2988 FN(spin_lock), \ 2989 FN(spin_unlock), \ 2990 FN(sk_fullsock), \ 2991 FN(tcp_sock), \ 2992 FN(skb_ecn_set_ce), \ 2993 FN(get_listener_sock), \ 2994 FN(skc_lookup_tcp), \ 2995 FN(tcp_check_syncookie), \ 2996 FN(sysctl_get_name), \ 2997 FN(sysctl_get_current_value), \ 2998 FN(sysctl_get_new_value), \ 2999 FN(sysctl_set_new_value), \ 3000 FN(strtol), \
3001 FN(strtoul), \ 3002 FN(sk_storage_get), \ 3003 FN(sk_storage_delete), \ 3004 FN(send_signal), \ 3005 FN(tcp_gen_syncookie), \ 3006 FN(skb_output), \ 3007 FN(probe_read_user), \ 3008 FN(probe_read_kernel), \ 3009 FN(probe_read_user_str), \ 3010 FN(probe_read_kernel_str), \ 3011 FN(tcp_send_ack), \ 3012 FN(send_signal_thread), \ 3013 FN(jiffies64), 3014 3015/* integer value in 'imm' field of BPF_CALL instruction selects which helper 3016 * function eBPF program intends to call 3017 */ 3018#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3019enum bpf_func_id { 3020 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3021 __BPF_FUNC_MAX_ID, 3022}; 3023#undef __BPF_ENUM_FN 3024 3025/* All flags used by eBPF helper functions, placed here. */ 3026 3027/* BPF_FUNC_skb_store_bytes flags. */ 3028#define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 3029#define BPF_F_INVALIDATE_HASH (1ULL << 1) 3030 3031/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3032 * First 4 bits are for passing the header field size. 3033 */ 3034#define BPF_F_HDR_FIELD_MASK 0xfULL 3035 3036/* BPF_FUNC_l4_csum_replace flags. */ 3037#define BPF_F_PSEUDO_HDR (1ULL << 4) 3038#define BPF_F_MARK_MANGLED_0 (1ULL << 5) 3039#define BPF_F_MARK_ENFORCE (1ULL << 6) 3040 3041/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3042#define BPF_F_INGRESS (1ULL << 0) 3043 3044/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3045#define BPF_F_TUNINFO_IPV6 (1ULL << 0) 3046 3047/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3048#define BPF_F_SKIP_FIELD_MASK 0xffULL 3049#define BPF_F_USER_STACK (1ULL << 8) 3050/* flags used by BPF_FUNC_get_stackid only. */ 3051#define BPF_F_FAST_STACK_CMP (1ULL << 9) 3052#define BPF_F_REUSE_STACKID (1ULL << 10) 3053/* flags used by BPF_FUNC_get_stack only. */ 3054#define BPF_F_USER_BUILD_ID (1ULL << 11) 3055 3056/* BPF_FUNC_skb_set_tunnel_key flags. */ 3057#define BPF_F_ZERO_CSUM_TX (1ULL << 1) 3058#define BPF_F_DONT_FRAGMENT (1ULL << 2) 3059#define BPF_F_SEQ_NUMBER (1ULL << 3) 3060 3061/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3062 * BPF_FUNC_perf_event_read_value flags. 3063 */ 3064#define BPF_F_INDEX_MASK 0xffffffffULL 3065#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 3066/* BPF_FUNC_perf_event_output for sk_buff input context. */ 3067#define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 3068 3069/* Current network namespace */ 3070#define BPF_F_CURRENT_NETNS (-1L) 3071 3072/* BPF_FUNC_skb_adjust_room flags. */ 3073#define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 3074 3075#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 3076#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 3077 3078#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 3079#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 3080#define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 3081#define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 3082#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3083 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3084 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3085 3086/* BPF_FUNC_sysctl_get_name flags. */ 3087#define BPF_F_SYSCTL_BASE_NAME (1ULL << 0) 3088 3089/* BPF_FUNC_sk_storage_get flags */ 3090#define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0) 3091 3092/* Mode for BPF_FUNC_skb_adjust_room helper. */ 3093enum bpf_adj_room_mode { 3094 BPF_ADJ_ROOM_NET, 3095 BPF_ADJ_ROOM_MAC, 3096}; 3097 3098/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3099enum bpf_hdr_start_off { 3100 BPF_HDR_START_MAC, 3101 BPF_HDR_START_NET, 3102}; 3103 3104/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3105enum bpf_lwt_encap_mode { 3106 BPF_LWT_ENCAP_SEG6, 3107 BPF_LWT_ENCAP_SEG6_INLINE, 3108 BPF_LWT_ENCAP_IP, 3109}; 3110 3111#define __bpf_md_ptr(type, name) \ 3112union { \ 3113 type name; \ 3114 __u64 :64; \ 3115} __attribute__((aligned(8))) 3116 3117/* user accessible mirror of in-kernel sk_buff. 3118 * new fields can only be added to the end of this structure 3119 */ 3120struct __sk_buff { 3121 __u32 len; 3122 __u32 pkt_type; 3123 __u32 mark; 3124 __u32 queue_mapping; 3125 __u32 protocol; 3126 __u32 vlan_present; 3127 __u32 vlan_tci; 3128 __u32 vlan_proto; 3129 __u32 priority; 3130 __u32 ingress_ifindex; 3131 __u32 ifindex; 3132 __u32 tc_index; 3133 __u32 cb[5]; 3134 __u32 hash; 3135 __u32 tc_classid; 3136 __u32 data; 3137 __u32 data_end; 3138 __u32 napi_id; 3139 3140 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3141 __u32 family; 3142 __u32 remote_ip4; /* Stored in network byte order */ 3143 __u32 local_ip4; /* Stored in network byte order */ 3144 __u32 remote_ip6[4]; /* Stored in network byte order */ 3145 __u32 local_ip6[4]; /* Stored in network byte order */ 3146 __u32 remote_port; /* Stored in network byte order */ 3147 __u32 local_port; /* stored in host byte order */ 3148 /* ... here. */ 3149 3150 __u32 data_meta; 3151 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3152 __u64 tstamp; 3153 __u32 wire_len; 3154 __u32 gso_segs; 3155 __bpf_md_ptr(struct bpf_sock *, sk); 3156}; 3157 3158struct bpf_tunnel_key { 3159 __u32 tunnel_id; 3160 union { 3161 __u32 remote_ipv4; 3162 __u32 remote_ipv6[4]; 3163 }; 3164 __u8 tunnel_tos; 3165 __u8 tunnel_ttl; 3166 __u16 tunnel_ext; /* Padding, future use. */ 3167 __u32 tunnel_label; 3168}; 3169 3170/* user accessible mirror of in-kernel xfrm_state. 3171 * new fields can only be added to the end of this structure 3172 */ 3173struct bpf_xfrm_state { 3174 __u32 reqid; 3175 __u32 spi; /* Stored in network byte order */ 3176 __u16 family; 3177 __u16 ext; /* Padding, future use. */ 3178 union { 3179 __u32 remote_ipv4; /* Stored in network byte order */ 3180 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3181 }; 3182}; 3183 3184/* Generic BPF return codes which all BPF program types may support. 3185 * The values are binary compatible with their TC_ACT_* counter-part to 3186 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3187 * programs. 3188 * 3189 * XDP is handled seprately, see XDP_*. 3190 */ 3191enum bpf_ret_code { 3192 BPF_OK = 0, 3193 /* 1 reserved */ 3194 BPF_DROP = 2, 3195 /* 3-6 reserved */ 3196 BPF_REDIRECT = 7, 3197 /* >127 are reserved for prog type specific return codes. 3198 * 3199 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3200 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3201 * changed and should be routed based on its new L3 header. 3202 * (This is an L3 redirect, as opposed to L2 redirect 3203 * represented by BPF_REDIRECT above). 3204 */ 3205 BPF_LWT_REROUTE = 128, 3206}; 3207 3208struct bpf_sock { 3209 __u32 bound_dev_if; 3210 __u32 family; 3211 __u32 type; 3212 __u32 protocol; 3213 __u32 mark; 3214 __u32 priority; 3215 /* IP address also allows 1 and 2 bytes access */ 3216 __u32 src_ip4; 3217 __u32 src_ip6[4]; 3218 __u32 src_port; /* host byte order */ 3219 __u32 dst_port; /* network byte order */ 3220 __u32 dst_ip4; 3221 __u32 dst_ip6[4]; 3222 __u32 state; 3223}; 3224 3225struct bpf_tcp_sock { 3226 __u32 snd_cwnd; /* Sending congestion window */ 3227 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3228 __u32 rtt_min; 3229 __u32 snd_ssthresh; /* Slow start size threshold */ 3230 __u32 rcv_nxt; /* What we want to receive next */ 3231 __u32 snd_nxt; /* Next sequence we send */ 3232 __u32 snd_una; /* First byte we want an ack for */ 3233 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3234 __u32 ecn_flags; /* ECN status bits. */ 3235 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3236 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3237 __u32 packets_out; /* Packets which are "in flight" */ 3238 __u32 retrans_out; /* Retransmitted packets out */ 3239 __u32 total_retrans; /* Total retransmits for entire connection */ 3240 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3241 * total number of segments in. 3242 */ 3243 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3244 * total number of data segments in. 3245 */ 3246 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3247 * The total number of segments sent. 3248 */ 3249 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3250 * total number of data segments sent. 3251 */ 3252 __u32 lost_out; /* Lost packets */ 3253 __u32 sacked_out; /* SACK'd packets */ 3254 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3255 * sum(delta(rcv_nxt)), or how many bytes 3256 * were acked. 3257 */ 3258 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3259 * sum(delta(snd_una)), or how many bytes 3260 * were acked. 3261 */ 3262 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3263 * total number of DSACK blocks received 3264 */ 3265 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3266 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3267 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3268}; 3269 3270struct bpf_sock_tuple { 3271 union { 3272 struct { 3273 __be32 saddr; 3274 __be32 daddr; 3275 __be16 sport; 3276 __be16 dport; 3277 } ipv4; 3278 struct { 3279 __be32 saddr[4]; 3280 __be32 daddr[4]; 3281 __be16 sport; 3282 __be16 dport; 3283 } ipv6; 3284 }; 3285}; 3286 3287struct bpf_xdp_sock { 3288 __u32 queue_id; 3289}; 3290 3291#define XDP_PACKET_HEADROOM 256 3292 3293/* User return codes for XDP prog type. 3294 * A valid XDP program must return one of these defined values. All other 3295 * return codes are reserved for future use. Unknown return codes will 3296 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3297 */ 3298enum xdp_action { 3299 XDP_ABORTED = 0, 3300 XDP_DROP, 3301 XDP_PASS, 3302 XDP_TX, 3303 XDP_REDIRECT, 3304}; 3305 3306/* user accessible metadata for XDP packet hook 3307 * new fields must be added to the end of this structure 3308 */ 3309struct xdp_md { 3310 __u32 data; 3311 __u32 data_end; 3312 __u32 data_meta; 3313 /* Below access go through struct xdp_rxq_info */ 3314 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3315 __u32 rx_queue_index; /* rxq->queue_index */ 3316}; 3317 3318enum sk_action { 3319 SK_DROP = 0, 3320 SK_PASS, 3321}; 3322 3323/* user accessible metadata for SK_MSG packet hook, new fields must 3324 * be added to the end of this structure 3325 */ 3326struct sk_msg_md { 3327 __bpf_md_ptr(void *, data); 3328 __bpf_md_ptr(void *, data_end); 3329 3330 __u32 family; 3331 __u32 remote_ip4; /* Stored in network byte order */ 3332 __u32 local_ip4; /* Stored in network byte order */ 3333 __u32 remote_ip6[4]; /* Stored in network byte order */ 3334 __u32 local_ip6[4]; /* Stored in network byte order */ 3335 __u32 remote_port; /* Stored in network byte order */ 3336 __u32 local_port; /* stored in host byte order */ 3337 __u32 size; /* Total size of sk_msg */ 3338}; 3339 3340struct sk_reuseport_md { 3341 /* 3342 * Start of directly accessible data. It begins from 3343 * the tcp/udp header. 3344 */ 3345 __bpf_md_ptr(void *, data); 3346 /* End of directly accessible data */ 3347 __bpf_md_ptr(void *, data_end); 3348 /* 3349 * Total length of packet (starting from the tcp/udp header). 3350 * Note that the directly accessible bytes (data_end - data) 3351 * could be less than this "len". Those bytes could be 3352 * indirectly read by a helper "bpf_skb_load_bytes()". 3353 */ 3354 __u32 len; 3355 /* 3356 * Eth protocol in the mac header (network byte order). e.g. 3357 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3358 */ 3359 __u32 eth_protocol; 3360 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3361 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3362 __u32 hash; /* A hash of the packet 4 tuples */ 3363}; 3364 3365#define BPF_TAG_SIZE 8 3366 3367struct bpf_prog_info { 3368 __u32 type; 3369 __u32 id; 3370 __u8 tag[BPF_TAG_SIZE]; 3371 __u32 jited_prog_len; 3372 __u32 xlated_prog_len; 3373 __aligned_u64 jited_prog_insns; 3374 __aligned_u64 xlated_prog_insns; 3375 __u64 load_time; /* ns since boottime */ 3376 __u32 created_by_uid; 3377 __u32 nr_map_ids; 3378 __aligned_u64 map_ids; 3379 char name[BPF_OBJ_NAME_LEN]; 3380 __u32 ifindex; 3381 __u32 gpl_compatible:1; 3382 __u32 :31; /* alignment pad */ 3383 __u64 netns_dev; 3384 __u64 netns_ino; 3385 __u32 nr_jited_ksyms; 3386 __u32 nr_jited_func_lens; 3387 __aligned_u64 jited_ksyms; 3388 __aligned_u64 jited_func_lens; 3389 __u32 btf_id; 3390 __u32 func_info_rec_size; 3391 __aligned_u64 func_info; 3392 __u32 nr_func_info; 3393 __u32 nr_line_info; 3394 __aligned_u64 line_info; 3395 __aligned_u64 jited_line_info; 3396 __u32 nr_jited_line_info; 3397 __u32 line_info_rec_size; 3398 __u32 jited_line_info_rec_size; 3399 __u32 nr_prog_tags; 3400 __aligned_u64 prog_tags; 3401 __u64 run_time_ns; 3402 __u64 run_cnt; 3403} __attribute__((aligned(8))); 3404 3405struct bpf_map_info { 3406 __u32 type; 3407 __u32 id; 3408 __u32 key_size; 3409 __u32 value_size; 3410 __u32 max_entries; 3411 __u32 map_flags; 3412 char name[BPF_OBJ_NAME_LEN]; 3413 __u32 ifindex; 3414 __u32 btf_vmlinux_value_type_id; 3415 __u64 netns_dev; 3416 __u64 netns_ino; 3417 __u32 btf_id; 3418 __u32 btf_key_type_id; 3419 __u32 btf_value_type_id; 3420} __attribute__((aligned(8))); 3421 3422struct bpf_btf_info { 3423 __aligned_u64 btf; 3424 __u32 btf_size; 3425 __u32 id; 3426} __attribute__((aligned(8))); 3427 3428/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3429 * by user and intended to be used by socket (e.g. to bind to, depends on 3430 * attach attach type). 3431 */ 3432struct bpf_sock_addr { 3433 __u32 user_family; /* Allows 4-byte read, but no write. */ 3434 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3435 * Stored in network byte order. 3436 */ 3437 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3438 * Stored in network byte order. 3439 */ 3440 __u32 user_port; /* Allows 4-byte read and write. 3441 * Stored in network byte order 3442 */ 3443 __u32 family; /* Allows 4-byte read, but no write */ 3444 __u32 type; /* Allows 4-byte read, but no write */ 3445 __u32 protocol; /* Allows 4-byte read, but no write */ 3446 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3447 * Stored in network byte order. 3448 */ 3449 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3450 * Stored in network byte order. 3451 */ 3452 __bpf_md_ptr(struct bpf_sock *, sk); 3453}; 3454 3455/* User bpf_sock_ops struct to access socket values and specify request ops 3456 * and their replies. 3457 * Some of this fields are in network (bigendian) byte order and may need 3458 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3459 * New fields can only be added at the end of this structure 3460 */ 3461struct bpf_sock_ops { 3462 __u32 op; 3463 union { 3464 __u32 args[4]; /* Optionally passed to bpf program */ 3465 __u32 reply; /* Returned by bpf program */ 3466 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3467 }; 3468 __u32 family; 3469 __u32 remote_ip4; /* Stored in network byte order */ 3470 __u32 local_ip4; /* Stored in network byte order */ 3471 __u32 remote_ip6[4]; /* Stored in network byte order */ 3472 __u32 local_ip6[4]; /* Stored in network byte order */ 3473 __u32 remote_port; /* Stored in network byte order */ 3474 __u32 local_port; /* stored in host byte order */ 3475 __u32 is_fullsock; /* Some TCP fields are only valid if 3476 * there is a full socket. If not, the 3477 * fields read as zero. 3478 */ 3479 __u32 snd_cwnd; 3480 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3481 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3482 __u32 state; 3483 __u32 rtt_min; 3484 __u32 snd_ssthresh; 3485 __u32 rcv_nxt; 3486 __u32 snd_nxt; 3487 __u32 snd_una; 3488 __u32 mss_cache; 3489 __u32 ecn_flags; 3490 __u32 rate_delivered; 3491 __u32 rate_interval_us; 3492 __u32 packets_out; 3493 __u32 retrans_out; 3494 __u32 total_retrans; 3495 __u32 segs_in; 3496 __u32 data_segs_in; 3497 __u32 segs_out; 3498 __u32 data_segs_out; 3499 __u32 lost_out; 3500 __u32 sacked_out; 3501 __u32 sk_txhash; 3502 __u64 bytes_received; 3503 __u64 bytes_acked; 3504 __bpf_md_ptr(struct bpf_sock *, sk); 3505}; 3506 3507/* Definitions for bpf_sock_ops_cb_flags */ 3508#define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3509#define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3510#define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3511#define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3) 3512#define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently 3513 * supported cb flags 3514 */ 3515 3516/* List of known BPF sock_ops operators. 3517 * New entries can only be added at the end 3518 */ 3519enum { 3520 BPF_SOCK_OPS_VOID, 3521 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3522 * -1 if default value should be used 3523 */ 3524 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3525 * window (in packets) or -1 if default 3526 * value should be used 3527 */ 3528 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3529 * active connection is initialized 3530 */ 3531 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3532 * active connection is 3533 * established 3534 */ 3535 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3536 * passive connection is 3537 * established 3538 */ 3539 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3540 * needs ECN 3541 */ 3542 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3543 * based on the path and may be 3544 * dependent on the congestion control 3545 * algorithm. In general it indicates 3546 * a congestion threshold. RTTs above 3547 * this indicate congestion 3548 */ 3549 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3550 * Arg1: value of icsk_retransmits 3551 * Arg2: value of icsk_rto 3552 * Arg3: whether RTO has expired 3553 */ 3554 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3555 * Arg1: sequence number of 1st byte 3556 * Arg2: # segments 3557 * Arg3: return value of 3558 * tcp_transmit_skb (0 => success) 3559 */ 3560 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3561 * Arg1: old_state 3562 * Arg2: new_state 3563 */ 3564 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3565 * socket transition to LISTEN state. 3566 */ 3567 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 3568 */ 3569}; 3570 3571/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3572 * changes between the TCP and BPF versions. Ideally this should never happen. 3573 * If it does, we need to add code to convert them before calling 3574 * the BPF sock_ops function. 3575 */ 3576enum { 3577 BPF_TCP_ESTABLISHED = 1, 3578 BPF_TCP_SYN_SENT, 3579 BPF_TCP_SYN_RECV, 3580 BPF_TCP_FIN_WAIT1, 3581 BPF_TCP_FIN_WAIT2, 3582 BPF_TCP_TIME_WAIT, 3583 BPF_TCP_CLOSE, 3584 BPF_TCP_CLOSE_WAIT, 3585 BPF_TCP_LAST_ACK, 3586 BPF_TCP_LISTEN, 3587 BPF_TCP_CLOSING, /* Now a valid state */ 3588 BPF_TCP_NEW_SYN_RECV, 3589 3590 BPF_TCP_MAX_STATES /* Leave at the end! */ 3591}; 3592 3593#define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3594#define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3595 3596struct bpf_perf_event_value { 3597 __u64 counter; 3598 __u64 enabled; 3599 __u64 running; 3600}; 3601 3602#define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3603#define BPF_DEVCG_ACC_READ (1ULL << 1) 3604#define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3605 3606#define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3607#define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3608 3609struct bpf_cgroup_dev_ctx { 3610 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3611 __u32 access_type; 3612 __u32 major; 3613 __u32 minor; 3614}; 3615 3616struct bpf_raw_tracepoint_args { 3617 __u64 args[0]; 3618}; 3619 3620/* DIRECT: Skip the FIB rules and go to FIB table associated with device 3621 * OUTPUT: Do lookup from egress perspective; default is ingress 3622 */ 3623#define BPF_FIB_LOOKUP_DIRECT (1U << 0) 3624#define BPF_FIB_LOOKUP_OUTPUT (1U << 1) 3625 3626enum { 3627 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3628 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3629 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3630 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3631 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3632 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3633 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3634 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3635 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3636}; 3637 3638struct bpf_fib_lookup { 3639 /* input: network family for lookup (AF_INET, AF_INET6) 3640 * output: network family of egress nexthop 3641 */ 3642 __u8 family; 3643 3644 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3645 __u8 l4_protocol; 3646 __be16 sport; 3647 __be16 dport; 3648 3649 /* total length of packet from network header - used for MTU check */ 3650 __u16 tot_len; 3651 3652 /* input: L3 device index for lookup 3653 * output: device index from FIB lookup 3654 */ 3655 __u32 ifindex; 3656 3657 union { 3658 /* inputs to lookup */ 3659 __u8 tos; /* AF_INET */ 3660 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3661 3662 /* output: metric of fib result (IPv4/IPv6 only) */ 3663 __u32 rt_metric; 3664 }; 3665 3666 union { 3667 __be32 ipv4_src; 3668 __u32 ipv6_src[4]; /* in6_addr; network order */ 3669 }; 3670 3671 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3672 * network header. output: bpf_fib_lookup sets to gateway address 3673 * if FIB lookup returns gateway route 3674 */ 3675 union { 3676 __be32 ipv4_dst; 3677 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3678 }; 3679 3680 /* output */ 3681 __be16 h_vlan_proto; 3682 __be16 h_vlan_TCI; 3683 __u8 smac[6]; /* ETH_ALEN */ 3684 __u8 dmac[6]; /* ETH_ALEN */ 3685}; 3686 3687enum bpf_task_fd_type { 3688 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3689 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3690 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3691 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3692 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3693 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3694}; 3695 3696#define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0) 3697#define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1) 3698#define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2) 3699 3700struct bpf_flow_keys { 3701 __u16 nhoff; 3702 __u16 thoff; 3703 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3704 __u8 is_frag; 3705 __u8 is_first_frag; 3706 __u8 is_encap; 3707 __u8 ip_proto; 3708 __be16 n_proto; 3709 __be16 sport; 3710 __be16 dport; 3711 union { 3712 struct { 3713 __be32 ipv4_src; 3714 __be32 ipv4_dst; 3715 }; 3716 struct { 3717 __u32 ipv6_src[4]; /* in6_addr; network order */ 3718 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3719 }; 3720 }; 3721 __u32 flags; 3722 __be32 flow_label; 3723}; 3724 3725struct bpf_func_info { 3726 __u32 insn_off; 3727 __u32 type_id; 3728}; 3729 3730#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3731#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3732 3733struct bpf_line_info { 3734 __u32 insn_off; 3735 __u32 file_name_off; 3736 __u32 line_off; 3737 __u32 line_col; 3738}; 3739 3740struct bpf_spin_lock { 3741 __u32 val; 3742}; 3743 3744struct bpf_sysctl { 3745 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3746 * Allows 1,2,4-byte read, but no write. 3747 */ 3748 __u32 file_pos; /* Sysctl file position to read from, write to. 3749 * Allows 1,2,4-byte read an 4-byte write. 3750 */ 3751}; 3752 3753struct bpf_sockopt { 3754 __bpf_md_ptr(struct bpf_sock *, sk); 3755 __bpf_md_ptr(void *, optval); 3756 __bpf_md_ptr(void *, optval_end); 3757 3758 __s32 level; 3759 __s32 optname; 3760 __s32 optlen; 3761 __s32 retval; 3762}; 3763 3764#endif /* _UAPI__LINUX_BPF_H__ */ 3765