1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 /* BPF syscall commands, see bpf(2) man-page for details. */ 85 enum bpf_cmd { 86 BPF_MAP_CREATE, 87 BPF_MAP_LOOKUP_ELEM, 88 BPF_MAP_UPDATE_ELEM, 89 BPF_MAP_DELETE_ELEM, 90 BPF_MAP_GET_NEXT_KEY, 91 BPF_PROG_LOAD, 92 BPF_OBJ_PIN, 93 BPF_OBJ_GET, 94 BPF_PROG_ATTACH, 95 BPF_PROG_DETACH, 96 BPF_PROG_TEST_RUN, 97 BPF_PROG_GET_NEXT_ID, 98 BPF_MAP_GET_NEXT_ID, 99 BPF_PROG_GET_FD_BY_ID, 100 BPF_MAP_GET_FD_BY_ID, 101 BPF_OBJ_GET_INFO_BY_FD, 102 BPF_PROG_QUERY, 103 BPF_RAW_TRACEPOINT_OPEN, 104 BPF_BTF_LOAD, 105 BPF_BTF_GET_FD_BY_ID, 106 BPF_TASK_FD_QUERY, 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 108 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 BPF_LINK_CREATE, 115 BPF_LINK_UPDATE, 116 }; 117 118 enum bpf_map_type { 119 BPF_MAP_TYPE_UNSPEC, 120 BPF_MAP_TYPE_HASH, 121 BPF_MAP_TYPE_ARRAY, 122 BPF_MAP_TYPE_PROG_ARRAY, 123 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 124 BPF_MAP_TYPE_PERCPU_HASH, 125 BPF_MAP_TYPE_PERCPU_ARRAY, 126 BPF_MAP_TYPE_STACK_TRACE, 127 BPF_MAP_TYPE_CGROUP_ARRAY, 128 BPF_MAP_TYPE_LRU_HASH, 129 BPF_MAP_TYPE_LRU_PERCPU_HASH, 130 BPF_MAP_TYPE_LPM_TRIE, 131 BPF_MAP_TYPE_ARRAY_OF_MAPS, 132 BPF_MAP_TYPE_HASH_OF_MAPS, 133 BPF_MAP_TYPE_DEVMAP, 134 BPF_MAP_TYPE_SOCKMAP, 135 BPF_MAP_TYPE_CPUMAP, 136 BPF_MAP_TYPE_XSKMAP, 137 BPF_MAP_TYPE_SOCKHASH, 138 BPF_MAP_TYPE_CGROUP_STORAGE, 139 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 140 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 141 BPF_MAP_TYPE_QUEUE, 142 BPF_MAP_TYPE_STACK, 143 BPF_MAP_TYPE_SK_STORAGE, 144 BPF_MAP_TYPE_DEVMAP_HASH, 145 BPF_MAP_TYPE_STRUCT_OPS, 146 }; 147 148 /* Note that tracing related programs such as 149 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 150 * are not subject to a stable API since kernel internal data 151 * structures can change from release to release and may 152 * therefore break existing tracing BPF programs. Tracing BPF 153 * programs correspond to /a/ specific kernel which is to be 154 * analyzed, and not /a/ specific kernel /and/ all future ones. 155 */ 156 enum bpf_prog_type { 157 BPF_PROG_TYPE_UNSPEC, 158 BPF_PROG_TYPE_SOCKET_FILTER, 159 BPF_PROG_TYPE_KPROBE, 160 BPF_PROG_TYPE_SCHED_CLS, 161 BPF_PROG_TYPE_SCHED_ACT, 162 BPF_PROG_TYPE_TRACEPOINT, 163 BPF_PROG_TYPE_XDP, 164 BPF_PROG_TYPE_PERF_EVENT, 165 BPF_PROG_TYPE_CGROUP_SKB, 166 BPF_PROG_TYPE_CGROUP_SOCK, 167 BPF_PROG_TYPE_LWT_IN, 168 BPF_PROG_TYPE_LWT_OUT, 169 BPF_PROG_TYPE_LWT_XMIT, 170 BPF_PROG_TYPE_SOCK_OPS, 171 BPF_PROG_TYPE_SK_SKB, 172 BPF_PROG_TYPE_CGROUP_DEVICE, 173 BPF_PROG_TYPE_SK_MSG, 174 BPF_PROG_TYPE_RAW_TRACEPOINT, 175 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 176 BPF_PROG_TYPE_LWT_SEG6LOCAL, 177 BPF_PROG_TYPE_LIRC_MODE2, 178 BPF_PROG_TYPE_SK_REUSEPORT, 179 BPF_PROG_TYPE_FLOW_DISSECTOR, 180 BPF_PROG_TYPE_CGROUP_SYSCTL, 181 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 182 BPF_PROG_TYPE_CGROUP_SOCKOPT, 183 BPF_PROG_TYPE_TRACING, 184 BPF_PROG_TYPE_STRUCT_OPS, 185 BPF_PROG_TYPE_EXT, 186 BPF_PROG_TYPE_LSM, 187 }; 188 189 enum bpf_attach_type { 190 BPF_CGROUP_INET_INGRESS, 191 BPF_CGROUP_INET_EGRESS, 192 BPF_CGROUP_INET_SOCK_CREATE, 193 BPF_CGROUP_SOCK_OPS, 194 BPF_SK_SKB_STREAM_PARSER, 195 BPF_SK_SKB_STREAM_VERDICT, 196 BPF_CGROUP_DEVICE, 197 BPF_SK_MSG_VERDICT, 198 BPF_CGROUP_INET4_BIND, 199 BPF_CGROUP_INET6_BIND, 200 BPF_CGROUP_INET4_CONNECT, 201 BPF_CGROUP_INET6_CONNECT, 202 BPF_CGROUP_INET4_POST_BIND, 203 BPF_CGROUP_INET6_POST_BIND, 204 BPF_CGROUP_UDP4_SENDMSG, 205 BPF_CGROUP_UDP6_SENDMSG, 206 BPF_LIRC_MODE2, 207 BPF_FLOW_DISSECTOR, 208 BPF_CGROUP_SYSCTL, 209 BPF_CGROUP_UDP4_RECVMSG, 210 BPF_CGROUP_UDP6_RECVMSG, 211 BPF_CGROUP_GETSOCKOPT, 212 BPF_CGROUP_SETSOCKOPT, 213 BPF_TRACE_RAW_TP, 214 BPF_TRACE_FENTRY, 215 BPF_TRACE_FEXIT, 216 BPF_MODIFY_RETURN, 217 BPF_LSM_MAC, 218 __MAX_BPF_ATTACH_TYPE 219 }; 220 221 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 222 223 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 224 * 225 * NONE(default): No further bpf programs allowed in the subtree. 226 * 227 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 228 * the program in this cgroup yields to sub-cgroup program. 229 * 230 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 231 * that cgroup program gets run in addition to the program in this cgroup. 232 * 233 * Only one program is allowed to be attached to a cgroup with 234 * NONE or BPF_F_ALLOW_OVERRIDE flag. 235 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 236 * release old program and attach the new one. Attach flags has to match. 237 * 238 * Multiple programs are allowed to be attached to a cgroup with 239 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 240 * (those that were attached first, run first) 241 * The programs of sub-cgroup are executed first, then programs of 242 * this cgroup and then programs of parent cgroup. 243 * When children program makes decision (like picking TCP CA or sock bind) 244 * parent program has a chance to override it. 245 * 246 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 247 * programs for a cgroup. Though it's possible to replace an old program at 248 * any position by also specifying BPF_F_REPLACE flag and position itself in 249 * replace_bpf_fd attribute. Old program at this position will be released. 250 * 251 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 252 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 253 * Ex1: 254 * cgrp1 (MULTI progs A, B) -> 255 * cgrp2 (OVERRIDE prog C) -> 256 * cgrp3 (MULTI prog D) -> 257 * cgrp4 (OVERRIDE prog E) -> 258 * cgrp5 (NONE prog F) 259 * the event in cgrp5 triggers execution of F,D,A,B in that order. 260 * if prog F is detached, the execution is E,D,A,B 261 * if prog F and D are detached, the execution is E,A,B 262 * if prog F, E and D are detached, the execution is C,A,B 263 * 264 * All eligible programs are executed regardless of return code from 265 * earlier programs. 266 */ 267 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 268 #define BPF_F_ALLOW_MULTI (1U << 1) 269 #define BPF_F_REPLACE (1U << 2) 270 271 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 272 * verifier will perform strict alignment checking as if the kernel 273 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 274 * and NET_IP_ALIGN defined to 2. 275 */ 276 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 277 278 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 279 * verifier will allow any alignment whatsoever. On platforms 280 * with strict alignment requirements for loads ands stores (such 281 * as sparc and mips) the verifier validates that all loads and 282 * stores provably follow this requirement. This flag turns that 283 * checking and enforcement off. 284 * 285 * It is mostly used for testing when we want to validate the 286 * context and memory access aspects of the verifier, but because 287 * of an unaligned access the alignment check would trigger before 288 * the one we are interested in. 289 */ 290 #define BPF_F_ANY_ALIGNMENT (1U << 1) 291 292 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 293 * Verifier does sub-register def/use analysis and identifies instructions whose 294 * def only matters for low 32-bit, high 32-bit is never referenced later 295 * through implicit zero extension. Therefore verifier notifies JIT back-ends 296 * that it is safe to ignore clearing high 32-bit for these instructions. This 297 * saves some back-ends a lot of code-gen. However such optimization is not 298 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 299 * hence hasn't used verifier's analysis result. But, we really want to have a 300 * way to be able to verify the correctness of the described optimization on 301 * x86_64 on which testsuites are frequently exercised. 302 * 303 * So, this flag is introduced. Once it is set, verifier will randomize high 304 * 32-bit for those instructions who has been identified as safe to ignore them. 305 * Then, if verifier is not doing correct analysis, such randomization will 306 * regress tests to expose bugs. 307 */ 308 #define BPF_F_TEST_RND_HI32 (1U << 2) 309 310 /* The verifier internal test flag. Behavior is undefined */ 311 #define BPF_F_TEST_STATE_FREQ (1U << 3) 312 313 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 314 * two extensions: 315 * 316 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 317 * insn[0].imm: map fd map fd 318 * insn[1].imm: 0 offset into value 319 * insn[0].off: 0 0 320 * insn[1].off: 0 0 321 * ldimm64 rewrite: address of map address of map[0]+offset 322 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 323 */ 324 #define BPF_PSEUDO_MAP_FD 1 325 #define BPF_PSEUDO_MAP_VALUE 2 326 327 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 328 * offset to another bpf function 329 */ 330 #define BPF_PSEUDO_CALL 1 331 332 /* flags for BPF_MAP_UPDATE_ELEM command */ 333 enum { 334 BPF_ANY = 0, /* create new element or update existing */ 335 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 336 BPF_EXIST = 2, /* update existing element */ 337 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 338 }; 339 340 /* flags for BPF_MAP_CREATE command */ 341 enum { 342 BPF_F_NO_PREALLOC = (1U << 0), 343 /* Instead of having one common LRU list in the 344 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 345 * which can scale and perform better. 346 * Note, the LRU nodes (including free nodes) cannot be moved 347 * across different LRU lists. 348 */ 349 BPF_F_NO_COMMON_LRU = (1U << 1), 350 /* Specify numa node during map creation */ 351 BPF_F_NUMA_NODE = (1U << 2), 352 353 /* Flags for accessing BPF object from syscall side. */ 354 BPF_F_RDONLY = (1U << 3), 355 BPF_F_WRONLY = (1U << 4), 356 357 /* Flag for stack_map, store build_id+offset instead of pointer */ 358 BPF_F_STACK_BUILD_ID = (1U << 5), 359 360 /* Zero-initialize hash function seed. This should only be used for testing. */ 361 BPF_F_ZERO_SEED = (1U << 6), 362 363 /* Flags for accessing BPF object from program side. */ 364 BPF_F_RDONLY_PROG = (1U << 7), 365 BPF_F_WRONLY_PROG = (1U << 8), 366 367 /* Clone map from listener for newly accepted socket */ 368 BPF_F_CLONE = (1U << 9), 369 370 /* Enable memory-mapping BPF map */ 371 BPF_F_MMAPABLE = (1U << 10), 372 }; 373 374 /* Flags for BPF_PROG_QUERY. */ 375 376 /* Query effective (directly attached + inherited from ancestor cgroups) 377 * programs that will be executed for events within a cgroup. 378 * attach_flags with this flag are returned only for directly attached programs. 379 */ 380 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 381 382 enum bpf_stack_build_id_status { 383 /* user space need an empty entry to identify end of a trace */ 384 BPF_STACK_BUILD_ID_EMPTY = 0, 385 /* with valid build_id and offset */ 386 BPF_STACK_BUILD_ID_VALID = 1, 387 /* couldn't get build_id, fallback to ip */ 388 BPF_STACK_BUILD_ID_IP = 2, 389 }; 390 391 #define BPF_BUILD_ID_SIZE 20 392 struct bpf_stack_build_id { 393 __s32 status; 394 unsigned char build_id[BPF_BUILD_ID_SIZE]; 395 union { 396 __u64 offset; 397 __u64 ip; 398 }; 399 }; 400 401 #define BPF_OBJ_NAME_LEN 16U 402 403 union bpf_attr { 404 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 405 __u32 map_type; /* one of enum bpf_map_type */ 406 __u32 key_size; /* size of key in bytes */ 407 __u32 value_size; /* size of value in bytes */ 408 __u32 max_entries; /* max number of entries in a map */ 409 __u32 map_flags; /* BPF_MAP_CREATE related 410 * flags defined above. 411 */ 412 __u32 inner_map_fd; /* fd pointing to the inner map */ 413 __u32 numa_node; /* numa node (effective only if 414 * BPF_F_NUMA_NODE is set). 415 */ 416 char map_name[BPF_OBJ_NAME_LEN]; 417 __u32 map_ifindex; /* ifindex of netdev to create on */ 418 __u32 btf_fd; /* fd pointing to a BTF type data */ 419 __u32 btf_key_type_id; /* BTF type_id of the key */ 420 __u32 btf_value_type_id; /* BTF type_id of the value */ 421 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 422 * struct stored as the 423 * map value 424 */ 425 }; 426 427 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 428 __u32 map_fd; 429 __aligned_u64 key; 430 union { 431 __aligned_u64 value; 432 __aligned_u64 next_key; 433 }; 434 __u64 flags; 435 }; 436 437 struct { /* struct used by BPF_MAP_*_BATCH commands */ 438 __aligned_u64 in_batch; /* start batch, 439 * NULL to start from beginning 440 */ 441 __aligned_u64 out_batch; /* output: next start batch */ 442 __aligned_u64 keys; 443 __aligned_u64 values; 444 __u32 count; /* input/output: 445 * input: # of key/value 446 * elements 447 * output: # of filled elements 448 */ 449 __u32 map_fd; 450 __u64 elem_flags; 451 __u64 flags; 452 } batch; 453 454 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 455 __u32 prog_type; /* one of enum bpf_prog_type */ 456 __u32 insn_cnt; 457 __aligned_u64 insns; 458 __aligned_u64 license; 459 __u32 log_level; /* verbosity level of verifier */ 460 __u32 log_size; /* size of user buffer */ 461 __aligned_u64 log_buf; /* user supplied buffer */ 462 __u32 kern_version; /* not used */ 463 __u32 prog_flags; 464 char prog_name[BPF_OBJ_NAME_LEN]; 465 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 466 /* For some prog types expected attach type must be known at 467 * load time to verify attach type specific parts of prog 468 * (context accesses, allowed helpers, etc). 469 */ 470 __u32 expected_attach_type; 471 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 472 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 473 __aligned_u64 func_info; /* func info */ 474 __u32 func_info_cnt; /* number of bpf_func_info records */ 475 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 476 __aligned_u64 line_info; /* line info */ 477 __u32 line_info_cnt; /* number of bpf_line_info records */ 478 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 479 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 480 }; 481 482 struct { /* anonymous struct used by BPF_OBJ_* commands */ 483 __aligned_u64 pathname; 484 __u32 bpf_fd; 485 __u32 file_flags; 486 }; 487 488 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 489 __u32 target_fd; /* container object to attach to */ 490 __u32 attach_bpf_fd; /* eBPF program to attach */ 491 __u32 attach_type; 492 __u32 attach_flags; 493 __u32 replace_bpf_fd; /* previously attached eBPF 494 * program to replace if 495 * BPF_F_REPLACE is used 496 */ 497 }; 498 499 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 500 __u32 prog_fd; 501 __u32 retval; 502 __u32 data_size_in; /* input: len of data_in */ 503 __u32 data_size_out; /* input/output: len of data_out 504 * returns ENOSPC if data_out 505 * is too small. 506 */ 507 __aligned_u64 data_in; 508 __aligned_u64 data_out; 509 __u32 repeat; 510 __u32 duration; 511 __u32 ctx_size_in; /* input: len of ctx_in */ 512 __u32 ctx_size_out; /* input/output: len of ctx_out 513 * returns ENOSPC if ctx_out 514 * is too small. 515 */ 516 __aligned_u64 ctx_in; 517 __aligned_u64 ctx_out; 518 } test; 519 520 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 521 union { 522 __u32 start_id; 523 __u32 prog_id; 524 __u32 map_id; 525 __u32 btf_id; 526 }; 527 __u32 next_id; 528 __u32 open_flags; 529 }; 530 531 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 532 __u32 bpf_fd; 533 __u32 info_len; 534 __aligned_u64 info; 535 } info; 536 537 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 538 __u32 target_fd; /* container object to query */ 539 __u32 attach_type; 540 __u32 query_flags; 541 __u32 attach_flags; 542 __aligned_u64 prog_ids; 543 __u32 prog_cnt; 544 } query; 545 546 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 547 __u64 name; 548 __u32 prog_fd; 549 } raw_tracepoint; 550 551 struct { /* anonymous struct for BPF_BTF_LOAD */ 552 __aligned_u64 btf; 553 __aligned_u64 btf_log_buf; 554 __u32 btf_size; 555 __u32 btf_log_size; 556 __u32 btf_log_level; 557 }; 558 559 struct { 560 __u32 pid; /* input: pid */ 561 __u32 fd; /* input: fd */ 562 __u32 flags; /* input: flags */ 563 __u32 buf_len; /* input/output: buf len */ 564 __aligned_u64 buf; /* input/output: 565 * tp_name for tracepoint 566 * symbol for kprobe 567 * filename for uprobe 568 */ 569 __u32 prog_id; /* output: prod_id */ 570 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 571 __u64 probe_offset; /* output: probe_offset */ 572 __u64 probe_addr; /* output: probe_addr */ 573 } task_fd_query; 574 575 struct { /* struct used by BPF_LINK_CREATE command */ 576 __u32 prog_fd; /* eBPF program to attach */ 577 __u32 target_fd; /* object to attach to */ 578 __u32 attach_type; /* attach type */ 579 __u32 flags; /* extra flags */ 580 } link_create; 581 582 struct { /* struct used by BPF_LINK_UPDATE command */ 583 __u32 link_fd; /* link fd */ 584 /* new program fd to update link with */ 585 __u32 new_prog_fd; 586 __u32 flags; /* extra flags */ 587 /* expected link's program fd; is specified only if 588 * BPF_F_REPLACE flag is set in flags */ 589 __u32 old_prog_fd; 590 } link_update; 591 592 } __attribute__((aligned(8))); 593 594 /* The description below is an attempt at providing documentation to eBPF 595 * developers about the multiple available eBPF helper functions. It can be 596 * parsed and used to produce a manual page. The workflow is the following, 597 * and requires the rst2man utility: 598 * 599 * $ ./scripts/bpf_helpers_doc.py \ 600 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 601 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 602 * $ man /tmp/bpf-helpers.7 603 * 604 * Note that in order to produce this external documentation, some RST 605 * formatting is used in the descriptions to get "bold" and "italics" in 606 * manual pages. Also note that the few trailing white spaces are 607 * intentional, removing them would break paragraphs for rst2man. 608 * 609 * Start of BPF helper function descriptions: 610 * 611 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 612 * Description 613 * Perform a lookup in *map* for an entry associated to *key*. 614 * Return 615 * Map value associated to *key*, or **NULL** if no entry was 616 * found. 617 * 618 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 619 * Description 620 * Add or update the value of the entry associated to *key* in 621 * *map* with *value*. *flags* is one of: 622 * 623 * **BPF_NOEXIST** 624 * The entry for *key* must not exist in the map. 625 * **BPF_EXIST** 626 * The entry for *key* must already exist in the map. 627 * **BPF_ANY** 628 * No condition on the existence of the entry for *key*. 629 * 630 * Flag value **BPF_NOEXIST** cannot be used for maps of types 631 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 632 * elements always exist), the helper would return an error. 633 * Return 634 * 0 on success, or a negative error in case of failure. 635 * 636 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 637 * Description 638 * Delete entry with *key* from *map*. 639 * Return 640 * 0 on success, or a negative error in case of failure. 641 * 642 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 643 * Description 644 * For tracing programs, safely attempt to read *size* bytes from 645 * kernel space address *unsafe_ptr* and store the data in *dst*. 646 * 647 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel() 648 * instead. 649 * Return 650 * 0 on success, or a negative error in case of failure. 651 * 652 * u64 bpf_ktime_get_ns(void) 653 * Description 654 * Return the time elapsed since system boot, in nanoseconds. 655 * Return 656 * Current *ktime*. 657 * 658 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 659 * Description 660 * This helper is a "printk()-like" facility for debugging. It 661 * prints a message defined by format *fmt* (of size *fmt_size*) 662 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 663 * available. It can take up to three additional **u64** 664 * arguments (as an eBPF helpers, the total number of arguments is 665 * limited to five). 666 * 667 * Each time the helper is called, it appends a line to the trace. 668 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 669 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 670 * The format of the trace is customizable, and the exact output 671 * one will get depends on the options set in 672 * *\/sys/kernel/debug/tracing/trace_options* (see also the 673 * *README* file under the same directory). However, it usually 674 * defaults to something like: 675 * 676 * :: 677 * 678 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 679 * 680 * In the above: 681 * 682 * * ``telnet`` is the name of the current task. 683 * * ``470`` is the PID of the current task. 684 * * ``001`` is the CPU number on which the task is 685 * running. 686 * * In ``.N..``, each character refers to a set of 687 * options (whether irqs are enabled, scheduling 688 * options, whether hard/softirqs are running, level of 689 * preempt_disabled respectively). **N** means that 690 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 691 * are set. 692 * * ``419421.045894`` is a timestamp. 693 * * ``0x00000001`` is a fake value used by BPF for the 694 * instruction pointer register. 695 * * ``<formatted msg>`` is the message formatted with 696 * *fmt*. 697 * 698 * The conversion specifiers supported by *fmt* are similar, but 699 * more limited than for printk(). They are **%d**, **%i**, 700 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 701 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 702 * of field, padding with zeroes, etc.) is available, and the 703 * helper will return **-EINVAL** (but print nothing) if it 704 * encounters an unknown specifier. 705 * 706 * Also, note that **bpf_trace_printk**\ () is slow, and should 707 * only be used for debugging purposes. For this reason, a notice 708 * bloc (spanning several lines) is printed to kernel logs and 709 * states that the helper should not be used "for production use" 710 * the first time this helper is used (or more precisely, when 711 * **trace_printk**\ () buffers are allocated). For passing values 712 * to user space, perf events should be preferred. 713 * Return 714 * The number of bytes written to the buffer, or a negative error 715 * in case of failure. 716 * 717 * u32 bpf_get_prandom_u32(void) 718 * Description 719 * Get a pseudo-random number. 720 * 721 * From a security point of view, this helper uses its own 722 * pseudo-random internal state, and cannot be used to infer the 723 * seed of other random functions in the kernel. However, it is 724 * essential to note that the generator used by the helper is not 725 * cryptographically secure. 726 * Return 727 * A random 32-bit unsigned value. 728 * 729 * u32 bpf_get_smp_processor_id(void) 730 * Description 731 * Get the SMP (symmetric multiprocessing) processor id. Note that 732 * all programs run with preemption disabled, which means that the 733 * SMP processor id is stable during all the execution of the 734 * program. 735 * Return 736 * The SMP id of the processor running the program. 737 * 738 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 739 * Description 740 * Store *len* bytes from address *from* into the packet 741 * associated to *skb*, at *offset*. *flags* are a combination of 742 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 743 * checksum for the packet after storing the bytes) and 744 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 745 * **->swhash** and *skb*\ **->l4hash** to 0). 746 * 747 * A call to this helper is susceptible to change the underlying 748 * packet buffer. Therefore, at load time, all checks on pointers 749 * previously done by the verifier are invalidated and must be 750 * performed again, if the helper is used in combination with 751 * direct packet access. 752 * Return 753 * 0 on success, or a negative error in case of failure. 754 * 755 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 756 * Description 757 * Recompute the layer 3 (e.g. IP) checksum for the packet 758 * associated to *skb*. Computation is incremental, so the helper 759 * must know the former value of the header field that was 760 * modified (*from*), the new value of this field (*to*), and the 761 * number of bytes (2 or 4) for this field, stored in *size*. 762 * Alternatively, it is possible to store the difference between 763 * the previous and the new values of the header field in *to*, by 764 * setting *from* and *size* to 0. For both methods, *offset* 765 * indicates the location of the IP checksum within the packet. 766 * 767 * This helper works in combination with **bpf_csum_diff**\ (), 768 * which does not update the checksum in-place, but offers more 769 * flexibility and can handle sizes larger than 2 or 4 for the 770 * checksum to update. 771 * 772 * A call to this helper is susceptible to change the underlying 773 * packet buffer. Therefore, at load time, all checks on pointers 774 * previously done by the verifier are invalidated and must be 775 * performed again, if the helper is used in combination with 776 * direct packet access. 777 * Return 778 * 0 on success, or a negative error in case of failure. 779 * 780 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 781 * Description 782 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 783 * packet associated to *skb*. Computation is incremental, so the 784 * helper must know the former value of the header field that was 785 * modified (*from*), the new value of this field (*to*), and the 786 * number of bytes (2 or 4) for this field, stored on the lowest 787 * four bits of *flags*. Alternatively, it is possible to store 788 * the difference between the previous and the new values of the 789 * header field in *to*, by setting *from* and the four lowest 790 * bits of *flags* to 0. For both methods, *offset* indicates the 791 * location of the IP checksum within the packet. In addition to 792 * the size of the field, *flags* can be added (bitwise OR) actual 793 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 794 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 795 * for updates resulting in a null checksum the value is set to 796 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 797 * the checksum is to be computed against a pseudo-header. 798 * 799 * This helper works in combination with **bpf_csum_diff**\ (), 800 * which does not update the checksum in-place, but offers more 801 * flexibility and can handle sizes larger than 2 or 4 for the 802 * checksum to update. 803 * 804 * A call to this helper is susceptible to change the underlying 805 * packet buffer. Therefore, at load time, all checks on pointers 806 * previously done by the verifier are invalidated and must be 807 * performed again, if the helper is used in combination with 808 * direct packet access. 809 * Return 810 * 0 on success, or a negative error in case of failure. 811 * 812 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 813 * Description 814 * This special helper is used to trigger a "tail call", or in 815 * other words, to jump into another eBPF program. The same stack 816 * frame is used (but values on stack and in registers for the 817 * caller are not accessible to the callee). This mechanism allows 818 * for program chaining, either for raising the maximum number of 819 * available eBPF instructions, or to execute given programs in 820 * conditional blocks. For security reasons, there is an upper 821 * limit to the number of successive tail calls that can be 822 * performed. 823 * 824 * Upon call of this helper, the program attempts to jump into a 825 * program referenced at index *index* in *prog_array_map*, a 826 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 827 * *ctx*, a pointer to the context. 828 * 829 * If the call succeeds, the kernel immediately runs the first 830 * instruction of the new program. This is not a function call, 831 * and it never returns to the previous program. If the call 832 * fails, then the helper has no effect, and the caller continues 833 * to run its subsequent instructions. A call can fail if the 834 * destination program for the jump does not exist (i.e. *index* 835 * is superior to the number of entries in *prog_array_map*), or 836 * if the maximum number of tail calls has been reached for this 837 * chain of programs. This limit is defined in the kernel by the 838 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 839 * which is currently set to 32. 840 * Return 841 * 0 on success, or a negative error in case of failure. 842 * 843 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 844 * Description 845 * Clone and redirect the packet associated to *skb* to another 846 * net device of index *ifindex*. Both ingress and egress 847 * interfaces can be used for redirection. The **BPF_F_INGRESS** 848 * value in *flags* is used to make the distinction (ingress path 849 * is selected if the flag is present, egress path otherwise). 850 * This is the only flag supported for now. 851 * 852 * In comparison with **bpf_redirect**\ () helper, 853 * **bpf_clone_redirect**\ () has the associated cost of 854 * duplicating the packet buffer, but this can be executed out of 855 * the eBPF program. Conversely, **bpf_redirect**\ () is more 856 * efficient, but it is handled through an action code where the 857 * redirection happens only after the eBPF program has returned. 858 * 859 * A call to this helper is susceptible to change the underlying 860 * packet buffer. Therefore, at load time, all checks on pointers 861 * previously done by the verifier are invalidated and must be 862 * performed again, if the helper is used in combination with 863 * direct packet access. 864 * Return 865 * 0 on success, or a negative error in case of failure. 866 * 867 * u64 bpf_get_current_pid_tgid(void) 868 * Return 869 * A 64-bit integer containing the current tgid and pid, and 870 * created as such: 871 * *current_task*\ **->tgid << 32 \|** 872 * *current_task*\ **->pid**. 873 * 874 * u64 bpf_get_current_uid_gid(void) 875 * Return 876 * A 64-bit integer containing the current GID and UID, and 877 * created as such: *current_gid* **<< 32 \|** *current_uid*. 878 * 879 * int bpf_get_current_comm(void *buf, u32 size_of_buf) 880 * Description 881 * Copy the **comm** attribute of the current task into *buf* of 882 * *size_of_buf*. The **comm** attribute contains the name of 883 * the executable (excluding the path) for the current task. The 884 * *size_of_buf* must be strictly positive. On success, the 885 * helper makes sure that the *buf* is NUL-terminated. On failure, 886 * it is filled with zeroes. 887 * Return 888 * 0 on success, or a negative error in case of failure. 889 * 890 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 891 * Description 892 * Retrieve the classid for the current task, i.e. for the net_cls 893 * cgroup to which *skb* belongs. 894 * 895 * This helper can be used on TC egress path, but not on ingress. 896 * 897 * The net_cls cgroup provides an interface to tag network packets 898 * based on a user-provided identifier for all traffic coming from 899 * the tasks belonging to the related cgroup. See also the related 900 * kernel documentation, available from the Linux sources in file 901 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 902 * 903 * The Linux kernel has two versions for cgroups: there are 904 * cgroups v1 and cgroups v2. Both are available to users, who can 905 * use a mixture of them, but note that the net_cls cgroup is for 906 * cgroup v1 only. This makes it incompatible with BPF programs 907 * run on cgroups, which is a cgroup-v2-only feature (a socket can 908 * only hold data for one version of cgroups at a time). 909 * 910 * This helper is only available is the kernel was compiled with 911 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 912 * "**y**" or to "**m**". 913 * Return 914 * The classid, or 0 for the default unconfigured classid. 915 * 916 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 917 * Description 918 * Push a *vlan_tci* (VLAN tag control information) of protocol 919 * *vlan_proto* to the packet associated to *skb*, then update 920 * the checksum. Note that if *vlan_proto* is different from 921 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 922 * be **ETH_P_8021Q**. 923 * 924 * A call to this helper is susceptible to change the underlying 925 * packet buffer. Therefore, at load time, all checks on pointers 926 * previously done by the verifier are invalidated and must be 927 * performed again, if the helper is used in combination with 928 * direct packet access. 929 * Return 930 * 0 on success, or a negative error in case of failure. 931 * 932 * int bpf_skb_vlan_pop(struct sk_buff *skb) 933 * Description 934 * Pop a VLAN header from the packet associated to *skb*. 935 * 936 * A call to this helper is susceptible to change the underlying 937 * packet buffer. Therefore, at load time, all checks on pointers 938 * previously done by the verifier are invalidated and must be 939 * performed again, if the helper is used in combination with 940 * direct packet access. 941 * Return 942 * 0 on success, or a negative error in case of failure. 943 * 944 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 945 * Description 946 * Get tunnel metadata. This helper takes a pointer *key* to an 947 * empty **struct bpf_tunnel_key** of **size**, that will be 948 * filled with tunnel metadata for the packet associated to *skb*. 949 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 950 * indicates that the tunnel is based on IPv6 protocol instead of 951 * IPv4. 952 * 953 * The **struct bpf_tunnel_key** is an object that generalizes the 954 * principal parameters used by various tunneling protocols into a 955 * single struct. This way, it can be used to easily make a 956 * decision based on the contents of the encapsulation header, 957 * "summarized" in this struct. In particular, it holds the IP 958 * address of the remote end (IPv4 or IPv6, depending on the case) 959 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 960 * this struct exposes the *key*\ **->tunnel_id**, which is 961 * generally mapped to a VNI (Virtual Network Identifier), making 962 * it programmable together with the **bpf_skb_set_tunnel_key**\ 963 * () helper. 964 * 965 * Let's imagine that the following code is part of a program 966 * attached to the TC ingress interface, on one end of a GRE 967 * tunnel, and is supposed to filter out all messages coming from 968 * remote ends with IPv4 address other than 10.0.0.1: 969 * 970 * :: 971 * 972 * int ret; 973 * struct bpf_tunnel_key key = {}; 974 * 975 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 976 * if (ret < 0) 977 * return TC_ACT_SHOT; // drop packet 978 * 979 * if (key.remote_ipv4 != 0x0a000001) 980 * return TC_ACT_SHOT; // drop packet 981 * 982 * return TC_ACT_OK; // accept packet 983 * 984 * This interface can also be used with all encapsulation devices 985 * that can operate in "collect metadata" mode: instead of having 986 * one network device per specific configuration, the "collect 987 * metadata" mode only requires a single device where the 988 * configuration can be extracted from this helper. 989 * 990 * This can be used together with various tunnels such as VXLan, 991 * Geneve, GRE or IP in IP (IPIP). 992 * Return 993 * 0 on success, or a negative error in case of failure. 994 * 995 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 996 * Description 997 * Populate tunnel metadata for packet associated to *skb.* The 998 * tunnel metadata is set to the contents of *key*, of *size*. The 999 * *flags* can be set to a combination of the following values: 1000 * 1001 * **BPF_F_TUNINFO_IPV6** 1002 * Indicate that the tunnel is based on IPv6 protocol 1003 * instead of IPv4. 1004 * **BPF_F_ZERO_CSUM_TX** 1005 * For IPv4 packets, add a flag to tunnel metadata 1006 * indicating that checksum computation should be skipped 1007 * and checksum set to zeroes. 1008 * **BPF_F_DONT_FRAGMENT** 1009 * Add a flag to tunnel metadata indicating that the 1010 * packet should not be fragmented. 1011 * **BPF_F_SEQ_NUMBER** 1012 * Add a flag to tunnel metadata indicating that a 1013 * sequence number should be added to tunnel header before 1014 * sending the packet. This flag was added for GRE 1015 * encapsulation, but might be used with other protocols 1016 * as well in the future. 1017 * 1018 * Here is a typical usage on the transmit path: 1019 * 1020 * :: 1021 * 1022 * struct bpf_tunnel_key key; 1023 * populate key ... 1024 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1025 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1026 * 1027 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1028 * helper for additional information. 1029 * Return 1030 * 0 on success, or a negative error in case of failure. 1031 * 1032 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1033 * Description 1034 * Read the value of a perf event counter. This helper relies on a 1035 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1036 * the perf event counter is selected when *map* is updated with 1037 * perf event file descriptors. The *map* is an array whose size 1038 * is the number of available CPUs, and each cell contains a value 1039 * relative to one CPU. The value to retrieve is indicated by 1040 * *flags*, that contains the index of the CPU to look up, masked 1041 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1042 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1043 * current CPU should be retrieved. 1044 * 1045 * Note that before Linux 4.13, only hardware perf event can be 1046 * retrieved. 1047 * 1048 * Also, be aware that the newer helper 1049 * **bpf_perf_event_read_value**\ () is recommended over 1050 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1051 * quirks where error and counter value are used as a return code 1052 * (which is wrong to do since ranges may overlap). This issue is 1053 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1054 * time provides more features over the **bpf_perf_event_read**\ 1055 * () interface. Please refer to the description of 1056 * **bpf_perf_event_read_value**\ () for details. 1057 * Return 1058 * The value of the perf event counter read from the map, or a 1059 * negative error code in case of failure. 1060 * 1061 * int bpf_redirect(u32 ifindex, u64 flags) 1062 * Description 1063 * Redirect the packet to another net device of index *ifindex*. 1064 * This helper is somewhat similar to **bpf_clone_redirect**\ 1065 * (), except that the packet is not cloned, which provides 1066 * increased performance. 1067 * 1068 * Except for XDP, both ingress and egress interfaces can be used 1069 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1070 * to make the distinction (ingress path is selected if the flag 1071 * is present, egress path otherwise). Currently, XDP only 1072 * supports redirection to the egress interface, and accepts no 1073 * flag at all. 1074 * 1075 * The same effect can also be attained with the more generic 1076 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1077 * redirect target instead of providing it directly to the helper. 1078 * Return 1079 * For XDP, the helper returns **XDP_REDIRECT** on success or 1080 * **XDP_ABORTED** on error. For other program types, the values 1081 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1082 * error. 1083 * 1084 * u32 bpf_get_route_realm(struct sk_buff *skb) 1085 * Description 1086 * Retrieve the realm or the route, that is to say the 1087 * **tclassid** field of the destination for the *skb*. The 1088 * indentifier retrieved is a user-provided tag, similar to the 1089 * one used with the net_cls cgroup (see description for 1090 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1091 * held by a route (a destination entry), not by a task. 1092 * 1093 * Retrieving this identifier works with the clsact TC egress hook 1094 * (see also **tc-bpf(8)**), or alternatively on conventional 1095 * classful egress qdiscs, but not on TC ingress path. In case of 1096 * clsact TC egress hook, this has the advantage that, internally, 1097 * the destination entry has not been dropped yet in the transmit 1098 * path. Therefore, the destination entry does not need to be 1099 * artificially held via **netif_keep_dst**\ () for a classful 1100 * qdisc until the *skb* is freed. 1101 * 1102 * This helper is available only if the kernel was compiled with 1103 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1104 * Return 1105 * The realm of the route for the packet associated to *skb*, or 0 1106 * if none was found. 1107 * 1108 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1109 * Description 1110 * Write raw *data* blob into a special BPF perf event held by 1111 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1112 * event must have the following attributes: **PERF_SAMPLE_RAW** 1113 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1114 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1115 * 1116 * The *flags* are used to indicate the index in *map* for which 1117 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1118 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1119 * to indicate that the index of the current CPU core should be 1120 * used. 1121 * 1122 * The value to write, of *size*, is passed through eBPF stack and 1123 * pointed by *data*. 1124 * 1125 * The context of the program *ctx* needs also be passed to the 1126 * helper. 1127 * 1128 * On user space, a program willing to read the values needs to 1129 * call **perf_event_open**\ () on the perf event (either for 1130 * one or for all CPUs) and to store the file descriptor into the 1131 * *map*. This must be done before the eBPF program can send data 1132 * into it. An example is available in file 1133 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1134 * tree (the eBPF program counterpart is in 1135 * *samples/bpf/trace_output_kern.c*). 1136 * 1137 * **bpf_perf_event_output**\ () achieves better performance 1138 * than **bpf_trace_printk**\ () for sharing data with user 1139 * space, and is much better suitable for streaming data from eBPF 1140 * programs. 1141 * 1142 * Note that this helper is not restricted to tracing use cases 1143 * and can be used with programs attached to TC or XDP as well, 1144 * where it allows for passing data to user space listeners. Data 1145 * can be: 1146 * 1147 * * Only custom structs, 1148 * * Only the packet payload, or 1149 * * A combination of both. 1150 * Return 1151 * 0 on success, or a negative error in case of failure. 1152 * 1153 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1154 * Description 1155 * This helper was provided as an easy way to load data from a 1156 * packet. It can be used to load *len* bytes from *offset* from 1157 * the packet associated to *skb*, into the buffer pointed by 1158 * *to*. 1159 * 1160 * Since Linux 4.7, usage of this helper has mostly been replaced 1161 * by "direct packet access", enabling packet data to be 1162 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1163 * pointing respectively to the first byte of packet data and to 1164 * the byte after the last byte of packet data. However, it 1165 * remains useful if one wishes to read large quantities of data 1166 * at once from a packet into the eBPF stack. 1167 * Return 1168 * 0 on success, or a negative error in case of failure. 1169 * 1170 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1171 * Description 1172 * Walk a user or a kernel stack and return its id. To achieve 1173 * this, the helper needs *ctx*, which is a pointer to the context 1174 * on which the tracing program is executed, and a pointer to a 1175 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1176 * 1177 * The last argument, *flags*, holds the number of stack frames to 1178 * skip (from 0 to 255), masked with 1179 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1180 * a combination of the following flags: 1181 * 1182 * **BPF_F_USER_STACK** 1183 * Collect a user space stack instead of a kernel stack. 1184 * **BPF_F_FAST_STACK_CMP** 1185 * Compare stacks by hash only. 1186 * **BPF_F_REUSE_STACKID** 1187 * If two different stacks hash into the same *stackid*, 1188 * discard the old one. 1189 * 1190 * The stack id retrieved is a 32 bit long integer handle which 1191 * can be further combined with other data (including other stack 1192 * ids) and used as a key into maps. This can be useful for 1193 * generating a variety of graphs (such as flame graphs or off-cpu 1194 * graphs). 1195 * 1196 * For walking a stack, this helper is an improvement over 1197 * **bpf_probe_read**\ (), which can be used with unrolled loops 1198 * but is not efficient and consumes a lot of eBPF instructions. 1199 * Instead, **bpf_get_stackid**\ () can collect up to 1200 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1201 * this limit can be controlled with the **sysctl** program, and 1202 * that it should be manually increased in order to profile long 1203 * user stacks (such as stacks for Java programs). To do so, use: 1204 * 1205 * :: 1206 * 1207 * # sysctl kernel.perf_event_max_stack=<new value> 1208 * Return 1209 * The positive or null stack id on success, or a negative error 1210 * in case of failure. 1211 * 1212 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1213 * Description 1214 * Compute a checksum difference, from the raw buffer pointed by 1215 * *from*, of length *from_size* (that must be a multiple of 4), 1216 * towards the raw buffer pointed by *to*, of size *to_size* 1217 * (same remark). An optional *seed* can be added to the value 1218 * (this can be cascaded, the seed may come from a previous call 1219 * to the helper). 1220 * 1221 * This is flexible enough to be used in several ways: 1222 * 1223 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1224 * checksum, it can be used when pushing new data. 1225 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1226 * checksum, it can be used when removing data from a packet. 1227 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1228 * can be used to compute a diff. Note that *from_size* and 1229 * *to_size* do not need to be equal. 1230 * 1231 * This helper can be used in combination with 1232 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1233 * which one can feed in the difference computed with 1234 * **bpf_csum_diff**\ (). 1235 * Return 1236 * The checksum result, or a negative error code in case of 1237 * failure. 1238 * 1239 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1240 * Description 1241 * Retrieve tunnel options metadata for the packet associated to 1242 * *skb*, and store the raw tunnel option data to the buffer *opt* 1243 * of *size*. 1244 * 1245 * This helper can be used with encapsulation devices that can 1246 * operate in "collect metadata" mode (please refer to the related 1247 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1248 * more details). A particular example where this can be used is 1249 * in combination with the Geneve encapsulation protocol, where it 1250 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1251 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1252 * the eBPF program. This allows for full customization of these 1253 * headers. 1254 * Return 1255 * The size of the option data retrieved. 1256 * 1257 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1258 * Description 1259 * Set tunnel options metadata for the packet associated to *skb* 1260 * to the option data contained in the raw buffer *opt* of *size*. 1261 * 1262 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1263 * helper for additional information. 1264 * Return 1265 * 0 on success, or a negative error in case of failure. 1266 * 1267 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1268 * Description 1269 * Change the protocol of the *skb* to *proto*. Currently 1270 * supported are transition from IPv4 to IPv6, and from IPv6 to 1271 * IPv4. The helper takes care of the groundwork for the 1272 * transition, including resizing the socket buffer. The eBPF 1273 * program is expected to fill the new headers, if any, via 1274 * **skb_store_bytes**\ () and to recompute the checksums with 1275 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1276 * (). The main case for this helper is to perform NAT64 1277 * operations out of an eBPF program. 1278 * 1279 * Internally, the GSO type is marked as dodgy so that headers are 1280 * checked and segments are recalculated by the GSO/GRO engine. 1281 * The size for GSO target is adapted as well. 1282 * 1283 * All values for *flags* are reserved for future usage, and must 1284 * be left at zero. 1285 * 1286 * A call to this helper is susceptible to change the underlying 1287 * packet buffer. Therefore, at load time, all checks on pointers 1288 * previously done by the verifier are invalidated and must be 1289 * performed again, if the helper is used in combination with 1290 * direct packet access. 1291 * Return 1292 * 0 on success, or a negative error in case of failure. 1293 * 1294 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1295 * Description 1296 * Change the packet type for the packet associated to *skb*. This 1297 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1298 * the eBPF program does not have a write access to *skb*\ 1299 * **->pkt_type** beside this helper. Using a helper here allows 1300 * for graceful handling of errors. 1301 * 1302 * The major use case is to change incoming *skb*s to 1303 * **PACKET_HOST** in a programmatic way instead of having to 1304 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1305 * example. 1306 * 1307 * Note that *type* only allows certain values. At this time, they 1308 * are: 1309 * 1310 * **PACKET_HOST** 1311 * Packet is for us. 1312 * **PACKET_BROADCAST** 1313 * Send packet to all. 1314 * **PACKET_MULTICAST** 1315 * Send packet to group. 1316 * **PACKET_OTHERHOST** 1317 * Send packet to someone else. 1318 * Return 1319 * 0 on success, or a negative error in case of failure. 1320 * 1321 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1322 * Description 1323 * Check whether *skb* is a descendant of the cgroup2 held by 1324 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1325 * Return 1326 * The return value depends on the result of the test, and can be: 1327 * 1328 * * 0, if the *skb* failed the cgroup2 descendant test. 1329 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1330 * * A negative error code, if an error occurred. 1331 * 1332 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1333 * Description 1334 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1335 * not set, in particular if the hash was cleared due to mangling, 1336 * recompute this hash. Later accesses to the hash can be done 1337 * directly with *skb*\ **->hash**. 1338 * 1339 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1340 * prototype with **bpf_skb_change_proto**\ (), or calling 1341 * **bpf_skb_store_bytes**\ () with the 1342 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1343 * the hash and to trigger a new computation for the next call to 1344 * **bpf_get_hash_recalc**\ (). 1345 * Return 1346 * The 32-bit hash. 1347 * 1348 * u64 bpf_get_current_task(void) 1349 * Return 1350 * A pointer to the current task struct. 1351 * 1352 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1353 * Description 1354 * Attempt in a safe way to write *len* bytes from the buffer 1355 * *src* to *dst* in memory. It only works for threads that are in 1356 * user context, and *dst* must be a valid user space address. 1357 * 1358 * This helper should not be used to implement any kind of 1359 * security mechanism because of TOC-TOU attacks, but rather to 1360 * debug, divert, and manipulate execution of semi-cooperative 1361 * processes. 1362 * 1363 * Keep in mind that this feature is meant for experiments, and it 1364 * has a risk of crashing the system and running programs. 1365 * Therefore, when an eBPF program using this helper is attached, 1366 * a warning including PID and process name is printed to kernel 1367 * logs. 1368 * Return 1369 * 0 on success, or a negative error in case of failure. 1370 * 1371 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1372 * Description 1373 * Check whether the probe is being run is the context of a given 1374 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1375 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1376 * Return 1377 * The return value depends on the result of the test, and can be: 1378 * 1379 * * 0, if the *skb* task belongs to the cgroup2. 1380 * * 1, if the *skb* task does not belong to the cgroup2. 1381 * * A negative error code, if an error occurred. 1382 * 1383 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1384 * Description 1385 * Resize (trim or grow) the packet associated to *skb* to the 1386 * new *len*. The *flags* are reserved for future usage, and must 1387 * be left at zero. 1388 * 1389 * The basic idea is that the helper performs the needed work to 1390 * change the size of the packet, then the eBPF program rewrites 1391 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1392 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1393 * and others. This helper is a slow path utility intended for 1394 * replies with control messages. And because it is targeted for 1395 * slow path, the helper itself can afford to be slow: it 1396 * implicitly linearizes, unclones and drops offloads from the 1397 * *skb*. 1398 * 1399 * A call to this helper is susceptible to change the underlying 1400 * packet buffer. Therefore, at load time, all checks on pointers 1401 * previously done by the verifier are invalidated and must be 1402 * performed again, if the helper is used in combination with 1403 * direct packet access. 1404 * Return 1405 * 0 on success, or a negative error in case of failure. 1406 * 1407 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1408 * Description 1409 * Pull in non-linear data in case the *skb* is non-linear and not 1410 * all of *len* are part of the linear section. Make *len* bytes 1411 * from *skb* readable and writable. If a zero value is passed for 1412 * *len*, then the whole length of the *skb* is pulled. 1413 * 1414 * This helper is only needed for reading and writing with direct 1415 * packet access. 1416 * 1417 * For direct packet access, testing that offsets to access 1418 * are within packet boundaries (test on *skb*\ **->data_end**) is 1419 * susceptible to fail if offsets are invalid, or if the requested 1420 * data is in non-linear parts of the *skb*. On failure the 1421 * program can just bail out, or in the case of a non-linear 1422 * buffer, use a helper to make the data available. The 1423 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1424 * the data. Another one consists in using **bpf_skb_pull_data** 1425 * to pull in once the non-linear parts, then retesting and 1426 * eventually access the data. 1427 * 1428 * At the same time, this also makes sure the *skb* is uncloned, 1429 * which is a necessary condition for direct write. As this needs 1430 * to be an invariant for the write part only, the verifier 1431 * detects writes and adds a prologue that is calling 1432 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1433 * the very beginning in case it is indeed cloned. 1434 * 1435 * A call to this helper is susceptible to change the underlying 1436 * packet buffer. Therefore, at load time, all checks on pointers 1437 * previously done by the verifier are invalidated and must be 1438 * performed again, if the helper is used in combination with 1439 * direct packet access. 1440 * Return 1441 * 0 on success, or a negative error in case of failure. 1442 * 1443 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1444 * Description 1445 * Add the checksum *csum* into *skb*\ **->csum** in case the 1446 * driver has supplied a checksum for the entire packet into that 1447 * field. Return an error otherwise. This helper is intended to be 1448 * used in combination with **bpf_csum_diff**\ (), in particular 1449 * when the checksum needs to be updated after data has been 1450 * written into the packet through direct packet access. 1451 * Return 1452 * The checksum on success, or a negative error code in case of 1453 * failure. 1454 * 1455 * void bpf_set_hash_invalid(struct sk_buff *skb) 1456 * Description 1457 * Invalidate the current *skb*\ **->hash**. It can be used after 1458 * mangling on headers through direct packet access, in order to 1459 * indicate that the hash is outdated and to trigger a 1460 * recalculation the next time the kernel tries to access this 1461 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1462 * 1463 * int bpf_get_numa_node_id(void) 1464 * Description 1465 * Return the id of the current NUMA node. The primary use case 1466 * for this helper is the selection of sockets for the local NUMA 1467 * node, when the program is attached to sockets using the 1468 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1469 * but the helper is also available to other eBPF program types, 1470 * similarly to **bpf_get_smp_processor_id**\ (). 1471 * Return 1472 * The id of current NUMA node. 1473 * 1474 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1475 * Description 1476 * Grows headroom of packet associated to *skb* and adjusts the 1477 * offset of the MAC header accordingly, adding *len* bytes of 1478 * space. It automatically extends and reallocates memory as 1479 * required. 1480 * 1481 * This helper can be used on a layer 3 *skb* to push a MAC header 1482 * for redirection into a layer 2 device. 1483 * 1484 * All values for *flags* are reserved for future usage, and must 1485 * be left at zero. 1486 * 1487 * A call to this helper is susceptible to change the underlying 1488 * packet buffer. Therefore, at load time, all checks on pointers 1489 * previously done by the verifier are invalidated and must be 1490 * performed again, if the helper is used in combination with 1491 * direct packet access. 1492 * Return 1493 * 0 on success, or a negative error in case of failure. 1494 * 1495 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1496 * Description 1497 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1498 * it is possible to use a negative value for *delta*. This helper 1499 * can be used to prepare the packet for pushing or popping 1500 * headers. 1501 * 1502 * A call to this helper is susceptible to change the underlying 1503 * packet buffer. Therefore, at load time, all checks on pointers 1504 * previously done by the verifier are invalidated and must be 1505 * performed again, if the helper is used in combination with 1506 * direct packet access. 1507 * Return 1508 * 0 on success, or a negative error in case of failure. 1509 * 1510 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1511 * Description 1512 * Copy a NUL terminated string from an unsafe kernel address 1513 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for 1514 * more details. 1515 * 1516 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str() 1517 * instead. 1518 * Return 1519 * On success, the strictly positive length of the string, 1520 * including the trailing NUL character. On error, a negative 1521 * value. 1522 * 1523 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1524 * Description 1525 * If the **struct sk_buff** pointed by *skb* has a known socket, 1526 * retrieve the cookie (generated by the kernel) of this socket. 1527 * If no cookie has been set yet, generate a new cookie. Once 1528 * generated, the socket cookie remains stable for the life of the 1529 * socket. This helper can be useful for monitoring per socket 1530 * networking traffic statistics as it provides a global socket 1531 * identifier that can be assumed unique. 1532 * Return 1533 * A 8-byte long non-decreasing number on success, or 0 if the 1534 * socket field is missing inside *skb*. 1535 * 1536 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1537 * Description 1538 * Equivalent to bpf_get_socket_cookie() helper that accepts 1539 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1540 * Return 1541 * A 8-byte long non-decreasing number. 1542 * 1543 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1544 * Description 1545 * Equivalent to bpf_get_socket_cookie() helper that accepts 1546 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1547 * Return 1548 * A 8-byte long non-decreasing number. 1549 * 1550 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1551 * Return 1552 * The owner UID of the socket associated to *skb*. If the socket 1553 * is **NULL**, or if it is not a full socket (i.e. if it is a 1554 * time-wait or a request socket instead), **overflowuid** value 1555 * is returned (note that **overflowuid** might also be the actual 1556 * UID value for the socket). 1557 * 1558 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1559 * Description 1560 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1561 * to value *hash*. 1562 * Return 1563 * 0 1564 * 1565 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1566 * Description 1567 * Emulate a call to **setsockopt()** on the socket associated to 1568 * *bpf_socket*, which must be a full socket. The *level* at 1569 * which the option resides and the name *optname* of the option 1570 * must be specified, see **setsockopt(2)** for more information. 1571 * The option value of length *optlen* is pointed by *optval*. 1572 * 1573 * This helper actually implements a subset of **setsockopt()**. 1574 * It supports the following *level*\ s: 1575 * 1576 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1577 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1578 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1579 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1580 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1581 * **TCP_BPF_SNDCWND_CLAMP**. 1582 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1583 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1584 * Return 1585 * 0 on success, or a negative error in case of failure. 1586 * 1587 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1588 * Description 1589 * Grow or shrink the room for data in the packet associated to 1590 * *skb* by *len_diff*, and according to the selected *mode*. 1591 * 1592 * There are two supported modes at this time: 1593 * 1594 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1595 * (room space is added or removed below the layer 2 header). 1596 * 1597 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1598 * (room space is added or removed below the layer 3 header). 1599 * 1600 * The following flags are supported at this time: 1601 * 1602 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1603 * Adjusting mss in this way is not allowed for datagrams. 1604 * 1605 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1606 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1607 * Any new space is reserved to hold a tunnel header. 1608 * Configure skb offsets and other fields accordingly. 1609 * 1610 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1611 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1612 * Use with ENCAP_L3 flags to further specify the tunnel type. 1613 * 1614 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1615 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1616 * type; *len* is the length of the inner MAC header. 1617 * 1618 * A call to this helper is susceptible to change the underlying 1619 * packet buffer. Therefore, at load time, all checks on pointers 1620 * previously done by the verifier are invalidated and must be 1621 * performed again, if the helper is used in combination with 1622 * direct packet access. 1623 * Return 1624 * 0 on success, or a negative error in case of failure. 1625 * 1626 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1627 * Description 1628 * Redirect the packet to the endpoint referenced by *map* at 1629 * index *key*. Depending on its type, this *map* can contain 1630 * references to net devices (for forwarding packets through other 1631 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1632 * but this is only implemented for native XDP (with driver 1633 * support) as of this writing). 1634 * 1635 * The lower two bits of *flags* are used as the return code if 1636 * the map lookup fails. This is so that the return value can be 1637 * one of the XDP program return codes up to XDP_TX, as chosen by 1638 * the caller. Any higher bits in the *flags* argument must be 1639 * unset. 1640 * 1641 * See also bpf_redirect(), which only supports redirecting to an 1642 * ifindex, but doesn't require a map to do so. 1643 * Return 1644 * **XDP_REDIRECT** on success, or the value of the two lower bits 1645 * of the **flags* argument on error. 1646 * 1647 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1648 * Description 1649 * Redirect the packet to the socket referenced by *map* (of type 1650 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1651 * egress interfaces can be used for redirection. The 1652 * **BPF_F_INGRESS** value in *flags* is used to make the 1653 * distinction (ingress path is selected if the flag is present, 1654 * egress path otherwise). This is the only flag supported for now. 1655 * Return 1656 * **SK_PASS** on success, or **SK_DROP** on error. 1657 * 1658 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1659 * Description 1660 * Add an entry to, or update a *map* referencing sockets. The 1661 * *skops* is used as a new value for the entry associated to 1662 * *key*. *flags* is one of: 1663 * 1664 * **BPF_NOEXIST** 1665 * The entry for *key* must not exist in the map. 1666 * **BPF_EXIST** 1667 * The entry for *key* must already exist in the map. 1668 * **BPF_ANY** 1669 * No condition on the existence of the entry for *key*. 1670 * 1671 * If the *map* has eBPF programs (parser and verdict), those will 1672 * be inherited by the socket being added. If the socket is 1673 * already attached to eBPF programs, this results in an error. 1674 * Return 1675 * 0 on success, or a negative error in case of failure. 1676 * 1677 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1678 * Description 1679 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1680 * *delta* (which can be positive or negative). Note that this 1681 * operation modifies the address stored in *xdp_md*\ **->data**, 1682 * so the latter must be loaded only after the helper has been 1683 * called. 1684 * 1685 * The use of *xdp_md*\ **->data_meta** is optional and programs 1686 * are not required to use it. The rationale is that when the 1687 * packet is processed with XDP (e.g. as DoS filter), it is 1688 * possible to push further meta data along with it before passing 1689 * to the stack, and to give the guarantee that an ingress eBPF 1690 * program attached as a TC classifier on the same device can pick 1691 * this up for further post-processing. Since TC works with socket 1692 * buffers, it remains possible to set from XDP the **mark** or 1693 * **priority** pointers, or other pointers for the socket buffer. 1694 * Having this scratch space generic and programmable allows for 1695 * more flexibility as the user is free to store whatever meta 1696 * data they need. 1697 * 1698 * A call to this helper is susceptible to change the underlying 1699 * packet buffer. Therefore, at load time, all checks on pointers 1700 * previously done by the verifier are invalidated and must be 1701 * performed again, if the helper is used in combination with 1702 * direct packet access. 1703 * Return 1704 * 0 on success, or a negative error in case of failure. 1705 * 1706 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1707 * Description 1708 * Read the value of a perf event counter, and store it into *buf* 1709 * of size *buf_size*. This helper relies on a *map* of type 1710 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1711 * counter is selected when *map* is updated with perf event file 1712 * descriptors. The *map* is an array whose size is the number of 1713 * available CPUs, and each cell contains a value relative to one 1714 * CPU. The value to retrieve is indicated by *flags*, that 1715 * contains the index of the CPU to look up, masked with 1716 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1717 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1718 * current CPU should be retrieved. 1719 * 1720 * This helper behaves in a way close to 1721 * **bpf_perf_event_read**\ () helper, save that instead of 1722 * just returning the value observed, it fills the *buf* 1723 * structure. This allows for additional data to be retrieved: in 1724 * particular, the enabled and running times (in *buf*\ 1725 * **->enabled** and *buf*\ **->running**, respectively) are 1726 * copied. In general, **bpf_perf_event_read_value**\ () is 1727 * recommended over **bpf_perf_event_read**\ (), which has some 1728 * ABI issues and provides fewer functionalities. 1729 * 1730 * These values are interesting, because hardware PMU (Performance 1731 * Monitoring Unit) counters are limited resources. When there are 1732 * more PMU based perf events opened than available counters, 1733 * kernel will multiplex these events so each event gets certain 1734 * percentage (but not all) of the PMU time. In case that 1735 * multiplexing happens, the number of samples or counter value 1736 * will not reflect the case compared to when no multiplexing 1737 * occurs. This makes comparison between different runs difficult. 1738 * Typically, the counter value should be normalized before 1739 * comparing to other experiments. The usual normalization is done 1740 * as follows. 1741 * 1742 * :: 1743 * 1744 * normalized_counter = counter * t_enabled / t_running 1745 * 1746 * Where t_enabled is the time enabled for event and t_running is 1747 * the time running for event since last normalization. The 1748 * enabled and running times are accumulated since the perf event 1749 * open. To achieve scaling factor between two invocations of an 1750 * eBPF program, users can can use CPU id as the key (which is 1751 * typical for perf array usage model) to remember the previous 1752 * value and do the calculation inside the eBPF program. 1753 * Return 1754 * 0 on success, or a negative error in case of failure. 1755 * 1756 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1757 * Description 1758 * For en eBPF program attached to a perf event, retrieve the 1759 * value of the event counter associated to *ctx* and store it in 1760 * the structure pointed by *buf* and of size *buf_size*. Enabled 1761 * and running times are also stored in the structure (see 1762 * description of helper **bpf_perf_event_read_value**\ () for 1763 * more details). 1764 * Return 1765 * 0 on success, or a negative error in case of failure. 1766 * 1767 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1768 * Description 1769 * Emulate a call to **getsockopt()** on the socket associated to 1770 * *bpf_socket*, which must be a full socket. The *level* at 1771 * which the option resides and the name *optname* of the option 1772 * must be specified, see **getsockopt(2)** for more information. 1773 * The retrieved value is stored in the structure pointed by 1774 * *opval* and of length *optlen*. 1775 * 1776 * This helper actually implements a subset of **getsockopt()**. 1777 * It supports the following *level*\ s: 1778 * 1779 * * **IPPROTO_TCP**, which supports *optname* 1780 * **TCP_CONGESTION**. 1781 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1782 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1783 * Return 1784 * 0 on success, or a negative error in case of failure. 1785 * 1786 * int bpf_override_return(struct pt_regs *regs, u64 rc) 1787 * Description 1788 * Used for error injection, this helper uses kprobes to override 1789 * the return value of the probed function, and to set it to *rc*. 1790 * The first argument is the context *regs* on which the kprobe 1791 * works. 1792 * 1793 * This helper works by setting setting the PC (program counter) 1794 * to an override function which is run in place of the original 1795 * probed function. This means the probed function is not run at 1796 * all. The replacement function just returns with the required 1797 * value. 1798 * 1799 * This helper has security implications, and thus is subject to 1800 * restrictions. It is only available if the kernel was compiled 1801 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1802 * option, and in this case it only works on functions tagged with 1803 * **ALLOW_ERROR_INJECTION** in the kernel code. 1804 * 1805 * Also, the helper is only available for the architectures having 1806 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1807 * x86 architecture is the only one to support this feature. 1808 * Return 1809 * 0 1810 * 1811 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1812 * Description 1813 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1814 * for the full TCP socket associated to *bpf_sock_ops* to 1815 * *argval*. 1816 * 1817 * The primary use of this field is to determine if there should 1818 * be calls to eBPF programs of type 1819 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1820 * code. A program of the same type can change its value, per 1821 * connection and as necessary, when the connection is 1822 * established. This field is directly accessible for reading, but 1823 * this helper must be used for updates in order to return an 1824 * error if an eBPF program tries to set a callback that is not 1825 * supported in the current kernel. 1826 * 1827 * *argval* is a flag array which can combine these flags: 1828 * 1829 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1830 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1831 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1832 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1833 * 1834 * Therefore, this function can be used to clear a callback flag by 1835 * setting the appropriate bit to zero. e.g. to disable the RTO 1836 * callback: 1837 * 1838 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1839 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1840 * 1841 * Here are some examples of where one could call such eBPF 1842 * program: 1843 * 1844 * * When RTO fires. 1845 * * When a packet is retransmitted. 1846 * * When the connection terminates. 1847 * * When a packet is sent. 1848 * * When a packet is received. 1849 * Return 1850 * Code **-EINVAL** if the socket is not a full TCP socket; 1851 * otherwise, a positive number containing the bits that could not 1852 * be set is returned (which comes down to 0 if all bits were set 1853 * as required). 1854 * 1855 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1856 * Description 1857 * This helper is used in programs implementing policies at the 1858 * socket level. If the message *msg* is allowed to pass (i.e. if 1859 * the verdict eBPF program returns **SK_PASS**), redirect it to 1860 * the socket referenced by *map* (of type 1861 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1862 * egress interfaces can be used for redirection. The 1863 * **BPF_F_INGRESS** value in *flags* is used to make the 1864 * distinction (ingress path is selected if the flag is present, 1865 * egress path otherwise). This is the only flag supported for now. 1866 * Return 1867 * **SK_PASS** on success, or **SK_DROP** on error. 1868 * 1869 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1870 * Description 1871 * For socket policies, apply the verdict of the eBPF program to 1872 * the next *bytes* (number of bytes) of message *msg*. 1873 * 1874 * For example, this helper can be used in the following cases: 1875 * 1876 * * A single **sendmsg**\ () or **sendfile**\ () system call 1877 * contains multiple logical messages that the eBPF program is 1878 * supposed to read and for which it should apply a verdict. 1879 * * An eBPF program only cares to read the first *bytes* of a 1880 * *msg*. If the message has a large payload, then setting up 1881 * and calling the eBPF program repeatedly for all bytes, even 1882 * though the verdict is already known, would create unnecessary 1883 * overhead. 1884 * 1885 * When called from within an eBPF program, the helper sets a 1886 * counter internal to the BPF infrastructure, that is used to 1887 * apply the last verdict to the next *bytes*. If *bytes* is 1888 * smaller than the current data being processed from a 1889 * **sendmsg**\ () or **sendfile**\ () system call, the first 1890 * *bytes* will be sent and the eBPF program will be re-run with 1891 * the pointer for start of data pointing to byte number *bytes* 1892 * **+ 1**. If *bytes* is larger than the current data being 1893 * processed, then the eBPF verdict will be applied to multiple 1894 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1895 * consumed. 1896 * 1897 * Note that if a socket closes with the internal counter holding 1898 * a non-zero value, this is not a problem because data is not 1899 * being buffered for *bytes* and is sent as it is received. 1900 * Return 1901 * 0 1902 * 1903 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1904 * Description 1905 * For socket policies, prevent the execution of the verdict eBPF 1906 * program for message *msg* until *bytes* (byte number) have been 1907 * accumulated. 1908 * 1909 * This can be used when one needs a specific number of bytes 1910 * before a verdict can be assigned, even if the data spans 1911 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1912 * case would be a user calling **sendmsg**\ () repeatedly with 1913 * 1-byte long message segments. Obviously, this is bad for 1914 * performance, but it is still valid. If the eBPF program needs 1915 * *bytes* bytes to validate a header, this helper can be used to 1916 * prevent the eBPF program to be called again until *bytes* have 1917 * been accumulated. 1918 * Return 1919 * 0 1920 * 1921 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1922 * Description 1923 * For socket policies, pull in non-linear data from user space 1924 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1925 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1926 * respectively. 1927 * 1928 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1929 * *msg* it can only parse data that the (**data**, **data_end**) 1930 * pointers have already consumed. For **sendmsg**\ () hooks this 1931 * is likely the first scatterlist element. But for calls relying 1932 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1933 * be the range (**0**, **0**) because the data is shared with 1934 * user space and by default the objective is to avoid allowing 1935 * user space to modify data while (or after) eBPF verdict is 1936 * being decided. This helper can be used to pull in data and to 1937 * set the start and end pointer to given values. Data will be 1938 * copied if necessary (i.e. if data was not linear and if start 1939 * and end pointers do not point to the same chunk). 1940 * 1941 * A call to this helper is susceptible to change the underlying 1942 * packet buffer. Therefore, at load time, all checks on pointers 1943 * previously done by the verifier are invalidated and must be 1944 * performed again, if the helper is used in combination with 1945 * direct packet access. 1946 * 1947 * All values for *flags* are reserved for future usage, and must 1948 * be left at zero. 1949 * Return 1950 * 0 on success, or a negative error in case of failure. 1951 * 1952 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1953 * Description 1954 * Bind the socket associated to *ctx* to the address pointed by 1955 * *addr*, of length *addr_len*. This allows for making outgoing 1956 * connection from the desired IP address, which can be useful for 1957 * example when all processes inside a cgroup should use one 1958 * single IP address on a host that has multiple IP configured. 1959 * 1960 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1961 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1962 * **AF_INET6**). Looking for a free port to bind to can be 1963 * expensive, therefore binding to port is not permitted by the 1964 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1965 * must be set to zero. 1966 * Return 1967 * 0 on success, or a negative error in case of failure. 1968 * 1969 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1970 * Description 1971 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1972 * only possible to shrink the packet as of this writing, 1973 * therefore *delta* must be a negative integer. 1974 * 1975 * A call to this helper is susceptible to change the underlying 1976 * packet buffer. Therefore, at load time, all checks on pointers 1977 * previously done by the verifier are invalidated and must be 1978 * performed again, if the helper is used in combination with 1979 * direct packet access. 1980 * Return 1981 * 0 on success, or a negative error in case of failure. 1982 * 1983 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1984 * Description 1985 * Retrieve the XFRM state (IP transform framework, see also 1986 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1987 * 1988 * The retrieved value is stored in the **struct bpf_xfrm_state** 1989 * pointed by *xfrm_state* and of length *size*. 1990 * 1991 * All values for *flags* are reserved for future usage, and must 1992 * be left at zero. 1993 * 1994 * This helper is available only if the kernel was compiled with 1995 * **CONFIG_XFRM** configuration option. 1996 * Return 1997 * 0 on success, or a negative error in case of failure. 1998 * 1999 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2000 * Description 2001 * Return a user or a kernel stack in bpf program provided buffer. 2002 * To achieve this, the helper needs *ctx*, which is a pointer 2003 * to the context on which the tracing program is executed. 2004 * To store the stacktrace, the bpf program provides *buf* with 2005 * a nonnegative *size*. 2006 * 2007 * The last argument, *flags*, holds the number of stack frames to 2008 * skip (from 0 to 255), masked with 2009 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2010 * the following flags: 2011 * 2012 * **BPF_F_USER_STACK** 2013 * Collect a user space stack instead of a kernel stack. 2014 * **BPF_F_USER_BUILD_ID** 2015 * Collect buildid+offset instead of ips for user stack, 2016 * only valid if **BPF_F_USER_STACK** is also specified. 2017 * 2018 * **bpf_get_stack**\ () can collect up to 2019 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2020 * to sufficient large buffer size. Note that 2021 * this limit can be controlled with the **sysctl** program, and 2022 * that it should be manually increased in order to profile long 2023 * user stacks (such as stacks for Java programs). To do so, use: 2024 * 2025 * :: 2026 * 2027 * # sysctl kernel.perf_event_max_stack=<new value> 2028 * Return 2029 * A non-negative value equal to or less than *size* on success, 2030 * or a negative error in case of failure. 2031 * 2032 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2033 * Description 2034 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2035 * it provides an easy way to load *len* bytes from *offset* 2036 * from the packet associated to *skb*, into the buffer pointed 2037 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2038 * a fifth argument *start_header* exists in order to select a 2039 * base offset to start from. *start_header* can be one of: 2040 * 2041 * **BPF_HDR_START_MAC** 2042 * Base offset to load data from is *skb*'s mac header. 2043 * **BPF_HDR_START_NET** 2044 * Base offset to load data from is *skb*'s network header. 2045 * 2046 * In general, "direct packet access" is the preferred method to 2047 * access packet data, however, this helper is in particular useful 2048 * in socket filters where *skb*\ **->data** does not always point 2049 * to the start of the mac header and where "direct packet access" 2050 * is not available. 2051 * Return 2052 * 0 on success, or a negative error in case of failure. 2053 * 2054 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2055 * Description 2056 * Do FIB lookup in kernel tables using parameters in *params*. 2057 * If lookup is successful and result shows packet is to be 2058 * forwarded, the neighbor tables are searched for the nexthop. 2059 * If successful (ie., FIB lookup shows forwarding and nexthop 2060 * is resolved), the nexthop address is returned in ipv4_dst 2061 * or ipv6_dst based on family, smac is set to mac address of 2062 * egress device, dmac is set to nexthop mac address, rt_metric 2063 * is set to metric from route (IPv4/IPv6 only), and ifindex 2064 * is set to the device index of the nexthop from the FIB lookup. 2065 * 2066 * *plen* argument is the size of the passed in struct. 2067 * *flags* argument can be a combination of one or more of the 2068 * following values: 2069 * 2070 * **BPF_FIB_LOOKUP_DIRECT** 2071 * Do a direct table lookup vs full lookup using FIB 2072 * rules. 2073 * **BPF_FIB_LOOKUP_OUTPUT** 2074 * Perform lookup from an egress perspective (default is 2075 * ingress). 2076 * 2077 * *ctx* is either **struct xdp_md** for XDP programs or 2078 * **struct sk_buff** tc cls_act programs. 2079 * Return 2080 * * < 0 if any input argument is invalid 2081 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2082 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2083 * packet is not forwarded or needs assist from full stack 2084 * 2085 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2086 * Description 2087 * Add an entry to, or update a sockhash *map* referencing sockets. 2088 * The *skops* is used as a new value for the entry associated to 2089 * *key*. *flags* is one of: 2090 * 2091 * **BPF_NOEXIST** 2092 * The entry for *key* must not exist in the map. 2093 * **BPF_EXIST** 2094 * The entry for *key* must already exist in the map. 2095 * **BPF_ANY** 2096 * No condition on the existence of the entry for *key*. 2097 * 2098 * If the *map* has eBPF programs (parser and verdict), those will 2099 * be inherited by the socket being added. If the socket is 2100 * already attached to eBPF programs, this results in an error. 2101 * Return 2102 * 0 on success, or a negative error in case of failure. 2103 * 2104 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2105 * Description 2106 * This helper is used in programs implementing policies at the 2107 * socket level. If the message *msg* is allowed to pass (i.e. if 2108 * the verdict eBPF program returns **SK_PASS**), redirect it to 2109 * the socket referenced by *map* (of type 2110 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2111 * egress interfaces can be used for redirection. The 2112 * **BPF_F_INGRESS** value in *flags* is used to make the 2113 * distinction (ingress path is selected if the flag is present, 2114 * egress path otherwise). This is the only flag supported for now. 2115 * Return 2116 * **SK_PASS** on success, or **SK_DROP** on error. 2117 * 2118 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2119 * Description 2120 * This helper is used in programs implementing policies at the 2121 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2122 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2123 * to the socket referenced by *map* (of type 2124 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2125 * egress interfaces can be used for redirection. The 2126 * **BPF_F_INGRESS** value in *flags* is used to make the 2127 * distinction (ingress path is selected if the flag is present, 2128 * egress otherwise). This is the only flag supported for now. 2129 * Return 2130 * **SK_PASS** on success, or **SK_DROP** on error. 2131 * 2132 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2133 * Description 2134 * Encapsulate the packet associated to *skb* within a Layer 3 2135 * protocol header. This header is provided in the buffer at 2136 * address *hdr*, with *len* its size in bytes. *type* indicates 2137 * the protocol of the header and can be one of: 2138 * 2139 * **BPF_LWT_ENCAP_SEG6** 2140 * IPv6 encapsulation with Segment Routing Header 2141 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2142 * the IPv6 header is computed by the kernel. 2143 * **BPF_LWT_ENCAP_SEG6_INLINE** 2144 * Only works if *skb* contains an IPv6 packet. Insert a 2145 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2146 * the IPv6 header. 2147 * **BPF_LWT_ENCAP_IP** 2148 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2149 * must be IPv4 or IPv6, followed by zero or more 2150 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2151 * total bytes in all prepended headers. Please note that 2152 * if **skb_is_gso**\ (*skb*) is true, no more than two 2153 * headers can be prepended, and the inner header, if 2154 * present, should be either GRE or UDP/GUE. 2155 * 2156 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2157 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2158 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2159 * **BPF_PROG_TYPE_LWT_XMIT**. 2160 * 2161 * A call to this helper is susceptible to change the underlying 2162 * packet buffer. Therefore, at load time, all checks on pointers 2163 * previously done by the verifier are invalidated and must be 2164 * performed again, if the helper is used in combination with 2165 * direct packet access. 2166 * Return 2167 * 0 on success, or a negative error in case of failure. 2168 * 2169 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2170 * Description 2171 * Store *len* bytes from address *from* into the packet 2172 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2173 * inside the outermost IPv6 Segment Routing Header can be 2174 * modified through this helper. 2175 * 2176 * A call to this helper is susceptible to change the underlying 2177 * packet buffer. Therefore, at load time, all checks on pointers 2178 * previously done by the verifier are invalidated and must be 2179 * performed again, if the helper is used in combination with 2180 * direct packet access. 2181 * Return 2182 * 0 on success, or a negative error in case of failure. 2183 * 2184 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2185 * Description 2186 * Adjust the size allocated to TLVs in the outermost IPv6 2187 * Segment Routing Header contained in the packet associated to 2188 * *skb*, at position *offset* by *delta* bytes. Only offsets 2189 * after the segments are accepted. *delta* can be as well 2190 * positive (growing) as negative (shrinking). 2191 * 2192 * A call to this helper is susceptible to change the underlying 2193 * packet buffer. Therefore, at load time, all checks on pointers 2194 * previously done by the verifier are invalidated and must be 2195 * performed again, if the helper is used in combination with 2196 * direct packet access. 2197 * Return 2198 * 0 on success, or a negative error in case of failure. 2199 * 2200 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2201 * Description 2202 * Apply an IPv6 Segment Routing action of type *action* to the 2203 * packet associated to *skb*. Each action takes a parameter 2204 * contained at address *param*, and of length *param_len* bytes. 2205 * *action* can be one of: 2206 * 2207 * **SEG6_LOCAL_ACTION_END_X** 2208 * End.X action: Endpoint with Layer-3 cross-connect. 2209 * Type of *param*: **struct in6_addr**. 2210 * **SEG6_LOCAL_ACTION_END_T** 2211 * End.T action: Endpoint with specific IPv6 table lookup. 2212 * Type of *param*: **int**. 2213 * **SEG6_LOCAL_ACTION_END_B6** 2214 * End.B6 action: Endpoint bound to an SRv6 policy. 2215 * Type of *param*: **struct ipv6_sr_hdr**. 2216 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2217 * End.B6.Encap action: Endpoint bound to an SRv6 2218 * encapsulation policy. 2219 * Type of *param*: **struct ipv6_sr_hdr**. 2220 * 2221 * A call to this helper is susceptible to change the underlying 2222 * packet buffer. Therefore, at load time, all checks on pointers 2223 * previously done by the verifier are invalidated and must be 2224 * performed again, if the helper is used in combination with 2225 * direct packet access. 2226 * Return 2227 * 0 on success, or a negative error in case of failure. 2228 * 2229 * int bpf_rc_repeat(void *ctx) 2230 * Description 2231 * This helper is used in programs implementing IR decoding, to 2232 * report a successfully decoded repeat key message. This delays 2233 * the generation of a key up event for previously generated 2234 * key down event. 2235 * 2236 * Some IR protocols like NEC have a special IR message for 2237 * repeating last button, for when a button is held down. 2238 * 2239 * The *ctx* should point to the lirc sample as passed into 2240 * the program. 2241 * 2242 * This helper is only available is the kernel was compiled with 2243 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2244 * "**y**". 2245 * Return 2246 * 0 2247 * 2248 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2249 * Description 2250 * This helper is used in programs implementing IR decoding, to 2251 * report a successfully decoded key press with *scancode*, 2252 * *toggle* value in the given *protocol*. The scancode will be 2253 * translated to a keycode using the rc keymap, and reported as 2254 * an input key down event. After a period a key up event is 2255 * generated. This period can be extended by calling either 2256 * **bpf_rc_keydown**\ () again with the same values, or calling 2257 * **bpf_rc_repeat**\ (). 2258 * 2259 * Some protocols include a toggle bit, in case the button was 2260 * released and pressed again between consecutive scancodes. 2261 * 2262 * The *ctx* should point to the lirc sample as passed into 2263 * the program. 2264 * 2265 * The *protocol* is the decoded protocol number (see 2266 * **enum rc_proto** for some predefined values). 2267 * 2268 * This helper is only available is the kernel was compiled with 2269 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2270 * "**y**". 2271 * Return 2272 * 0 2273 * 2274 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2275 * Description 2276 * Return the cgroup v2 id of the socket associated with the *skb*. 2277 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2278 * helper for cgroup v1 by providing a tag resp. identifier that 2279 * can be matched on or used for map lookups e.g. to implement 2280 * policy. The cgroup v2 id of a given path in the hierarchy is 2281 * exposed in user space through the f_handle API in order to get 2282 * to the same 64-bit id. 2283 * 2284 * This helper can be used on TC egress path, but not on ingress, 2285 * and is available only if the kernel was compiled with the 2286 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2287 * Return 2288 * The id is returned or 0 in case the id could not be retrieved. 2289 * 2290 * u64 bpf_get_current_cgroup_id(void) 2291 * Return 2292 * A 64-bit integer containing the current cgroup id based 2293 * on the cgroup within which the current task is running. 2294 * 2295 * void *bpf_get_local_storage(void *map, u64 flags) 2296 * Description 2297 * Get the pointer to the local storage area. 2298 * The type and the size of the local storage is defined 2299 * by the *map* argument. 2300 * The *flags* meaning is specific for each map type, 2301 * and has to be 0 for cgroup local storage. 2302 * 2303 * Depending on the BPF program type, a local storage area 2304 * can be shared between multiple instances of the BPF program, 2305 * running simultaneously. 2306 * 2307 * A user should care about the synchronization by himself. 2308 * For example, by using the **BPF_STX_XADD** instruction to alter 2309 * the shared data. 2310 * Return 2311 * A pointer to the local storage area. 2312 * 2313 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2314 * Description 2315 * Select a **SO_REUSEPORT** socket from a 2316 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2317 * It checks the selected socket is matching the incoming 2318 * request in the socket buffer. 2319 * Return 2320 * 0 on success, or a negative error in case of failure. 2321 * 2322 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2323 * Description 2324 * Return id of cgroup v2 that is ancestor of cgroup associated 2325 * with the *skb* at the *ancestor_level*. The root cgroup is at 2326 * *ancestor_level* zero and each step down the hierarchy 2327 * increments the level. If *ancestor_level* == level of cgroup 2328 * associated with *skb*, then return value will be same as that 2329 * of **bpf_skb_cgroup_id**\ (). 2330 * 2331 * The helper is useful to implement policies based on cgroups 2332 * that are upper in hierarchy than immediate cgroup associated 2333 * with *skb*. 2334 * 2335 * The format of returned id and helper limitations are same as in 2336 * **bpf_skb_cgroup_id**\ (). 2337 * Return 2338 * The id is returned or 0 in case the id could not be retrieved. 2339 * 2340 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2341 * Description 2342 * Look for TCP socket matching *tuple*, optionally in a child 2343 * network namespace *netns*. The return value must be checked, 2344 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2345 * 2346 * The *ctx* should point to the context of the program, such as 2347 * the skb or socket (depending on the hook in use). This is used 2348 * to determine the base network namespace for the lookup. 2349 * 2350 * *tuple_size* must be one of: 2351 * 2352 * **sizeof**\ (*tuple*\ **->ipv4**) 2353 * Look for an IPv4 socket. 2354 * **sizeof**\ (*tuple*\ **->ipv6**) 2355 * Look for an IPv6 socket. 2356 * 2357 * If the *netns* is a negative signed 32-bit integer, then the 2358 * socket lookup table in the netns associated with the *ctx* will 2359 * will be used. For the TC hooks, this is the netns of the device 2360 * in the skb. For socket hooks, this is the netns of the socket. 2361 * If *netns* is any other signed 32-bit value greater than or 2362 * equal to zero then it specifies the ID of the netns relative to 2363 * the netns associated with the *ctx*. *netns* values beyond the 2364 * range of 32-bit integers are reserved for future use. 2365 * 2366 * All values for *flags* are reserved for future usage, and must 2367 * be left at zero. 2368 * 2369 * This helper is available only if the kernel was compiled with 2370 * **CONFIG_NET** configuration option. 2371 * Return 2372 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2373 * For sockets with reuseport option, the **struct bpf_sock** 2374 * result is from *reuse*\ **->socks**\ [] using the hash of the 2375 * tuple. 2376 * 2377 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2378 * Description 2379 * Look for UDP socket matching *tuple*, optionally in a child 2380 * network namespace *netns*. The return value must be checked, 2381 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2382 * 2383 * The *ctx* should point to the context of the program, such as 2384 * the skb or socket (depending on the hook in use). This is used 2385 * to determine the base network namespace for the lookup. 2386 * 2387 * *tuple_size* must be one of: 2388 * 2389 * **sizeof**\ (*tuple*\ **->ipv4**) 2390 * Look for an IPv4 socket. 2391 * **sizeof**\ (*tuple*\ **->ipv6**) 2392 * Look for an IPv6 socket. 2393 * 2394 * If the *netns* is a negative signed 32-bit integer, then the 2395 * socket lookup table in the netns associated with the *ctx* will 2396 * will be used. For the TC hooks, this is the netns of the device 2397 * in the skb. For socket hooks, this is the netns of the socket. 2398 * If *netns* is any other signed 32-bit value greater than or 2399 * equal to zero then it specifies the ID of the netns relative to 2400 * the netns associated with the *ctx*. *netns* values beyond the 2401 * range of 32-bit integers are reserved for future use. 2402 * 2403 * All values for *flags* are reserved for future usage, and must 2404 * be left at zero. 2405 * 2406 * This helper is available only if the kernel was compiled with 2407 * **CONFIG_NET** configuration option. 2408 * Return 2409 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2410 * For sockets with reuseport option, the **struct bpf_sock** 2411 * result is from *reuse*\ **->socks**\ [] using the hash of the 2412 * tuple. 2413 * 2414 * int bpf_sk_release(struct bpf_sock *sock) 2415 * Description 2416 * Release the reference held by *sock*. *sock* must be a 2417 * non-**NULL** pointer that was returned from 2418 * **bpf_sk_lookup_xxx**\ (). 2419 * Return 2420 * 0 on success, or a negative error in case of failure. 2421 * 2422 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2423 * Description 2424 * Push an element *value* in *map*. *flags* is one of: 2425 * 2426 * **BPF_EXIST** 2427 * If the queue/stack is full, the oldest element is 2428 * removed to make room for this. 2429 * Return 2430 * 0 on success, or a negative error in case of failure. 2431 * 2432 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2433 * Description 2434 * Pop an element from *map*. 2435 * Return 2436 * 0 on success, or a negative error in case of failure. 2437 * 2438 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2439 * Description 2440 * Get an element from *map* without removing it. 2441 * Return 2442 * 0 on success, or a negative error in case of failure. 2443 * 2444 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2445 * Description 2446 * For socket policies, insert *len* bytes into *msg* at offset 2447 * *start*. 2448 * 2449 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2450 * *msg* it may want to insert metadata or options into the *msg*. 2451 * This can later be read and used by any of the lower layer BPF 2452 * hooks. 2453 * 2454 * This helper may fail if under memory pressure (a malloc 2455 * fails) in these cases BPF programs will get an appropriate 2456 * error and BPF programs will need to handle them. 2457 * Return 2458 * 0 on success, or a negative error in case of failure. 2459 * 2460 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2461 * Description 2462 * Will remove *len* bytes from a *msg* starting at byte *start*. 2463 * This may result in **ENOMEM** errors under certain situations if 2464 * an allocation and copy are required due to a full ring buffer. 2465 * However, the helper will try to avoid doing the allocation 2466 * if possible. Other errors can occur if input parameters are 2467 * invalid either due to *start* byte not being valid part of *msg* 2468 * payload and/or *pop* value being to large. 2469 * Return 2470 * 0 on success, or a negative error in case of failure. 2471 * 2472 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2473 * Description 2474 * This helper is used in programs implementing IR decoding, to 2475 * report a successfully decoded pointer movement. 2476 * 2477 * The *ctx* should point to the lirc sample as passed into 2478 * the program. 2479 * 2480 * This helper is only available is the kernel was compiled with 2481 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2482 * "**y**". 2483 * Return 2484 * 0 2485 * 2486 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2487 * Description 2488 * Acquire a spinlock represented by the pointer *lock*, which is 2489 * stored as part of a value of a map. Taking the lock allows to 2490 * safely update the rest of the fields in that value. The 2491 * spinlock can (and must) later be released with a call to 2492 * **bpf_spin_unlock**\ (\ *lock*\ ). 2493 * 2494 * Spinlocks in BPF programs come with a number of restrictions 2495 * and constraints: 2496 * 2497 * * **bpf_spin_lock** objects are only allowed inside maps of 2498 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2499 * list could be extended in the future). 2500 * * BTF description of the map is mandatory. 2501 * * The BPF program can take ONE lock at a time, since taking two 2502 * or more could cause dead locks. 2503 * * Only one **struct bpf_spin_lock** is allowed per map element. 2504 * * When the lock is taken, calls (either BPF to BPF or helpers) 2505 * are not allowed. 2506 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2507 * allowed inside a spinlock-ed region. 2508 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2509 * the lock, on all execution paths, before it returns. 2510 * * The BPF program can access **struct bpf_spin_lock** only via 2511 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2512 * helpers. Loading or storing data into the **struct 2513 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2514 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2515 * of the map value must be a struct and have **struct 2516 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2517 * Nested lock inside another struct is not allowed. 2518 * * The **struct bpf_spin_lock** *lock* field in a map value must 2519 * be aligned on a multiple of 4 bytes in that value. 2520 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2521 * the **bpf_spin_lock** field to user space. 2522 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2523 * a BPF program, do not update the **bpf_spin_lock** field. 2524 * * **bpf_spin_lock** cannot be on the stack or inside a 2525 * networking packet (it can only be inside of a map values). 2526 * * **bpf_spin_lock** is available to root only. 2527 * * Tracing programs and socket filter programs cannot use 2528 * **bpf_spin_lock**\ () due to insufficient preemption checks 2529 * (but this may change in the future). 2530 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2531 * Return 2532 * 0 2533 * 2534 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2535 * Description 2536 * Release the *lock* previously locked by a call to 2537 * **bpf_spin_lock**\ (\ *lock*\ ). 2538 * Return 2539 * 0 2540 * 2541 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2542 * Description 2543 * This helper gets a **struct bpf_sock** pointer such 2544 * that all the fields in this **bpf_sock** can be accessed. 2545 * Return 2546 * A **struct bpf_sock** pointer on success, or **NULL** in 2547 * case of failure. 2548 * 2549 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2550 * Description 2551 * This helper gets a **struct bpf_tcp_sock** pointer from a 2552 * **struct bpf_sock** pointer. 2553 * Return 2554 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2555 * case of failure. 2556 * 2557 * int bpf_skb_ecn_set_ce(struct sk_buff *skb) 2558 * Description 2559 * Set ECN (Explicit Congestion Notification) field of IP header 2560 * to **CE** (Congestion Encountered) if current value is **ECT** 2561 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2562 * and IPv4. 2563 * Return 2564 * 1 if the **CE** flag is set (either by the current helper call 2565 * or because it was already present), 0 if it is not set. 2566 * 2567 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2568 * Description 2569 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2570 * **bpf_sk_release**\ () is unnecessary and not allowed. 2571 * Return 2572 * A **struct bpf_sock** pointer on success, or **NULL** in 2573 * case of failure. 2574 * 2575 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2576 * Description 2577 * Look for TCP socket matching *tuple*, optionally in a child 2578 * network namespace *netns*. The return value must be checked, 2579 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2580 * 2581 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2582 * that it also returns timewait or request sockets. Use 2583 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2584 * full structure. 2585 * 2586 * This helper is available only if the kernel was compiled with 2587 * **CONFIG_NET** configuration option. 2588 * Return 2589 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2590 * For sockets with reuseport option, the **struct bpf_sock** 2591 * result is from *reuse*\ **->socks**\ [] using the hash of the 2592 * tuple. 2593 * 2594 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2595 * Description 2596 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2597 * the listening socket in *sk*. 2598 * 2599 * *iph* points to the start of the IPv4 or IPv6 header, while 2600 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2601 * **sizeof**\ (**struct ip6hdr**). 2602 * 2603 * *th* points to the start of the TCP header, while *th_len* 2604 * contains **sizeof**\ (**struct tcphdr**). 2605 * 2606 * Return 2607 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2608 * error otherwise. 2609 * 2610 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2611 * Description 2612 * Get name of sysctl in /proc/sys/ and copy it into provided by 2613 * program buffer *buf* of size *buf_len*. 2614 * 2615 * The buffer is always NUL terminated, unless it's zero-sized. 2616 * 2617 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2618 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2619 * only (e.g. "tcp_mem"). 2620 * Return 2621 * Number of character copied (not including the trailing NUL). 2622 * 2623 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2624 * truncated name in this case). 2625 * 2626 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2627 * Description 2628 * Get current value of sysctl as it is presented in /proc/sys 2629 * (incl. newline, etc), and copy it as a string into provided 2630 * by program buffer *buf* of size *buf_len*. 2631 * 2632 * The whole value is copied, no matter what file position user 2633 * space issued e.g. sys_read at. 2634 * 2635 * The buffer is always NUL terminated, unless it's zero-sized. 2636 * Return 2637 * Number of character copied (not including the trailing NUL). 2638 * 2639 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2640 * truncated name in this case). 2641 * 2642 * **-EINVAL** if current value was unavailable, e.g. because 2643 * sysctl is uninitialized and read returns -EIO for it. 2644 * 2645 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2646 * Description 2647 * Get new value being written by user space to sysctl (before 2648 * the actual write happens) and copy it as a string into 2649 * provided by program buffer *buf* of size *buf_len*. 2650 * 2651 * User space may write new value at file position > 0. 2652 * 2653 * The buffer is always NUL terminated, unless it's zero-sized. 2654 * Return 2655 * Number of character copied (not including the trailing NUL). 2656 * 2657 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2658 * truncated name in this case). 2659 * 2660 * **-EINVAL** if sysctl is being read. 2661 * 2662 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2663 * Description 2664 * Override new value being written by user space to sysctl with 2665 * value provided by program in buffer *buf* of size *buf_len*. 2666 * 2667 * *buf* should contain a string in same form as provided by user 2668 * space on sysctl write. 2669 * 2670 * User space may write new value at file position > 0. To override 2671 * the whole sysctl value file position should be set to zero. 2672 * Return 2673 * 0 on success. 2674 * 2675 * **-E2BIG** if the *buf_len* is too big. 2676 * 2677 * **-EINVAL** if sysctl is being read. 2678 * 2679 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2680 * Description 2681 * Convert the initial part of the string from buffer *buf* of 2682 * size *buf_len* to a long integer according to the given base 2683 * and save the result in *res*. 2684 * 2685 * The string may begin with an arbitrary amount of white space 2686 * (as determined by **isspace**\ (3)) followed by a single 2687 * optional '**-**' sign. 2688 * 2689 * Five least significant bits of *flags* encode base, other bits 2690 * are currently unused. 2691 * 2692 * Base must be either 8, 10, 16 or 0 to detect it automatically 2693 * similar to user space **strtol**\ (3). 2694 * Return 2695 * Number of characters consumed on success. Must be positive but 2696 * no more than *buf_len*. 2697 * 2698 * **-EINVAL** if no valid digits were found or unsupported base 2699 * was provided. 2700 * 2701 * **-ERANGE** if resulting value was out of range. 2702 * 2703 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2704 * Description 2705 * Convert the initial part of the string from buffer *buf* of 2706 * size *buf_len* to an unsigned long integer according to the 2707 * given base and save the result in *res*. 2708 * 2709 * The string may begin with an arbitrary amount of white space 2710 * (as determined by **isspace**\ (3)). 2711 * 2712 * Five least significant bits of *flags* encode base, other bits 2713 * are currently unused. 2714 * 2715 * Base must be either 8, 10, 16 or 0 to detect it automatically 2716 * similar to user space **strtoul**\ (3). 2717 * Return 2718 * Number of characters consumed on success. Must be positive but 2719 * no more than *buf_len*. 2720 * 2721 * **-EINVAL** if no valid digits were found or unsupported base 2722 * was provided. 2723 * 2724 * **-ERANGE** if resulting value was out of range. 2725 * 2726 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2727 * Description 2728 * Get a bpf-local-storage from a *sk*. 2729 * 2730 * Logically, it could be thought of getting the value from 2731 * a *map* with *sk* as the **key**. From this 2732 * perspective, the usage is not much different from 2733 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2734 * helper enforces the key must be a full socket and the map must 2735 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2736 * 2737 * Underneath, the value is stored locally at *sk* instead of 2738 * the *map*. The *map* is used as the bpf-local-storage 2739 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2740 * searched against all bpf-local-storages residing at *sk*. 2741 * 2742 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2743 * used such that a new bpf-local-storage will be 2744 * created if one does not exist. *value* can be used 2745 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2746 * the initial value of a bpf-local-storage. If *value* is 2747 * **NULL**, the new bpf-local-storage will be zero initialized. 2748 * Return 2749 * A bpf-local-storage pointer is returned on success. 2750 * 2751 * **NULL** if not found or there was an error in adding 2752 * a new bpf-local-storage. 2753 * 2754 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2755 * Description 2756 * Delete a bpf-local-storage from a *sk*. 2757 * Return 2758 * 0 on success. 2759 * 2760 * **-ENOENT** if the bpf-local-storage cannot be found. 2761 * 2762 * int bpf_send_signal(u32 sig) 2763 * Description 2764 * Send signal *sig* to the process of the current task. 2765 * The signal may be delivered to any of this process's threads. 2766 * Return 2767 * 0 on success or successfully queued. 2768 * 2769 * **-EBUSY** if work queue under nmi is full. 2770 * 2771 * **-EINVAL** if *sig* is invalid. 2772 * 2773 * **-EPERM** if no permission to send the *sig*. 2774 * 2775 * **-EAGAIN** if bpf program can try again. 2776 * 2777 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2778 * Description 2779 * Try to issue a SYN cookie for the packet with corresponding 2780 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2781 * 2782 * *iph* points to the start of the IPv4 or IPv6 header, while 2783 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2784 * **sizeof**\ (**struct ip6hdr**). 2785 * 2786 * *th* points to the start of the TCP header, while *th_len* 2787 * contains the length of the TCP header. 2788 * 2789 * Return 2790 * On success, lower 32 bits hold the generated SYN cookie in 2791 * followed by 16 bits which hold the MSS value for that cookie, 2792 * and the top 16 bits are unused. 2793 * 2794 * On failure, the returned value is one of the following: 2795 * 2796 * **-EINVAL** SYN cookie cannot be issued due to error 2797 * 2798 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2799 * 2800 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2801 * 2802 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2803 * 2804 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2805 * Description 2806 * Write raw *data* blob into a special BPF perf event held by 2807 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2808 * event must have the following attributes: **PERF_SAMPLE_RAW** 2809 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2810 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2811 * 2812 * The *flags* are used to indicate the index in *map* for which 2813 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2814 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2815 * to indicate that the index of the current CPU core should be 2816 * used. 2817 * 2818 * The value to write, of *size*, is passed through eBPF stack and 2819 * pointed by *data*. 2820 * 2821 * *ctx* is a pointer to in-kernel struct sk_buff. 2822 * 2823 * This helper is similar to **bpf_perf_event_output**\ () but 2824 * restricted to raw_tracepoint bpf programs. 2825 * Return 2826 * 0 on success, or a negative error in case of failure. 2827 * 2828 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2829 * Description 2830 * Safely attempt to read *size* bytes from user space address 2831 * *unsafe_ptr* and store the data in *dst*. 2832 * Return 2833 * 0 on success, or a negative error in case of failure. 2834 * 2835 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2836 * Description 2837 * Safely attempt to read *size* bytes from kernel space address 2838 * *unsafe_ptr* and store the data in *dst*. 2839 * Return 2840 * 0 on success, or a negative error in case of failure. 2841 * 2842 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2843 * Description 2844 * Copy a NUL terminated string from an unsafe user address 2845 * *unsafe_ptr* to *dst*. The *size* should include the 2846 * terminating NUL byte. In case the string length is smaller than 2847 * *size*, the target is not padded with further NUL bytes. If the 2848 * string length is larger than *size*, just *size*-1 bytes are 2849 * copied and the last byte is set to NUL. 2850 * 2851 * On success, the length of the copied string is returned. This 2852 * makes this helper useful in tracing programs for reading 2853 * strings, and more importantly to get its length at runtime. See 2854 * the following snippet: 2855 * 2856 * :: 2857 * 2858 * SEC("kprobe/sys_open") 2859 * void bpf_sys_open(struct pt_regs *ctx) 2860 * { 2861 * char buf[PATHLEN]; // PATHLEN is defined to 256 2862 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2863 * ctx->di); 2864 * 2865 * // Consume buf, for example push it to 2866 * // userspace via bpf_perf_event_output(); we 2867 * // can use res (the string length) as event 2868 * // size, after checking its boundaries. 2869 * } 2870 * 2871 * In comparison, using **bpf_probe_read_user()** helper here 2872 * instead to read the string would require to estimate the length 2873 * at compile time, and would often result in copying more memory 2874 * than necessary. 2875 * 2876 * Another useful use case is when parsing individual process 2877 * arguments or individual environment variables navigating 2878 * *current*\ **->mm->arg_start** and *current*\ 2879 * **->mm->env_start**: using this helper and the return value, 2880 * one can quickly iterate at the right offset of the memory area. 2881 * Return 2882 * On success, the strictly positive length of the string, 2883 * including the trailing NUL character. On error, a negative 2884 * value. 2885 * 2886 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 2887 * Description 2888 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2889 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply. 2890 * Return 2891 * On success, the strictly positive length of the string, including 2892 * the trailing NUL character. On error, a negative value. 2893 * 2894 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 2895 * Description 2896 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock. 2897 * *rcv_nxt* is the ack_seq to be sent out. 2898 * Return 2899 * 0 on success, or a negative error in case of failure. 2900 * 2901 * int bpf_send_signal_thread(u32 sig) 2902 * Description 2903 * Send signal *sig* to the thread corresponding to the current task. 2904 * Return 2905 * 0 on success or successfully queued. 2906 * 2907 * **-EBUSY** if work queue under nmi is full. 2908 * 2909 * **-EINVAL** if *sig* is invalid. 2910 * 2911 * **-EPERM** if no permission to send the *sig*. 2912 * 2913 * **-EAGAIN** if bpf program can try again. 2914 * 2915 * u64 bpf_jiffies64(void) 2916 * Description 2917 * Obtain the 64bit jiffies 2918 * Return 2919 * The 64 bit jiffies 2920 * 2921 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 2922 * Description 2923 * For an eBPF program attached to a perf event, retrieve the 2924 * branch records (struct perf_branch_entry) associated to *ctx* 2925 * and store it in the buffer pointed by *buf* up to size 2926 * *size* bytes. 2927 * Return 2928 * On success, number of bytes written to *buf*. On error, a 2929 * negative value. 2930 * 2931 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 2932 * instead return the number of bytes required to store all the 2933 * branch entries. If this flag is set, *buf* may be NULL. 2934 * 2935 * **-EINVAL** if arguments invalid or **size** not a multiple 2936 * of sizeof(struct perf_branch_entry). 2937 * 2938 * **-ENOENT** if architecture does not support branch records. 2939 * 2940 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 2941 * Description 2942 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 2943 * *namespace* will be returned in *nsdata*. 2944 * 2945 * On failure, the returned value is one of the following: 2946 * 2947 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 2948 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 2949 * 2950 * **-ENOENT** if pidns does not exists for the current task. 2951 * 2952 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2953 * Description 2954 * Write raw *data* blob into a special BPF perf event held by 2955 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2956 * event must have the following attributes: **PERF_SAMPLE_RAW** 2957 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2958 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2959 * 2960 * The *flags* are used to indicate the index in *map* for which 2961 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2962 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2963 * to indicate that the index of the current CPU core should be 2964 * used. 2965 * 2966 * The value to write, of *size*, is passed through eBPF stack and 2967 * pointed by *data*. 2968 * 2969 * *ctx* is a pointer to in-kernel struct xdp_buff. 2970 * 2971 * This helper is similar to **bpf_perf_eventoutput**\ () but 2972 * restricted to raw_tracepoint bpf programs. 2973 * Return 2974 * 0 on success, or a negative error in case of failure. 2975 * 2976 * u64 bpf_get_netns_cookie(void *ctx) 2977 * Description 2978 * Retrieve the cookie (generated by the kernel) of the network 2979 * namespace the input *ctx* is associated with. The network 2980 * namespace cookie remains stable for its lifetime and provides 2981 * a global identifier that can be assumed unique. If *ctx* is 2982 * NULL, then the helper returns the cookie for the initial 2983 * network namespace. The cookie itself is very similar to that 2984 * of bpf_get_socket_cookie() helper, but for network namespaces 2985 * instead of sockets. 2986 * Return 2987 * A 8-byte long opaque number. 2988 * 2989 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 2990 * Description 2991 * Return id of cgroup v2 that is ancestor of the cgroup associated 2992 * with the current task at the *ancestor_level*. The root cgroup 2993 * is at *ancestor_level* zero and each step down the hierarchy 2994 * increments the level. If *ancestor_level* == level of cgroup 2995 * associated with the current task, then return value will be the 2996 * same as that of **bpf_get_current_cgroup_id**\ (). 2997 * 2998 * The helper is useful to implement policies based on cgroups 2999 * that are upper in hierarchy than immediate cgroup associated 3000 * with the current task. 3001 * 3002 * The format of returned id and helper limitations are same as in 3003 * **bpf_get_current_cgroup_id**\ (). 3004 * Return 3005 * The id is returned or 0 in case the id could not be retrieved. 3006 * 3007 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags) 3008 * Description 3009 * Assign the *sk* to the *skb*. When combined with appropriate 3010 * routing configuration to receive the packet towards the socket, 3011 * will cause *skb* to be delivered to the specified socket. 3012 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3013 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3014 * interfere with successful delivery to the socket. 3015 * 3016 * This operation is only valid from TC ingress path. 3017 * 3018 * The *flags* argument must be zero. 3019 * Return 3020 * 0 on success, or a negative errno in case of failure. 3021 * 3022 * * **-EINVAL** Unsupported flags specified. 3023 * * **-ENOENT** Socket is unavailable for assignment. 3024 * * **-ENETUNREACH** Socket is unreachable (wrong netns). 3025 * * **-EOPNOTSUPP** Unsupported operation, for example a 3026 * call from outside of TC ingress. 3027 * * **-ESOCKTNOSUPPORT** Socket type not supported (reuseport). 3028 */ 3029 #define __BPF_FUNC_MAPPER(FN) \ 3030 FN(unspec), \ 3031 FN(map_lookup_elem), \ 3032 FN(map_update_elem), \ 3033 FN(map_delete_elem), \ 3034 FN(probe_read), \ 3035 FN(ktime_get_ns), \ 3036 FN(trace_printk), \ 3037 FN(get_prandom_u32), \ 3038 FN(get_smp_processor_id), \ 3039 FN(skb_store_bytes), \ 3040 FN(l3_csum_replace), \ 3041 FN(l4_csum_replace), \ 3042 FN(tail_call), \ 3043 FN(clone_redirect), \ 3044 FN(get_current_pid_tgid), \ 3045 FN(get_current_uid_gid), \ 3046 FN(get_current_comm), \ 3047 FN(get_cgroup_classid), \ 3048 FN(skb_vlan_push), \ 3049 FN(skb_vlan_pop), \ 3050 FN(skb_get_tunnel_key), \ 3051 FN(skb_set_tunnel_key), \ 3052 FN(perf_event_read), \ 3053 FN(redirect), \ 3054 FN(get_route_realm), \ 3055 FN(perf_event_output), \ 3056 FN(skb_load_bytes), \ 3057 FN(get_stackid), \ 3058 FN(csum_diff), \ 3059 FN(skb_get_tunnel_opt), \ 3060 FN(skb_set_tunnel_opt), \ 3061 FN(skb_change_proto), \ 3062 FN(skb_change_type), \ 3063 FN(skb_under_cgroup), \ 3064 FN(get_hash_recalc), \ 3065 FN(get_current_task), \ 3066 FN(probe_write_user), \ 3067 FN(current_task_under_cgroup), \ 3068 FN(skb_change_tail), \ 3069 FN(skb_pull_data), \ 3070 FN(csum_update), \ 3071 FN(set_hash_invalid), \ 3072 FN(get_numa_node_id), \ 3073 FN(skb_change_head), \ 3074 FN(xdp_adjust_head), \ 3075 FN(probe_read_str), \ 3076 FN(get_socket_cookie), \ 3077 FN(get_socket_uid), \ 3078 FN(set_hash), \ 3079 FN(setsockopt), \ 3080 FN(skb_adjust_room), \ 3081 FN(redirect_map), \ 3082 FN(sk_redirect_map), \ 3083 FN(sock_map_update), \ 3084 FN(xdp_adjust_meta), \ 3085 FN(perf_event_read_value), \ 3086 FN(perf_prog_read_value), \ 3087 FN(getsockopt), \ 3088 FN(override_return), \ 3089 FN(sock_ops_cb_flags_set), \ 3090 FN(msg_redirect_map), \ 3091 FN(msg_apply_bytes), \ 3092 FN(msg_cork_bytes), \ 3093 FN(msg_pull_data), \ 3094 FN(bind), \ 3095 FN(xdp_adjust_tail), \ 3096 FN(skb_get_xfrm_state), \ 3097 FN(get_stack), \ 3098 FN(skb_load_bytes_relative), \ 3099 FN(fib_lookup), \ 3100 FN(sock_hash_update), \ 3101 FN(msg_redirect_hash), \ 3102 FN(sk_redirect_hash), \ 3103 FN(lwt_push_encap), \ 3104 FN(lwt_seg6_store_bytes), \ 3105 FN(lwt_seg6_adjust_srh), \ 3106 FN(lwt_seg6_action), \ 3107 FN(rc_repeat), \ 3108 FN(rc_keydown), \ 3109 FN(skb_cgroup_id), \ 3110 FN(get_current_cgroup_id), \ 3111 FN(get_local_storage), \ 3112 FN(sk_select_reuseport), \ 3113 FN(skb_ancestor_cgroup_id), \ 3114 FN(sk_lookup_tcp), \ 3115 FN(sk_lookup_udp), \ 3116 FN(sk_release), \ 3117 FN(map_push_elem), \ 3118 FN(map_pop_elem), \ 3119 FN(map_peek_elem), \ 3120 FN(msg_push_data), \ 3121 FN(msg_pop_data), \ 3122 FN(rc_pointer_rel), \ 3123 FN(spin_lock), \ 3124 FN(spin_unlock), \ 3125 FN(sk_fullsock), \ 3126 FN(tcp_sock), \ 3127 FN(skb_ecn_set_ce), \ 3128 FN(get_listener_sock), \ 3129 FN(skc_lookup_tcp), \ 3130 FN(tcp_check_syncookie), \ 3131 FN(sysctl_get_name), \ 3132 FN(sysctl_get_current_value), \ 3133 FN(sysctl_get_new_value), \ 3134 FN(sysctl_set_new_value), \ 3135 FN(strtol), \ 3136 FN(strtoul), \ 3137 FN(sk_storage_get), \ 3138 FN(sk_storage_delete), \ 3139 FN(send_signal), \ 3140 FN(tcp_gen_syncookie), \ 3141 FN(skb_output), \ 3142 FN(probe_read_user), \ 3143 FN(probe_read_kernel), \ 3144 FN(probe_read_user_str), \ 3145 FN(probe_read_kernel_str), \ 3146 FN(tcp_send_ack), \ 3147 FN(send_signal_thread), \ 3148 FN(jiffies64), \ 3149 FN(read_branch_records), \ 3150 FN(get_ns_current_pid_tgid), \ 3151 FN(xdp_output), \ 3152 FN(get_netns_cookie), \ 3153 FN(get_current_ancestor_cgroup_id), \ 3154 FN(sk_assign), 3155 3156 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3157 * function eBPF program intends to call 3158 */ 3159 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3160 enum bpf_func_id { 3161 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3162 __BPF_FUNC_MAX_ID, 3163 }; 3164 #undef __BPF_ENUM_FN 3165 3166 /* All flags used by eBPF helper functions, placed here. */ 3167 3168 /* BPF_FUNC_skb_store_bytes flags. */ 3169 enum { 3170 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 3171 BPF_F_INVALIDATE_HASH = (1ULL << 1), 3172 }; 3173 3174 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3175 * First 4 bits are for passing the header field size. 3176 */ 3177 enum { 3178 BPF_F_HDR_FIELD_MASK = 0xfULL, 3179 }; 3180 3181 /* BPF_FUNC_l4_csum_replace flags. */ 3182 enum { 3183 BPF_F_PSEUDO_HDR = (1ULL << 4), 3184 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 3185 BPF_F_MARK_ENFORCE = (1ULL << 6), 3186 }; 3187 3188 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3189 enum { 3190 BPF_F_INGRESS = (1ULL << 0), 3191 }; 3192 3193 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3194 enum { 3195 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 3196 }; 3197 3198 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3199 enum { 3200 BPF_F_SKIP_FIELD_MASK = 0xffULL, 3201 BPF_F_USER_STACK = (1ULL << 8), 3202 /* flags used by BPF_FUNC_get_stackid only. */ 3203 BPF_F_FAST_STACK_CMP = (1ULL << 9), 3204 BPF_F_REUSE_STACKID = (1ULL << 10), 3205 /* flags used by BPF_FUNC_get_stack only. */ 3206 BPF_F_USER_BUILD_ID = (1ULL << 11), 3207 }; 3208 3209 /* BPF_FUNC_skb_set_tunnel_key flags. */ 3210 enum { 3211 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 3212 BPF_F_DONT_FRAGMENT = (1ULL << 2), 3213 BPF_F_SEQ_NUMBER = (1ULL << 3), 3214 }; 3215 3216 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3217 * BPF_FUNC_perf_event_read_value flags. 3218 */ 3219 enum { 3220 BPF_F_INDEX_MASK = 0xffffffffULL, 3221 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 3222 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 3223 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 3224 }; 3225 3226 /* Current network namespace */ 3227 enum { 3228 BPF_F_CURRENT_NETNS = (-1L), 3229 }; 3230 3231 /* BPF_FUNC_skb_adjust_room flags. */ 3232 enum { 3233 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 3234 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 3235 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 3236 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 3237 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 3238 }; 3239 3240 enum { 3241 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 3242 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 3243 }; 3244 3245 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3246 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3247 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3248 3249 /* BPF_FUNC_sysctl_get_name flags. */ 3250 enum { 3251 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 3252 }; 3253 3254 /* BPF_FUNC_sk_storage_get flags */ 3255 enum { 3256 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0), 3257 }; 3258 3259 /* BPF_FUNC_read_branch_records flags. */ 3260 enum { 3261 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 3262 }; 3263 3264 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 3265 enum bpf_adj_room_mode { 3266 BPF_ADJ_ROOM_NET, 3267 BPF_ADJ_ROOM_MAC, 3268 }; 3269 3270 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3271 enum bpf_hdr_start_off { 3272 BPF_HDR_START_MAC, 3273 BPF_HDR_START_NET, 3274 }; 3275 3276 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3277 enum bpf_lwt_encap_mode { 3278 BPF_LWT_ENCAP_SEG6, 3279 BPF_LWT_ENCAP_SEG6_INLINE, 3280 BPF_LWT_ENCAP_IP, 3281 }; 3282 3283 #define __bpf_md_ptr(type, name) \ 3284 union { \ 3285 type name; \ 3286 __u64 :64; \ 3287 } __attribute__((aligned(8))) 3288 3289 /* user accessible mirror of in-kernel sk_buff. 3290 * new fields can only be added to the end of this structure 3291 */ 3292 struct __sk_buff { 3293 __u32 len; 3294 __u32 pkt_type; 3295 __u32 mark; 3296 __u32 queue_mapping; 3297 __u32 protocol; 3298 __u32 vlan_present; 3299 __u32 vlan_tci; 3300 __u32 vlan_proto; 3301 __u32 priority; 3302 __u32 ingress_ifindex; 3303 __u32 ifindex; 3304 __u32 tc_index; 3305 __u32 cb[5]; 3306 __u32 hash; 3307 __u32 tc_classid; 3308 __u32 data; 3309 __u32 data_end; 3310 __u32 napi_id; 3311 3312 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3313 __u32 family; 3314 __u32 remote_ip4; /* Stored in network byte order */ 3315 __u32 local_ip4; /* Stored in network byte order */ 3316 __u32 remote_ip6[4]; /* Stored in network byte order */ 3317 __u32 local_ip6[4]; /* Stored in network byte order */ 3318 __u32 remote_port; /* Stored in network byte order */ 3319 __u32 local_port; /* stored in host byte order */ 3320 /* ... here. */ 3321 3322 __u32 data_meta; 3323 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3324 __u64 tstamp; 3325 __u32 wire_len; 3326 __u32 gso_segs; 3327 __bpf_md_ptr(struct bpf_sock *, sk); 3328 __u32 gso_size; 3329 }; 3330 3331 struct bpf_tunnel_key { 3332 __u32 tunnel_id; 3333 union { 3334 __u32 remote_ipv4; 3335 __u32 remote_ipv6[4]; 3336 }; 3337 __u8 tunnel_tos; 3338 __u8 tunnel_ttl; 3339 __u16 tunnel_ext; /* Padding, future use. */ 3340 __u32 tunnel_label; 3341 }; 3342 3343 /* user accessible mirror of in-kernel xfrm_state. 3344 * new fields can only be added to the end of this structure 3345 */ 3346 struct bpf_xfrm_state { 3347 __u32 reqid; 3348 __u32 spi; /* Stored in network byte order */ 3349 __u16 family; 3350 __u16 ext; /* Padding, future use. */ 3351 union { 3352 __u32 remote_ipv4; /* Stored in network byte order */ 3353 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3354 }; 3355 }; 3356 3357 /* Generic BPF return codes which all BPF program types may support. 3358 * The values are binary compatible with their TC_ACT_* counter-part to 3359 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3360 * programs. 3361 * 3362 * XDP is handled seprately, see XDP_*. 3363 */ 3364 enum bpf_ret_code { 3365 BPF_OK = 0, 3366 /* 1 reserved */ 3367 BPF_DROP = 2, 3368 /* 3-6 reserved */ 3369 BPF_REDIRECT = 7, 3370 /* >127 are reserved for prog type specific return codes. 3371 * 3372 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3373 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3374 * changed and should be routed based on its new L3 header. 3375 * (This is an L3 redirect, as opposed to L2 redirect 3376 * represented by BPF_REDIRECT above). 3377 */ 3378 BPF_LWT_REROUTE = 128, 3379 }; 3380 3381 struct bpf_sock { 3382 __u32 bound_dev_if; 3383 __u32 family; 3384 __u32 type; 3385 __u32 protocol; 3386 __u32 mark; 3387 __u32 priority; 3388 /* IP address also allows 1 and 2 bytes access */ 3389 __u32 src_ip4; 3390 __u32 src_ip6[4]; 3391 __u32 src_port; /* host byte order */ 3392 __u32 dst_port; /* network byte order */ 3393 __u32 dst_ip4; 3394 __u32 dst_ip6[4]; 3395 __u32 state; 3396 }; 3397 3398 struct bpf_tcp_sock { 3399 __u32 snd_cwnd; /* Sending congestion window */ 3400 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3401 __u32 rtt_min; 3402 __u32 snd_ssthresh; /* Slow start size threshold */ 3403 __u32 rcv_nxt; /* What we want to receive next */ 3404 __u32 snd_nxt; /* Next sequence we send */ 3405 __u32 snd_una; /* First byte we want an ack for */ 3406 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3407 __u32 ecn_flags; /* ECN status bits. */ 3408 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3409 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3410 __u32 packets_out; /* Packets which are "in flight" */ 3411 __u32 retrans_out; /* Retransmitted packets out */ 3412 __u32 total_retrans; /* Total retransmits for entire connection */ 3413 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3414 * total number of segments in. 3415 */ 3416 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3417 * total number of data segments in. 3418 */ 3419 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3420 * The total number of segments sent. 3421 */ 3422 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3423 * total number of data segments sent. 3424 */ 3425 __u32 lost_out; /* Lost packets */ 3426 __u32 sacked_out; /* SACK'd packets */ 3427 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3428 * sum(delta(rcv_nxt)), or how many bytes 3429 * were acked. 3430 */ 3431 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3432 * sum(delta(snd_una)), or how many bytes 3433 * were acked. 3434 */ 3435 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3436 * total number of DSACK blocks received 3437 */ 3438 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3439 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3440 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3441 }; 3442 3443 struct bpf_sock_tuple { 3444 union { 3445 struct { 3446 __be32 saddr; 3447 __be32 daddr; 3448 __be16 sport; 3449 __be16 dport; 3450 } ipv4; 3451 struct { 3452 __be32 saddr[4]; 3453 __be32 daddr[4]; 3454 __be16 sport; 3455 __be16 dport; 3456 } ipv6; 3457 }; 3458 }; 3459 3460 struct bpf_xdp_sock { 3461 __u32 queue_id; 3462 }; 3463 3464 #define XDP_PACKET_HEADROOM 256 3465 3466 /* User return codes for XDP prog type. 3467 * A valid XDP program must return one of these defined values. All other 3468 * return codes are reserved for future use. Unknown return codes will 3469 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3470 */ 3471 enum xdp_action { 3472 XDP_ABORTED = 0, 3473 XDP_DROP, 3474 XDP_PASS, 3475 XDP_TX, 3476 XDP_REDIRECT, 3477 }; 3478 3479 /* user accessible metadata for XDP packet hook 3480 * new fields must be added to the end of this structure 3481 */ 3482 struct xdp_md { 3483 __u32 data; 3484 __u32 data_end; 3485 __u32 data_meta; 3486 /* Below access go through struct xdp_rxq_info */ 3487 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3488 __u32 rx_queue_index; /* rxq->queue_index */ 3489 }; 3490 3491 enum sk_action { 3492 SK_DROP = 0, 3493 SK_PASS, 3494 }; 3495 3496 /* user accessible metadata for SK_MSG packet hook, new fields must 3497 * be added to the end of this structure 3498 */ 3499 struct sk_msg_md { 3500 __bpf_md_ptr(void *, data); 3501 __bpf_md_ptr(void *, data_end); 3502 3503 __u32 family; 3504 __u32 remote_ip4; /* Stored in network byte order */ 3505 __u32 local_ip4; /* Stored in network byte order */ 3506 __u32 remote_ip6[4]; /* Stored in network byte order */ 3507 __u32 local_ip6[4]; /* Stored in network byte order */ 3508 __u32 remote_port; /* Stored in network byte order */ 3509 __u32 local_port; /* stored in host byte order */ 3510 __u32 size; /* Total size of sk_msg */ 3511 }; 3512 3513 struct sk_reuseport_md { 3514 /* 3515 * Start of directly accessible data. It begins from 3516 * the tcp/udp header. 3517 */ 3518 __bpf_md_ptr(void *, data); 3519 /* End of directly accessible data */ 3520 __bpf_md_ptr(void *, data_end); 3521 /* 3522 * Total length of packet (starting from the tcp/udp header). 3523 * Note that the directly accessible bytes (data_end - data) 3524 * could be less than this "len". Those bytes could be 3525 * indirectly read by a helper "bpf_skb_load_bytes()". 3526 */ 3527 __u32 len; 3528 /* 3529 * Eth protocol in the mac header (network byte order). e.g. 3530 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3531 */ 3532 __u32 eth_protocol; 3533 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3534 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3535 __u32 hash; /* A hash of the packet 4 tuples */ 3536 }; 3537 3538 #define BPF_TAG_SIZE 8 3539 3540 struct bpf_prog_info { 3541 __u32 type; 3542 __u32 id; 3543 __u8 tag[BPF_TAG_SIZE]; 3544 __u32 jited_prog_len; 3545 __u32 xlated_prog_len; 3546 __aligned_u64 jited_prog_insns; 3547 __aligned_u64 xlated_prog_insns; 3548 __u64 load_time; /* ns since boottime */ 3549 __u32 created_by_uid; 3550 __u32 nr_map_ids; 3551 __aligned_u64 map_ids; 3552 char name[BPF_OBJ_NAME_LEN]; 3553 __u32 ifindex; 3554 __u32 gpl_compatible:1; 3555 __u32 :31; /* alignment pad */ 3556 __u64 netns_dev; 3557 __u64 netns_ino; 3558 __u32 nr_jited_ksyms; 3559 __u32 nr_jited_func_lens; 3560 __aligned_u64 jited_ksyms; 3561 __aligned_u64 jited_func_lens; 3562 __u32 btf_id; 3563 __u32 func_info_rec_size; 3564 __aligned_u64 func_info; 3565 __u32 nr_func_info; 3566 __u32 nr_line_info; 3567 __aligned_u64 line_info; 3568 __aligned_u64 jited_line_info; 3569 __u32 nr_jited_line_info; 3570 __u32 line_info_rec_size; 3571 __u32 jited_line_info_rec_size; 3572 __u32 nr_prog_tags; 3573 __aligned_u64 prog_tags; 3574 __u64 run_time_ns; 3575 __u64 run_cnt; 3576 } __attribute__((aligned(8))); 3577 3578 struct bpf_map_info { 3579 __u32 type; 3580 __u32 id; 3581 __u32 key_size; 3582 __u32 value_size; 3583 __u32 max_entries; 3584 __u32 map_flags; 3585 char name[BPF_OBJ_NAME_LEN]; 3586 __u32 ifindex; 3587 __u32 btf_vmlinux_value_type_id; 3588 __u64 netns_dev; 3589 __u64 netns_ino; 3590 __u32 btf_id; 3591 __u32 btf_key_type_id; 3592 __u32 btf_value_type_id; 3593 } __attribute__((aligned(8))); 3594 3595 struct bpf_btf_info { 3596 __aligned_u64 btf; 3597 __u32 btf_size; 3598 __u32 id; 3599 } __attribute__((aligned(8))); 3600 3601 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3602 * by user and intended to be used by socket (e.g. to bind to, depends on 3603 * attach attach type). 3604 */ 3605 struct bpf_sock_addr { 3606 __u32 user_family; /* Allows 4-byte read, but no write. */ 3607 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3608 * Stored in network byte order. 3609 */ 3610 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3611 * Stored in network byte order. 3612 */ 3613 __u32 user_port; /* Allows 4-byte read and write. 3614 * Stored in network byte order 3615 */ 3616 __u32 family; /* Allows 4-byte read, but no write */ 3617 __u32 type; /* Allows 4-byte read, but no write */ 3618 __u32 protocol; /* Allows 4-byte read, but no write */ 3619 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3620 * Stored in network byte order. 3621 */ 3622 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3623 * Stored in network byte order. 3624 */ 3625 __bpf_md_ptr(struct bpf_sock *, sk); 3626 }; 3627 3628 /* User bpf_sock_ops struct to access socket values and specify request ops 3629 * and their replies. 3630 * Some of this fields are in network (bigendian) byte order and may need 3631 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3632 * New fields can only be added at the end of this structure 3633 */ 3634 struct bpf_sock_ops { 3635 __u32 op; 3636 union { 3637 __u32 args[4]; /* Optionally passed to bpf program */ 3638 __u32 reply; /* Returned by bpf program */ 3639 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3640 }; 3641 __u32 family; 3642 __u32 remote_ip4; /* Stored in network byte order */ 3643 __u32 local_ip4; /* Stored in network byte order */ 3644 __u32 remote_ip6[4]; /* Stored in network byte order */ 3645 __u32 local_ip6[4]; /* Stored in network byte order */ 3646 __u32 remote_port; /* Stored in network byte order */ 3647 __u32 local_port; /* stored in host byte order */ 3648 __u32 is_fullsock; /* Some TCP fields are only valid if 3649 * there is a full socket. If not, the 3650 * fields read as zero. 3651 */ 3652 __u32 snd_cwnd; 3653 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3654 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3655 __u32 state; 3656 __u32 rtt_min; 3657 __u32 snd_ssthresh; 3658 __u32 rcv_nxt; 3659 __u32 snd_nxt; 3660 __u32 snd_una; 3661 __u32 mss_cache; 3662 __u32 ecn_flags; 3663 __u32 rate_delivered; 3664 __u32 rate_interval_us; 3665 __u32 packets_out; 3666 __u32 retrans_out; 3667 __u32 total_retrans; 3668 __u32 segs_in; 3669 __u32 data_segs_in; 3670 __u32 segs_out; 3671 __u32 data_segs_out; 3672 __u32 lost_out; 3673 __u32 sacked_out; 3674 __u32 sk_txhash; 3675 __u64 bytes_received; 3676 __u64 bytes_acked; 3677 __bpf_md_ptr(struct bpf_sock *, sk); 3678 }; 3679 3680 /* Definitions for bpf_sock_ops_cb_flags */ 3681 enum { 3682 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 3683 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 3684 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 3685 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 3686 /* Mask of all currently supported cb flags */ 3687 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF, 3688 }; 3689 3690 /* List of known BPF sock_ops operators. 3691 * New entries can only be added at the end 3692 */ 3693 enum { 3694 BPF_SOCK_OPS_VOID, 3695 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3696 * -1 if default value should be used 3697 */ 3698 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3699 * window (in packets) or -1 if default 3700 * value should be used 3701 */ 3702 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3703 * active connection is initialized 3704 */ 3705 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3706 * active connection is 3707 * established 3708 */ 3709 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3710 * passive connection is 3711 * established 3712 */ 3713 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3714 * needs ECN 3715 */ 3716 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3717 * based on the path and may be 3718 * dependent on the congestion control 3719 * algorithm. In general it indicates 3720 * a congestion threshold. RTTs above 3721 * this indicate congestion 3722 */ 3723 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3724 * Arg1: value of icsk_retransmits 3725 * Arg2: value of icsk_rto 3726 * Arg3: whether RTO has expired 3727 */ 3728 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3729 * Arg1: sequence number of 1st byte 3730 * Arg2: # segments 3731 * Arg3: return value of 3732 * tcp_transmit_skb (0 => success) 3733 */ 3734 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3735 * Arg1: old_state 3736 * Arg2: new_state 3737 */ 3738 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3739 * socket transition to LISTEN state. 3740 */ 3741 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 3742 */ 3743 }; 3744 3745 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3746 * changes between the TCP and BPF versions. Ideally this should never happen. 3747 * If it does, we need to add code to convert them before calling 3748 * the BPF sock_ops function. 3749 */ 3750 enum { 3751 BPF_TCP_ESTABLISHED = 1, 3752 BPF_TCP_SYN_SENT, 3753 BPF_TCP_SYN_RECV, 3754 BPF_TCP_FIN_WAIT1, 3755 BPF_TCP_FIN_WAIT2, 3756 BPF_TCP_TIME_WAIT, 3757 BPF_TCP_CLOSE, 3758 BPF_TCP_CLOSE_WAIT, 3759 BPF_TCP_LAST_ACK, 3760 BPF_TCP_LISTEN, 3761 BPF_TCP_CLOSING, /* Now a valid state */ 3762 BPF_TCP_NEW_SYN_RECV, 3763 3764 BPF_TCP_MAX_STATES /* Leave at the end! */ 3765 }; 3766 3767 enum { 3768 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 3769 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 3770 }; 3771 3772 struct bpf_perf_event_value { 3773 __u64 counter; 3774 __u64 enabled; 3775 __u64 running; 3776 }; 3777 3778 enum { 3779 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 3780 BPF_DEVCG_ACC_READ = (1ULL << 1), 3781 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 3782 }; 3783 3784 enum { 3785 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 3786 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 3787 }; 3788 3789 struct bpf_cgroup_dev_ctx { 3790 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3791 __u32 access_type; 3792 __u32 major; 3793 __u32 minor; 3794 }; 3795 3796 struct bpf_raw_tracepoint_args { 3797 __u64 args[0]; 3798 }; 3799 3800 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3801 * OUTPUT: Do lookup from egress perspective; default is ingress 3802 */ 3803 enum { 3804 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 3805 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 3806 }; 3807 3808 enum { 3809 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3810 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3811 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3812 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3813 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3814 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3815 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3816 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3817 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3818 }; 3819 3820 struct bpf_fib_lookup { 3821 /* input: network family for lookup (AF_INET, AF_INET6) 3822 * output: network family of egress nexthop 3823 */ 3824 __u8 family; 3825 3826 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3827 __u8 l4_protocol; 3828 __be16 sport; 3829 __be16 dport; 3830 3831 /* total length of packet from network header - used for MTU check */ 3832 __u16 tot_len; 3833 3834 /* input: L3 device index for lookup 3835 * output: device index from FIB lookup 3836 */ 3837 __u32 ifindex; 3838 3839 union { 3840 /* inputs to lookup */ 3841 __u8 tos; /* AF_INET */ 3842 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3843 3844 /* output: metric of fib result (IPv4/IPv6 only) */ 3845 __u32 rt_metric; 3846 }; 3847 3848 union { 3849 __be32 ipv4_src; 3850 __u32 ipv6_src[4]; /* in6_addr; network order */ 3851 }; 3852 3853 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3854 * network header. output: bpf_fib_lookup sets to gateway address 3855 * if FIB lookup returns gateway route 3856 */ 3857 union { 3858 __be32 ipv4_dst; 3859 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3860 }; 3861 3862 /* output */ 3863 __be16 h_vlan_proto; 3864 __be16 h_vlan_TCI; 3865 __u8 smac[6]; /* ETH_ALEN */ 3866 __u8 dmac[6]; /* ETH_ALEN */ 3867 }; 3868 3869 enum bpf_task_fd_type { 3870 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3871 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3872 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3873 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3874 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3875 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3876 }; 3877 3878 enum { 3879 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 3880 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 3881 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 3882 }; 3883 3884 struct bpf_flow_keys { 3885 __u16 nhoff; 3886 __u16 thoff; 3887 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3888 __u8 is_frag; 3889 __u8 is_first_frag; 3890 __u8 is_encap; 3891 __u8 ip_proto; 3892 __be16 n_proto; 3893 __be16 sport; 3894 __be16 dport; 3895 union { 3896 struct { 3897 __be32 ipv4_src; 3898 __be32 ipv4_dst; 3899 }; 3900 struct { 3901 __u32 ipv6_src[4]; /* in6_addr; network order */ 3902 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3903 }; 3904 }; 3905 __u32 flags; 3906 __be32 flow_label; 3907 }; 3908 3909 struct bpf_func_info { 3910 __u32 insn_off; 3911 __u32 type_id; 3912 }; 3913 3914 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3915 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3916 3917 struct bpf_line_info { 3918 __u32 insn_off; 3919 __u32 file_name_off; 3920 __u32 line_off; 3921 __u32 line_col; 3922 }; 3923 3924 struct bpf_spin_lock { 3925 __u32 val; 3926 }; 3927 3928 struct bpf_sysctl { 3929 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3930 * Allows 1,2,4-byte read, but no write. 3931 */ 3932 __u32 file_pos; /* Sysctl file position to read from, write to. 3933 * Allows 1,2,4-byte read an 4-byte write. 3934 */ 3935 }; 3936 3937 struct bpf_sockopt { 3938 __bpf_md_ptr(struct bpf_sock *, sk); 3939 __bpf_md_ptr(void *, optval); 3940 __bpf_md_ptr(void *, optval_end); 3941 3942 __s32 level; 3943 __s32 optname; 3944 __s32 optlen; 3945 __s32 retval; 3946 }; 3947 3948 struct bpf_pidns_info { 3949 __u32 pid; 3950 __u32 tgid; 3951 }; 3952 #endif /* _UAPI__LINUX_BPF_H__ */ 3953