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_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24 25 /* alu/jmp fields */ 26 #define BPF_MOV 0xb0 /* mov reg to reg */ 27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28 29 /* change endianness of a register */ 30 #define BPF_END 0xd0 /* flags for endianness conversion: */ 31 #define BPF_TO_LE 0x00 /* convert to little-endian */ 32 #define BPF_TO_BE 0x08 /* convert to big-endian */ 33 #define BPF_FROM_LE BPF_TO_LE 34 #define BPF_FROM_BE BPF_TO_BE 35 36 /* jmp encodings */ 37 #define BPF_JNE 0x50 /* jump != */ 38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44 #define BPF_CALL 0x80 /* function call */ 45 #define BPF_EXIT 0x90 /* function return */ 46 47 /* atomic op type fields (stored in immediate) */ 48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 51 52 /* Register numbers */ 53 enum { 54 BPF_REG_0 = 0, 55 BPF_REG_1, 56 BPF_REG_2, 57 BPF_REG_3, 58 BPF_REG_4, 59 BPF_REG_5, 60 BPF_REG_6, 61 BPF_REG_7, 62 BPF_REG_8, 63 BPF_REG_9, 64 BPF_REG_10, 65 __MAX_BPF_REG, 66 }; 67 68 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 69 #define MAX_BPF_REG __MAX_BPF_REG 70 71 struct bpf_insn { 72 __u8 code; /* opcode */ 73 __u8 dst_reg:4; /* dest register */ 74 __u8 src_reg:4; /* source register */ 75 __s16 off; /* signed offset */ 76 __s32 imm; /* signed immediate constant */ 77 }; 78 79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80 struct bpf_lpm_trie_key { 81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82 __u8 data[0]; /* Arbitrary size */ 83 }; 84 85 struct bpf_cgroup_storage_key { 86 __u64 cgroup_inode_id; /* cgroup inode id */ 87 __u32 attach_type; /* program attach type */ 88 }; 89 90 union bpf_iter_link_info { 91 struct { 92 __u32 map_fd; 93 } map; 94 }; 95 96 /* BPF syscall commands, see bpf(2) man-page for details. */ 97 enum bpf_cmd { 98 BPF_MAP_CREATE, 99 BPF_MAP_LOOKUP_ELEM, 100 BPF_MAP_UPDATE_ELEM, 101 BPF_MAP_DELETE_ELEM, 102 BPF_MAP_GET_NEXT_KEY, 103 BPF_PROG_LOAD, 104 BPF_OBJ_PIN, 105 BPF_OBJ_GET, 106 BPF_PROG_ATTACH, 107 BPF_PROG_DETACH, 108 BPF_PROG_TEST_RUN, 109 BPF_PROG_GET_NEXT_ID, 110 BPF_MAP_GET_NEXT_ID, 111 BPF_PROG_GET_FD_BY_ID, 112 BPF_MAP_GET_FD_BY_ID, 113 BPF_OBJ_GET_INFO_BY_FD, 114 BPF_PROG_QUERY, 115 BPF_RAW_TRACEPOINT_OPEN, 116 BPF_BTF_LOAD, 117 BPF_BTF_GET_FD_BY_ID, 118 BPF_TASK_FD_QUERY, 119 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 120 BPF_MAP_FREEZE, 121 BPF_BTF_GET_NEXT_ID, 122 BPF_MAP_LOOKUP_BATCH, 123 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 124 BPF_MAP_UPDATE_BATCH, 125 BPF_MAP_DELETE_BATCH, 126 BPF_LINK_CREATE, 127 BPF_LINK_UPDATE, 128 BPF_LINK_GET_FD_BY_ID, 129 BPF_LINK_GET_NEXT_ID, 130 BPF_ENABLE_STATS, 131 BPF_ITER_CREATE, 132 BPF_LINK_DETACH, 133 BPF_PROG_BIND_MAP, 134 }; 135 136 enum bpf_map_type { 137 BPF_MAP_TYPE_UNSPEC, 138 BPF_MAP_TYPE_HASH, 139 BPF_MAP_TYPE_ARRAY, 140 BPF_MAP_TYPE_PROG_ARRAY, 141 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 142 BPF_MAP_TYPE_PERCPU_HASH, 143 BPF_MAP_TYPE_PERCPU_ARRAY, 144 BPF_MAP_TYPE_STACK_TRACE, 145 BPF_MAP_TYPE_CGROUP_ARRAY, 146 BPF_MAP_TYPE_LRU_HASH, 147 BPF_MAP_TYPE_LRU_PERCPU_HASH, 148 BPF_MAP_TYPE_LPM_TRIE, 149 BPF_MAP_TYPE_ARRAY_OF_MAPS, 150 BPF_MAP_TYPE_HASH_OF_MAPS, 151 BPF_MAP_TYPE_DEVMAP, 152 BPF_MAP_TYPE_SOCKMAP, 153 BPF_MAP_TYPE_CPUMAP, 154 BPF_MAP_TYPE_XSKMAP, 155 BPF_MAP_TYPE_SOCKHASH, 156 BPF_MAP_TYPE_CGROUP_STORAGE, 157 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 158 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 159 BPF_MAP_TYPE_QUEUE, 160 BPF_MAP_TYPE_STACK, 161 BPF_MAP_TYPE_SK_STORAGE, 162 BPF_MAP_TYPE_DEVMAP_HASH, 163 BPF_MAP_TYPE_STRUCT_OPS, 164 BPF_MAP_TYPE_RINGBUF, 165 BPF_MAP_TYPE_INODE_STORAGE, 166 BPF_MAP_TYPE_TASK_STORAGE, 167 }; 168 169 /* Note that tracing related programs such as 170 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 171 * are not subject to a stable API since kernel internal data 172 * structures can change from release to release and may 173 * therefore break existing tracing BPF programs. Tracing BPF 174 * programs correspond to /a/ specific kernel which is to be 175 * analyzed, and not /a/ specific kernel /and/ all future ones. 176 */ 177 enum bpf_prog_type { 178 BPF_PROG_TYPE_UNSPEC, 179 BPF_PROG_TYPE_SOCKET_FILTER, 180 BPF_PROG_TYPE_KPROBE, 181 BPF_PROG_TYPE_SCHED_CLS, 182 BPF_PROG_TYPE_SCHED_ACT, 183 BPF_PROG_TYPE_TRACEPOINT, 184 BPF_PROG_TYPE_XDP, 185 BPF_PROG_TYPE_PERF_EVENT, 186 BPF_PROG_TYPE_CGROUP_SKB, 187 BPF_PROG_TYPE_CGROUP_SOCK, 188 BPF_PROG_TYPE_LWT_IN, 189 BPF_PROG_TYPE_LWT_OUT, 190 BPF_PROG_TYPE_LWT_XMIT, 191 BPF_PROG_TYPE_SOCK_OPS, 192 BPF_PROG_TYPE_SK_SKB, 193 BPF_PROG_TYPE_CGROUP_DEVICE, 194 BPF_PROG_TYPE_SK_MSG, 195 BPF_PROG_TYPE_RAW_TRACEPOINT, 196 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 197 BPF_PROG_TYPE_LWT_SEG6LOCAL, 198 BPF_PROG_TYPE_LIRC_MODE2, 199 BPF_PROG_TYPE_SK_REUSEPORT, 200 BPF_PROG_TYPE_FLOW_DISSECTOR, 201 BPF_PROG_TYPE_CGROUP_SYSCTL, 202 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 203 BPF_PROG_TYPE_CGROUP_SOCKOPT, 204 BPF_PROG_TYPE_TRACING, 205 BPF_PROG_TYPE_STRUCT_OPS, 206 BPF_PROG_TYPE_EXT, 207 BPF_PROG_TYPE_LSM, 208 BPF_PROG_TYPE_SK_LOOKUP, 209 }; 210 211 enum bpf_attach_type { 212 BPF_CGROUP_INET_INGRESS, 213 BPF_CGROUP_INET_EGRESS, 214 BPF_CGROUP_INET_SOCK_CREATE, 215 BPF_CGROUP_SOCK_OPS, 216 BPF_SK_SKB_STREAM_PARSER, 217 BPF_SK_SKB_STREAM_VERDICT, 218 BPF_CGROUP_DEVICE, 219 BPF_SK_MSG_VERDICT, 220 BPF_CGROUP_INET4_BIND, 221 BPF_CGROUP_INET6_BIND, 222 BPF_CGROUP_INET4_CONNECT, 223 BPF_CGROUP_INET6_CONNECT, 224 BPF_CGROUP_INET4_POST_BIND, 225 BPF_CGROUP_INET6_POST_BIND, 226 BPF_CGROUP_UDP4_SENDMSG, 227 BPF_CGROUP_UDP6_SENDMSG, 228 BPF_LIRC_MODE2, 229 BPF_FLOW_DISSECTOR, 230 BPF_CGROUP_SYSCTL, 231 BPF_CGROUP_UDP4_RECVMSG, 232 BPF_CGROUP_UDP6_RECVMSG, 233 BPF_CGROUP_GETSOCKOPT, 234 BPF_CGROUP_SETSOCKOPT, 235 BPF_TRACE_RAW_TP, 236 BPF_TRACE_FENTRY, 237 BPF_TRACE_FEXIT, 238 BPF_MODIFY_RETURN, 239 BPF_LSM_MAC, 240 BPF_TRACE_ITER, 241 BPF_CGROUP_INET4_GETPEERNAME, 242 BPF_CGROUP_INET6_GETPEERNAME, 243 BPF_CGROUP_INET4_GETSOCKNAME, 244 BPF_CGROUP_INET6_GETSOCKNAME, 245 BPF_XDP_DEVMAP, 246 BPF_CGROUP_INET_SOCK_RELEASE, 247 BPF_XDP_CPUMAP, 248 BPF_SK_LOOKUP, 249 BPF_XDP, 250 __MAX_BPF_ATTACH_TYPE 251 }; 252 253 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 254 255 enum bpf_link_type { 256 BPF_LINK_TYPE_UNSPEC = 0, 257 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 258 BPF_LINK_TYPE_TRACING = 2, 259 BPF_LINK_TYPE_CGROUP = 3, 260 BPF_LINK_TYPE_ITER = 4, 261 BPF_LINK_TYPE_NETNS = 5, 262 BPF_LINK_TYPE_XDP = 6, 263 264 MAX_BPF_LINK_TYPE, 265 }; 266 267 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 268 * 269 * NONE(default): No further bpf programs allowed in the subtree. 270 * 271 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 272 * the program in this cgroup yields to sub-cgroup program. 273 * 274 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 275 * that cgroup program gets run in addition to the program in this cgroup. 276 * 277 * Only one program is allowed to be attached to a cgroup with 278 * NONE or BPF_F_ALLOW_OVERRIDE flag. 279 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 280 * release old program and attach the new one. Attach flags has to match. 281 * 282 * Multiple programs are allowed to be attached to a cgroup with 283 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 284 * (those that were attached first, run first) 285 * The programs of sub-cgroup are executed first, then programs of 286 * this cgroup and then programs of parent cgroup. 287 * When children program makes decision (like picking TCP CA or sock bind) 288 * parent program has a chance to override it. 289 * 290 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 291 * programs for a cgroup. Though it's possible to replace an old program at 292 * any position by also specifying BPF_F_REPLACE flag and position itself in 293 * replace_bpf_fd attribute. Old program at this position will be released. 294 * 295 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 296 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 297 * Ex1: 298 * cgrp1 (MULTI progs A, B) -> 299 * cgrp2 (OVERRIDE prog C) -> 300 * cgrp3 (MULTI prog D) -> 301 * cgrp4 (OVERRIDE prog E) -> 302 * cgrp5 (NONE prog F) 303 * the event in cgrp5 triggers execution of F,D,A,B in that order. 304 * if prog F is detached, the execution is E,D,A,B 305 * if prog F and D are detached, the execution is E,A,B 306 * if prog F, E and D are detached, the execution is C,A,B 307 * 308 * All eligible programs are executed regardless of return code from 309 * earlier programs. 310 */ 311 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 312 #define BPF_F_ALLOW_MULTI (1U << 1) 313 #define BPF_F_REPLACE (1U << 2) 314 315 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 316 * verifier will perform strict alignment checking as if the kernel 317 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 318 * and NET_IP_ALIGN defined to 2. 319 */ 320 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 321 322 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 323 * verifier will allow any alignment whatsoever. On platforms 324 * with strict alignment requirements for loads ands stores (such 325 * as sparc and mips) the verifier validates that all loads and 326 * stores provably follow this requirement. This flag turns that 327 * checking and enforcement off. 328 * 329 * It is mostly used for testing when we want to validate the 330 * context and memory access aspects of the verifier, but because 331 * of an unaligned access the alignment check would trigger before 332 * the one we are interested in. 333 */ 334 #define BPF_F_ANY_ALIGNMENT (1U << 1) 335 336 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 337 * Verifier does sub-register def/use analysis and identifies instructions whose 338 * def only matters for low 32-bit, high 32-bit is never referenced later 339 * through implicit zero extension. Therefore verifier notifies JIT back-ends 340 * that it is safe to ignore clearing high 32-bit for these instructions. This 341 * saves some back-ends a lot of code-gen. However such optimization is not 342 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 343 * hence hasn't used verifier's analysis result. But, we really want to have a 344 * way to be able to verify the correctness of the described optimization on 345 * x86_64 on which testsuites are frequently exercised. 346 * 347 * So, this flag is introduced. Once it is set, verifier will randomize high 348 * 32-bit for those instructions who has been identified as safe to ignore them. 349 * Then, if verifier is not doing correct analysis, such randomization will 350 * regress tests to expose bugs. 351 */ 352 #define BPF_F_TEST_RND_HI32 (1U << 2) 353 354 /* The verifier internal test flag. Behavior is undefined */ 355 #define BPF_F_TEST_STATE_FREQ (1U << 3) 356 357 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 358 * restrict map and helper usage for such programs. Sleepable BPF programs can 359 * only be attached to hooks where kernel execution context allows sleeping. 360 * Such programs are allowed to use helpers that may sleep like 361 * bpf_copy_from_user(). 362 */ 363 #define BPF_F_SLEEPABLE (1U << 4) 364 365 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 366 * the following extensions: 367 * 368 * insn[0].src_reg: BPF_PSEUDO_MAP_FD 369 * insn[0].imm: map fd 370 * insn[1].imm: 0 371 * insn[0].off: 0 372 * insn[1].off: 0 373 * ldimm64 rewrite: address of map 374 * verifier type: CONST_PTR_TO_MAP 375 */ 376 #define BPF_PSEUDO_MAP_FD 1 377 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE 378 * insn[0].imm: map fd 379 * insn[1].imm: offset into value 380 * insn[0].off: 0 381 * insn[1].off: 0 382 * ldimm64 rewrite: address of map[0]+offset 383 * verifier type: PTR_TO_MAP_VALUE 384 */ 385 #define BPF_PSEUDO_MAP_VALUE 2 386 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 387 * insn[0].imm: kernel btd id of VAR 388 * insn[1].imm: 0 389 * insn[0].off: 0 390 * insn[1].off: 0 391 * ldimm64 rewrite: address of the kernel variable 392 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 393 * is struct/union. 394 */ 395 #define BPF_PSEUDO_BTF_ID 3 396 397 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 398 * offset to another bpf function 399 */ 400 #define BPF_PSEUDO_CALL 1 401 402 /* flags for BPF_MAP_UPDATE_ELEM command */ 403 enum { 404 BPF_ANY = 0, /* create new element or update existing */ 405 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 406 BPF_EXIST = 2, /* update existing element */ 407 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 408 }; 409 410 /* flags for BPF_MAP_CREATE command */ 411 enum { 412 BPF_F_NO_PREALLOC = (1U << 0), 413 /* Instead of having one common LRU list in the 414 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 415 * which can scale and perform better. 416 * Note, the LRU nodes (including free nodes) cannot be moved 417 * across different LRU lists. 418 */ 419 BPF_F_NO_COMMON_LRU = (1U << 1), 420 /* Specify numa node during map creation */ 421 BPF_F_NUMA_NODE = (1U << 2), 422 423 /* Flags for accessing BPF object from syscall side. */ 424 BPF_F_RDONLY = (1U << 3), 425 BPF_F_WRONLY = (1U << 4), 426 427 /* Flag for stack_map, store build_id+offset instead of pointer */ 428 BPF_F_STACK_BUILD_ID = (1U << 5), 429 430 /* Zero-initialize hash function seed. This should only be used for testing. */ 431 BPF_F_ZERO_SEED = (1U << 6), 432 433 /* Flags for accessing BPF object from program side. */ 434 BPF_F_RDONLY_PROG = (1U << 7), 435 BPF_F_WRONLY_PROG = (1U << 8), 436 437 /* Clone map from listener for newly accepted socket */ 438 BPF_F_CLONE = (1U << 9), 439 440 /* Enable memory-mapping BPF map */ 441 BPF_F_MMAPABLE = (1U << 10), 442 443 /* Share perf_event among processes */ 444 BPF_F_PRESERVE_ELEMS = (1U << 11), 445 446 /* Create a map that is suitable to be an inner map with dynamic max entries */ 447 BPF_F_INNER_MAP = (1U << 12), 448 }; 449 450 /* Flags for BPF_PROG_QUERY. */ 451 452 /* Query effective (directly attached + inherited from ancestor cgroups) 453 * programs that will be executed for events within a cgroup. 454 * attach_flags with this flag are returned only for directly attached programs. 455 */ 456 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 457 458 /* Flags for BPF_PROG_TEST_RUN */ 459 460 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 461 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 462 463 /* type for BPF_ENABLE_STATS */ 464 enum bpf_stats_type { 465 /* enabled run_time_ns and run_cnt */ 466 BPF_STATS_RUN_TIME = 0, 467 }; 468 469 enum bpf_stack_build_id_status { 470 /* user space need an empty entry to identify end of a trace */ 471 BPF_STACK_BUILD_ID_EMPTY = 0, 472 /* with valid build_id and offset */ 473 BPF_STACK_BUILD_ID_VALID = 1, 474 /* couldn't get build_id, fallback to ip */ 475 BPF_STACK_BUILD_ID_IP = 2, 476 }; 477 478 #define BPF_BUILD_ID_SIZE 20 479 struct bpf_stack_build_id { 480 __s32 status; 481 unsigned char build_id[BPF_BUILD_ID_SIZE]; 482 union { 483 __u64 offset; 484 __u64 ip; 485 }; 486 }; 487 488 #define BPF_OBJ_NAME_LEN 16U 489 490 union bpf_attr { 491 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 492 __u32 map_type; /* one of enum bpf_map_type */ 493 __u32 key_size; /* size of key in bytes */ 494 __u32 value_size; /* size of value in bytes */ 495 __u32 max_entries; /* max number of entries in a map */ 496 __u32 map_flags; /* BPF_MAP_CREATE related 497 * flags defined above. 498 */ 499 __u32 inner_map_fd; /* fd pointing to the inner map */ 500 __u32 numa_node; /* numa node (effective only if 501 * BPF_F_NUMA_NODE is set). 502 */ 503 char map_name[BPF_OBJ_NAME_LEN]; 504 __u32 map_ifindex; /* ifindex of netdev to create on */ 505 __u32 btf_fd; /* fd pointing to a BTF type data */ 506 __u32 btf_key_type_id; /* BTF type_id of the key */ 507 __u32 btf_value_type_id; /* BTF type_id of the value */ 508 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 509 * struct stored as the 510 * map value 511 */ 512 }; 513 514 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 515 __u32 map_fd; 516 __aligned_u64 key; 517 union { 518 __aligned_u64 value; 519 __aligned_u64 next_key; 520 }; 521 __u64 flags; 522 }; 523 524 struct { /* struct used by BPF_MAP_*_BATCH commands */ 525 __aligned_u64 in_batch; /* start batch, 526 * NULL to start from beginning 527 */ 528 __aligned_u64 out_batch; /* output: next start batch */ 529 __aligned_u64 keys; 530 __aligned_u64 values; 531 __u32 count; /* input/output: 532 * input: # of key/value 533 * elements 534 * output: # of filled elements 535 */ 536 __u32 map_fd; 537 __u64 elem_flags; 538 __u64 flags; 539 } batch; 540 541 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 542 __u32 prog_type; /* one of enum bpf_prog_type */ 543 __u32 insn_cnt; 544 __aligned_u64 insns; 545 __aligned_u64 license; 546 __u32 log_level; /* verbosity level of verifier */ 547 __u32 log_size; /* size of user buffer */ 548 __aligned_u64 log_buf; /* user supplied buffer */ 549 __u32 kern_version; /* not used */ 550 __u32 prog_flags; 551 char prog_name[BPF_OBJ_NAME_LEN]; 552 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 553 /* For some prog types expected attach type must be known at 554 * load time to verify attach type specific parts of prog 555 * (context accesses, allowed helpers, etc). 556 */ 557 __u32 expected_attach_type; 558 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 559 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 560 __aligned_u64 func_info; /* func info */ 561 __u32 func_info_cnt; /* number of bpf_func_info records */ 562 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 563 __aligned_u64 line_info; /* line info */ 564 __u32 line_info_cnt; /* number of bpf_line_info records */ 565 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 566 union { 567 /* valid prog_fd to attach to bpf prog */ 568 __u32 attach_prog_fd; 569 /* or valid module BTF object fd or 0 to attach to vmlinux */ 570 __u32 attach_btf_obj_fd; 571 }; 572 }; 573 574 struct { /* anonymous struct used by BPF_OBJ_* commands */ 575 __aligned_u64 pathname; 576 __u32 bpf_fd; 577 __u32 file_flags; 578 }; 579 580 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 581 __u32 target_fd; /* container object to attach to */ 582 __u32 attach_bpf_fd; /* eBPF program to attach */ 583 __u32 attach_type; 584 __u32 attach_flags; 585 __u32 replace_bpf_fd; /* previously attached eBPF 586 * program to replace if 587 * BPF_F_REPLACE is used 588 */ 589 }; 590 591 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 592 __u32 prog_fd; 593 __u32 retval; 594 __u32 data_size_in; /* input: len of data_in */ 595 __u32 data_size_out; /* input/output: len of data_out 596 * returns ENOSPC if data_out 597 * is too small. 598 */ 599 __aligned_u64 data_in; 600 __aligned_u64 data_out; 601 __u32 repeat; 602 __u32 duration; 603 __u32 ctx_size_in; /* input: len of ctx_in */ 604 __u32 ctx_size_out; /* input/output: len of ctx_out 605 * returns ENOSPC if ctx_out 606 * is too small. 607 */ 608 __aligned_u64 ctx_in; 609 __aligned_u64 ctx_out; 610 __u32 flags; 611 __u32 cpu; 612 } test; 613 614 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 615 union { 616 __u32 start_id; 617 __u32 prog_id; 618 __u32 map_id; 619 __u32 btf_id; 620 __u32 link_id; 621 }; 622 __u32 next_id; 623 __u32 open_flags; 624 }; 625 626 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 627 __u32 bpf_fd; 628 __u32 info_len; 629 __aligned_u64 info; 630 } info; 631 632 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 633 __u32 target_fd; /* container object to query */ 634 __u32 attach_type; 635 __u32 query_flags; 636 __u32 attach_flags; 637 __aligned_u64 prog_ids; 638 __u32 prog_cnt; 639 } query; 640 641 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 642 __u64 name; 643 __u32 prog_fd; 644 } raw_tracepoint; 645 646 struct { /* anonymous struct for BPF_BTF_LOAD */ 647 __aligned_u64 btf; 648 __aligned_u64 btf_log_buf; 649 __u32 btf_size; 650 __u32 btf_log_size; 651 __u32 btf_log_level; 652 }; 653 654 struct { 655 __u32 pid; /* input: pid */ 656 __u32 fd; /* input: fd */ 657 __u32 flags; /* input: flags */ 658 __u32 buf_len; /* input/output: buf len */ 659 __aligned_u64 buf; /* input/output: 660 * tp_name for tracepoint 661 * symbol for kprobe 662 * filename for uprobe 663 */ 664 __u32 prog_id; /* output: prod_id */ 665 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 666 __u64 probe_offset; /* output: probe_offset */ 667 __u64 probe_addr; /* output: probe_addr */ 668 } task_fd_query; 669 670 struct { /* struct used by BPF_LINK_CREATE command */ 671 __u32 prog_fd; /* eBPF program to attach */ 672 union { 673 __u32 target_fd; /* object to attach to */ 674 __u32 target_ifindex; /* target ifindex */ 675 }; 676 __u32 attach_type; /* attach type */ 677 __u32 flags; /* extra flags */ 678 union { 679 __u32 target_btf_id; /* btf_id of target to attach to */ 680 struct { 681 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 682 __u32 iter_info_len; /* iter_info length */ 683 }; 684 }; 685 } link_create; 686 687 struct { /* struct used by BPF_LINK_UPDATE command */ 688 __u32 link_fd; /* link fd */ 689 /* new program fd to update link with */ 690 __u32 new_prog_fd; 691 __u32 flags; /* extra flags */ 692 /* expected link's program fd; is specified only if 693 * BPF_F_REPLACE flag is set in flags */ 694 __u32 old_prog_fd; 695 } link_update; 696 697 struct { 698 __u32 link_fd; 699 } link_detach; 700 701 struct { /* struct used by BPF_ENABLE_STATS command */ 702 __u32 type; 703 } enable_stats; 704 705 struct { /* struct used by BPF_ITER_CREATE command */ 706 __u32 link_fd; 707 __u32 flags; 708 } iter_create; 709 710 struct { /* struct used by BPF_PROG_BIND_MAP command */ 711 __u32 prog_fd; 712 __u32 map_fd; 713 __u32 flags; /* extra flags */ 714 } prog_bind_map; 715 716 } __attribute__((aligned(8))); 717 718 /* The description below is an attempt at providing documentation to eBPF 719 * developers about the multiple available eBPF helper functions. It can be 720 * parsed and used to produce a manual page. The workflow is the following, 721 * and requires the rst2man utility: 722 * 723 * $ ./scripts/bpf_helpers_doc.py \ 724 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 725 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 726 * $ man /tmp/bpf-helpers.7 727 * 728 * Note that in order to produce this external documentation, some RST 729 * formatting is used in the descriptions to get "bold" and "italics" in 730 * manual pages. Also note that the few trailing white spaces are 731 * intentional, removing them would break paragraphs for rst2man. 732 * 733 * Start of BPF helper function descriptions: 734 * 735 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 736 * Description 737 * Perform a lookup in *map* for an entry associated to *key*. 738 * Return 739 * Map value associated to *key*, or **NULL** if no entry was 740 * found. 741 * 742 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 743 * Description 744 * Add or update the value of the entry associated to *key* in 745 * *map* with *value*. *flags* is one of: 746 * 747 * **BPF_NOEXIST** 748 * The entry for *key* must not exist in the map. 749 * **BPF_EXIST** 750 * The entry for *key* must already exist in the map. 751 * **BPF_ANY** 752 * No condition on the existence of the entry for *key*. 753 * 754 * Flag value **BPF_NOEXIST** cannot be used for maps of types 755 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 756 * elements always exist), the helper would return an error. 757 * Return 758 * 0 on success, or a negative error in case of failure. 759 * 760 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 761 * Description 762 * Delete entry with *key* from *map*. 763 * Return 764 * 0 on success, or a negative error in case of failure. 765 * 766 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 767 * Description 768 * For tracing programs, safely attempt to read *size* bytes from 769 * kernel space address *unsafe_ptr* and store the data in *dst*. 770 * 771 * Generally, use **bpf_probe_read_user**\ () or 772 * **bpf_probe_read_kernel**\ () instead. 773 * Return 774 * 0 on success, or a negative error in case of failure. 775 * 776 * u64 bpf_ktime_get_ns(void) 777 * Description 778 * Return the time elapsed since system boot, in nanoseconds. 779 * Does not include time the system was suspended. 780 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 781 * Return 782 * Current *ktime*. 783 * 784 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 785 * Description 786 * This helper is a "printk()-like" facility for debugging. It 787 * prints a message defined by format *fmt* (of size *fmt_size*) 788 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 789 * available. It can take up to three additional **u64** 790 * arguments (as an eBPF helpers, the total number of arguments is 791 * limited to five). 792 * 793 * Each time the helper is called, it appends a line to the trace. 794 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 795 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 796 * The format of the trace is customizable, and the exact output 797 * one will get depends on the options set in 798 * *\/sys/kernel/debug/tracing/trace_options* (see also the 799 * *README* file under the same directory). However, it usually 800 * defaults to something like: 801 * 802 * :: 803 * 804 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 805 * 806 * In the above: 807 * 808 * * ``telnet`` is the name of the current task. 809 * * ``470`` is the PID of the current task. 810 * * ``001`` is the CPU number on which the task is 811 * running. 812 * * In ``.N..``, each character refers to a set of 813 * options (whether irqs are enabled, scheduling 814 * options, whether hard/softirqs are running, level of 815 * preempt_disabled respectively). **N** means that 816 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 817 * are set. 818 * * ``419421.045894`` is a timestamp. 819 * * ``0x00000001`` is a fake value used by BPF for the 820 * instruction pointer register. 821 * * ``<formatted msg>`` is the message formatted with 822 * *fmt*. 823 * 824 * The conversion specifiers supported by *fmt* are similar, but 825 * more limited than for printk(). They are **%d**, **%i**, 826 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 827 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 828 * of field, padding with zeroes, etc.) is available, and the 829 * helper will return **-EINVAL** (but print nothing) if it 830 * encounters an unknown specifier. 831 * 832 * Also, note that **bpf_trace_printk**\ () is slow, and should 833 * only be used for debugging purposes. For this reason, a notice 834 * block (spanning several lines) is printed to kernel logs and 835 * states that the helper should not be used "for production use" 836 * the first time this helper is used (or more precisely, when 837 * **trace_printk**\ () buffers are allocated). For passing values 838 * to user space, perf events should be preferred. 839 * Return 840 * The number of bytes written to the buffer, or a negative error 841 * in case of failure. 842 * 843 * u32 bpf_get_prandom_u32(void) 844 * Description 845 * Get a pseudo-random number. 846 * 847 * From a security point of view, this helper uses its own 848 * pseudo-random internal state, and cannot be used to infer the 849 * seed of other random functions in the kernel. However, it is 850 * essential to note that the generator used by the helper is not 851 * cryptographically secure. 852 * Return 853 * A random 32-bit unsigned value. 854 * 855 * u32 bpf_get_smp_processor_id(void) 856 * Description 857 * Get the SMP (symmetric multiprocessing) processor id. Note that 858 * all programs run with preemption disabled, which means that the 859 * SMP processor id is stable during all the execution of the 860 * program. 861 * Return 862 * The SMP id of the processor running the program. 863 * 864 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 865 * Description 866 * Store *len* bytes from address *from* into the packet 867 * associated to *skb*, at *offset*. *flags* are a combination of 868 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 869 * checksum for the packet after storing the bytes) and 870 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 871 * **->swhash** and *skb*\ **->l4hash** to 0). 872 * 873 * A call to this helper is susceptible to change the underlying 874 * packet buffer. Therefore, at load time, all checks on pointers 875 * previously done by the verifier are invalidated and must be 876 * performed again, if the helper is used in combination with 877 * direct packet access. 878 * Return 879 * 0 on success, or a negative error in case of failure. 880 * 881 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 882 * Description 883 * Recompute the layer 3 (e.g. IP) checksum for the packet 884 * associated to *skb*. Computation is incremental, so the helper 885 * must know the former value of the header field that was 886 * modified (*from*), the new value of this field (*to*), and the 887 * number of bytes (2 or 4) for this field, stored in *size*. 888 * Alternatively, it is possible to store the difference between 889 * the previous and the new values of the header field in *to*, by 890 * setting *from* and *size* to 0. For both methods, *offset* 891 * indicates the location of the IP checksum within the packet. 892 * 893 * This helper works in combination with **bpf_csum_diff**\ (), 894 * which does not update the checksum in-place, but offers more 895 * flexibility and can handle sizes larger than 2 or 4 for the 896 * checksum to update. 897 * 898 * A call to this helper is susceptible to change the underlying 899 * packet buffer. Therefore, at load time, all checks on pointers 900 * previously done by the verifier are invalidated and must be 901 * performed again, if the helper is used in combination with 902 * direct packet access. 903 * Return 904 * 0 on success, or a negative error in case of failure. 905 * 906 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 907 * Description 908 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 909 * packet associated to *skb*. Computation is incremental, so the 910 * helper must know the former value of the header field that was 911 * modified (*from*), the new value of this field (*to*), and the 912 * number of bytes (2 or 4) for this field, stored on the lowest 913 * four bits of *flags*. Alternatively, it is possible to store 914 * the difference between the previous and the new values of the 915 * header field in *to*, by setting *from* and the four lowest 916 * bits of *flags* to 0. For both methods, *offset* indicates the 917 * location of the IP checksum within the packet. In addition to 918 * the size of the field, *flags* can be added (bitwise OR) actual 919 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 920 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 921 * for updates resulting in a null checksum the value is set to 922 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 923 * the checksum is to be computed against a pseudo-header. 924 * 925 * This helper works in combination with **bpf_csum_diff**\ (), 926 * which does not update the checksum in-place, but offers more 927 * flexibility and can handle sizes larger than 2 or 4 for the 928 * checksum to update. 929 * 930 * A call to this helper is susceptible to change the underlying 931 * packet buffer. Therefore, at load time, all checks on pointers 932 * previously done by the verifier are invalidated and must be 933 * performed again, if the helper is used in combination with 934 * direct packet access. 935 * Return 936 * 0 on success, or a negative error in case of failure. 937 * 938 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 939 * Description 940 * This special helper is used to trigger a "tail call", or in 941 * other words, to jump into another eBPF program. The same stack 942 * frame is used (but values on stack and in registers for the 943 * caller are not accessible to the callee). This mechanism allows 944 * for program chaining, either for raising the maximum number of 945 * available eBPF instructions, or to execute given programs in 946 * conditional blocks. For security reasons, there is an upper 947 * limit to the number of successive tail calls that can be 948 * performed. 949 * 950 * Upon call of this helper, the program attempts to jump into a 951 * program referenced at index *index* in *prog_array_map*, a 952 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 953 * *ctx*, a pointer to the context. 954 * 955 * If the call succeeds, the kernel immediately runs the first 956 * instruction of the new program. This is not a function call, 957 * and it never returns to the previous program. If the call 958 * fails, then the helper has no effect, and the caller continues 959 * to run its subsequent instructions. A call can fail if the 960 * destination program for the jump does not exist (i.e. *index* 961 * is superior to the number of entries in *prog_array_map*), or 962 * if the maximum number of tail calls has been reached for this 963 * chain of programs. This limit is defined in the kernel by the 964 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 965 * which is currently set to 32. 966 * Return 967 * 0 on success, or a negative error in case of failure. 968 * 969 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 970 * Description 971 * Clone and redirect the packet associated to *skb* to another 972 * net device of index *ifindex*. Both ingress and egress 973 * interfaces can be used for redirection. The **BPF_F_INGRESS** 974 * value in *flags* is used to make the distinction (ingress path 975 * is selected if the flag is present, egress path otherwise). 976 * This is the only flag supported for now. 977 * 978 * In comparison with **bpf_redirect**\ () helper, 979 * **bpf_clone_redirect**\ () has the associated cost of 980 * duplicating the packet buffer, but this can be executed out of 981 * the eBPF program. Conversely, **bpf_redirect**\ () is more 982 * efficient, but it is handled through an action code where the 983 * redirection happens only after the eBPF program has returned. 984 * 985 * A call to this helper is susceptible to change the underlying 986 * packet buffer. Therefore, at load time, all checks on pointers 987 * previously done by the verifier are invalidated and must be 988 * performed again, if the helper is used in combination with 989 * direct packet access. 990 * Return 991 * 0 on success, or a negative error in case of failure. 992 * 993 * u64 bpf_get_current_pid_tgid(void) 994 * Return 995 * A 64-bit integer containing the current tgid and pid, and 996 * created as such: 997 * *current_task*\ **->tgid << 32 \|** 998 * *current_task*\ **->pid**. 999 * 1000 * u64 bpf_get_current_uid_gid(void) 1001 * Return 1002 * A 64-bit integer containing the current GID and UID, and 1003 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1004 * 1005 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1006 * Description 1007 * Copy the **comm** attribute of the current task into *buf* of 1008 * *size_of_buf*. The **comm** attribute contains the name of 1009 * the executable (excluding the path) for the current task. The 1010 * *size_of_buf* must be strictly positive. On success, the 1011 * helper makes sure that the *buf* is NUL-terminated. On failure, 1012 * it is filled with zeroes. 1013 * Return 1014 * 0 on success, or a negative error in case of failure. 1015 * 1016 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1017 * Description 1018 * Retrieve the classid for the current task, i.e. for the net_cls 1019 * cgroup to which *skb* belongs. 1020 * 1021 * This helper can be used on TC egress path, but not on ingress. 1022 * 1023 * The net_cls cgroup provides an interface to tag network packets 1024 * based on a user-provided identifier for all traffic coming from 1025 * the tasks belonging to the related cgroup. See also the related 1026 * kernel documentation, available from the Linux sources in file 1027 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1028 * 1029 * The Linux kernel has two versions for cgroups: there are 1030 * cgroups v1 and cgroups v2. Both are available to users, who can 1031 * use a mixture of them, but note that the net_cls cgroup is for 1032 * cgroup v1 only. This makes it incompatible with BPF programs 1033 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1034 * only hold data for one version of cgroups at a time). 1035 * 1036 * This helper is only available is the kernel was compiled with 1037 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1038 * "**y**" or to "**m**". 1039 * Return 1040 * The classid, or 0 for the default unconfigured classid. 1041 * 1042 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1043 * Description 1044 * Push a *vlan_tci* (VLAN tag control information) of protocol 1045 * *vlan_proto* to the packet associated to *skb*, then update 1046 * the checksum. Note that if *vlan_proto* is different from 1047 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1048 * be **ETH_P_8021Q**. 1049 * 1050 * A call to this helper is susceptible to change the underlying 1051 * packet buffer. Therefore, at load time, all checks on pointers 1052 * previously done by the verifier are invalidated and must be 1053 * performed again, if the helper is used in combination with 1054 * direct packet access. 1055 * Return 1056 * 0 on success, or a negative error in case of failure. 1057 * 1058 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1059 * Description 1060 * Pop a VLAN header from the packet associated to *skb*. 1061 * 1062 * A call to this helper is susceptible to change the underlying 1063 * packet buffer. Therefore, at load time, all checks on pointers 1064 * previously done by the verifier are invalidated and must be 1065 * performed again, if the helper is used in combination with 1066 * direct packet access. 1067 * Return 1068 * 0 on success, or a negative error in case of failure. 1069 * 1070 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1071 * Description 1072 * Get tunnel metadata. This helper takes a pointer *key* to an 1073 * empty **struct bpf_tunnel_key** of **size**, that will be 1074 * filled with tunnel metadata for the packet associated to *skb*. 1075 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1076 * indicates that the tunnel is based on IPv6 protocol instead of 1077 * IPv4. 1078 * 1079 * The **struct bpf_tunnel_key** is an object that generalizes the 1080 * principal parameters used by various tunneling protocols into a 1081 * single struct. This way, it can be used to easily make a 1082 * decision based on the contents of the encapsulation header, 1083 * "summarized" in this struct. In particular, it holds the IP 1084 * address of the remote end (IPv4 or IPv6, depending on the case) 1085 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1086 * this struct exposes the *key*\ **->tunnel_id**, which is 1087 * generally mapped to a VNI (Virtual Network Identifier), making 1088 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1089 * () helper. 1090 * 1091 * Let's imagine that the following code is part of a program 1092 * attached to the TC ingress interface, on one end of a GRE 1093 * tunnel, and is supposed to filter out all messages coming from 1094 * remote ends with IPv4 address other than 10.0.0.1: 1095 * 1096 * :: 1097 * 1098 * int ret; 1099 * struct bpf_tunnel_key key = {}; 1100 * 1101 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1102 * if (ret < 0) 1103 * return TC_ACT_SHOT; // drop packet 1104 * 1105 * if (key.remote_ipv4 != 0x0a000001) 1106 * return TC_ACT_SHOT; // drop packet 1107 * 1108 * return TC_ACT_OK; // accept packet 1109 * 1110 * This interface can also be used with all encapsulation devices 1111 * that can operate in "collect metadata" mode: instead of having 1112 * one network device per specific configuration, the "collect 1113 * metadata" mode only requires a single device where the 1114 * configuration can be extracted from this helper. 1115 * 1116 * This can be used together with various tunnels such as VXLan, 1117 * Geneve, GRE or IP in IP (IPIP). 1118 * Return 1119 * 0 on success, or a negative error in case of failure. 1120 * 1121 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1122 * Description 1123 * Populate tunnel metadata for packet associated to *skb.* The 1124 * tunnel metadata is set to the contents of *key*, of *size*. The 1125 * *flags* can be set to a combination of the following values: 1126 * 1127 * **BPF_F_TUNINFO_IPV6** 1128 * Indicate that the tunnel is based on IPv6 protocol 1129 * instead of IPv4. 1130 * **BPF_F_ZERO_CSUM_TX** 1131 * For IPv4 packets, add a flag to tunnel metadata 1132 * indicating that checksum computation should be skipped 1133 * and checksum set to zeroes. 1134 * **BPF_F_DONT_FRAGMENT** 1135 * Add a flag to tunnel metadata indicating that the 1136 * packet should not be fragmented. 1137 * **BPF_F_SEQ_NUMBER** 1138 * Add a flag to tunnel metadata indicating that a 1139 * sequence number should be added to tunnel header before 1140 * sending the packet. This flag was added for GRE 1141 * encapsulation, but might be used with other protocols 1142 * as well in the future. 1143 * 1144 * Here is a typical usage on the transmit path: 1145 * 1146 * :: 1147 * 1148 * struct bpf_tunnel_key key; 1149 * populate key ... 1150 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1151 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1152 * 1153 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1154 * helper for additional information. 1155 * Return 1156 * 0 on success, or a negative error in case of failure. 1157 * 1158 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1159 * Description 1160 * Read the value of a perf event counter. This helper relies on a 1161 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1162 * the perf event counter is selected when *map* is updated with 1163 * perf event file descriptors. The *map* is an array whose size 1164 * is the number of available CPUs, and each cell contains a value 1165 * relative to one CPU. The value to retrieve is indicated by 1166 * *flags*, that contains the index of the CPU to look up, masked 1167 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1168 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1169 * current CPU should be retrieved. 1170 * 1171 * Note that before Linux 4.13, only hardware perf event can be 1172 * retrieved. 1173 * 1174 * Also, be aware that the newer helper 1175 * **bpf_perf_event_read_value**\ () is recommended over 1176 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1177 * quirks where error and counter value are used as a return code 1178 * (which is wrong to do since ranges may overlap). This issue is 1179 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1180 * time provides more features over the **bpf_perf_event_read**\ 1181 * () interface. Please refer to the description of 1182 * **bpf_perf_event_read_value**\ () for details. 1183 * Return 1184 * The value of the perf event counter read from the map, or a 1185 * negative error code in case of failure. 1186 * 1187 * long bpf_redirect(u32 ifindex, u64 flags) 1188 * Description 1189 * Redirect the packet to another net device of index *ifindex*. 1190 * This helper is somewhat similar to **bpf_clone_redirect**\ 1191 * (), except that the packet is not cloned, which provides 1192 * increased performance. 1193 * 1194 * Except for XDP, both ingress and egress interfaces can be used 1195 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1196 * to make the distinction (ingress path is selected if the flag 1197 * is present, egress path otherwise). Currently, XDP only 1198 * supports redirection to the egress interface, and accepts no 1199 * flag at all. 1200 * 1201 * The same effect can also be attained with the more generic 1202 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1203 * redirect target instead of providing it directly to the helper. 1204 * Return 1205 * For XDP, the helper returns **XDP_REDIRECT** on success or 1206 * **XDP_ABORTED** on error. For other program types, the values 1207 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1208 * error. 1209 * 1210 * u32 bpf_get_route_realm(struct sk_buff *skb) 1211 * Description 1212 * Retrieve the realm or the route, that is to say the 1213 * **tclassid** field of the destination for the *skb*. The 1214 * identifier retrieved is a user-provided tag, similar to the 1215 * one used with the net_cls cgroup (see description for 1216 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1217 * held by a route (a destination entry), not by a task. 1218 * 1219 * Retrieving this identifier works with the clsact TC egress hook 1220 * (see also **tc-bpf(8)**), or alternatively on conventional 1221 * classful egress qdiscs, but not on TC ingress path. In case of 1222 * clsact TC egress hook, this has the advantage that, internally, 1223 * the destination entry has not been dropped yet in the transmit 1224 * path. Therefore, the destination entry does not need to be 1225 * artificially held via **netif_keep_dst**\ () for a classful 1226 * qdisc until the *skb* is freed. 1227 * 1228 * This helper is available only if the kernel was compiled with 1229 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1230 * Return 1231 * The realm of the route for the packet associated to *skb*, or 0 1232 * if none was found. 1233 * 1234 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1235 * Description 1236 * Write raw *data* blob into a special BPF perf event held by 1237 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1238 * event must have the following attributes: **PERF_SAMPLE_RAW** 1239 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1240 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1241 * 1242 * The *flags* are used to indicate the index in *map* for which 1243 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1244 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1245 * to indicate that the index of the current CPU core should be 1246 * used. 1247 * 1248 * The value to write, of *size*, is passed through eBPF stack and 1249 * pointed by *data*. 1250 * 1251 * The context of the program *ctx* needs also be passed to the 1252 * helper. 1253 * 1254 * On user space, a program willing to read the values needs to 1255 * call **perf_event_open**\ () on the perf event (either for 1256 * one or for all CPUs) and to store the file descriptor into the 1257 * *map*. This must be done before the eBPF program can send data 1258 * into it. An example is available in file 1259 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1260 * tree (the eBPF program counterpart is in 1261 * *samples/bpf/trace_output_kern.c*). 1262 * 1263 * **bpf_perf_event_output**\ () achieves better performance 1264 * than **bpf_trace_printk**\ () for sharing data with user 1265 * space, and is much better suitable for streaming data from eBPF 1266 * programs. 1267 * 1268 * Note that this helper is not restricted to tracing use cases 1269 * and can be used with programs attached to TC or XDP as well, 1270 * where it allows for passing data to user space listeners. Data 1271 * can be: 1272 * 1273 * * Only custom structs, 1274 * * Only the packet payload, or 1275 * * A combination of both. 1276 * Return 1277 * 0 on success, or a negative error in case of failure. 1278 * 1279 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1280 * Description 1281 * This helper was provided as an easy way to load data from a 1282 * packet. It can be used to load *len* bytes from *offset* from 1283 * the packet associated to *skb*, into the buffer pointed by 1284 * *to*. 1285 * 1286 * Since Linux 4.7, usage of this helper has mostly been replaced 1287 * by "direct packet access", enabling packet data to be 1288 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1289 * pointing respectively to the first byte of packet data and to 1290 * the byte after the last byte of packet data. However, it 1291 * remains useful if one wishes to read large quantities of data 1292 * at once from a packet into the eBPF stack. 1293 * Return 1294 * 0 on success, or a negative error in case of failure. 1295 * 1296 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1297 * Description 1298 * Walk a user or a kernel stack and return its id. To achieve 1299 * this, the helper needs *ctx*, which is a pointer to the context 1300 * on which the tracing program is executed, and a pointer to a 1301 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1302 * 1303 * The last argument, *flags*, holds the number of stack frames to 1304 * skip (from 0 to 255), masked with 1305 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1306 * a combination of the following flags: 1307 * 1308 * **BPF_F_USER_STACK** 1309 * Collect a user space stack instead of a kernel stack. 1310 * **BPF_F_FAST_STACK_CMP** 1311 * Compare stacks by hash only. 1312 * **BPF_F_REUSE_STACKID** 1313 * If two different stacks hash into the same *stackid*, 1314 * discard the old one. 1315 * 1316 * The stack id retrieved is a 32 bit long integer handle which 1317 * can be further combined with other data (including other stack 1318 * ids) and used as a key into maps. This can be useful for 1319 * generating a variety of graphs (such as flame graphs or off-cpu 1320 * graphs). 1321 * 1322 * For walking a stack, this helper is an improvement over 1323 * **bpf_probe_read**\ (), which can be used with unrolled loops 1324 * but is not efficient and consumes a lot of eBPF instructions. 1325 * Instead, **bpf_get_stackid**\ () can collect up to 1326 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1327 * this limit can be controlled with the **sysctl** program, and 1328 * that it should be manually increased in order to profile long 1329 * user stacks (such as stacks for Java programs). To do so, use: 1330 * 1331 * :: 1332 * 1333 * # sysctl kernel.perf_event_max_stack=<new value> 1334 * Return 1335 * The positive or null stack id on success, or a negative error 1336 * in case of failure. 1337 * 1338 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1339 * Description 1340 * Compute a checksum difference, from the raw buffer pointed by 1341 * *from*, of length *from_size* (that must be a multiple of 4), 1342 * towards the raw buffer pointed by *to*, of size *to_size* 1343 * (same remark). An optional *seed* can be added to the value 1344 * (this can be cascaded, the seed may come from a previous call 1345 * to the helper). 1346 * 1347 * This is flexible enough to be used in several ways: 1348 * 1349 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1350 * checksum, it can be used when pushing new data. 1351 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1352 * checksum, it can be used when removing data from a packet. 1353 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1354 * can be used to compute a diff. Note that *from_size* and 1355 * *to_size* do not need to be equal. 1356 * 1357 * This helper can be used in combination with 1358 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1359 * which one can feed in the difference computed with 1360 * **bpf_csum_diff**\ (). 1361 * Return 1362 * The checksum result, or a negative error code in case of 1363 * failure. 1364 * 1365 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1366 * Description 1367 * Retrieve tunnel options metadata for the packet associated to 1368 * *skb*, and store the raw tunnel option data to the buffer *opt* 1369 * of *size*. 1370 * 1371 * This helper can be used with encapsulation devices that can 1372 * operate in "collect metadata" mode (please refer to the related 1373 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1374 * more details). A particular example where this can be used is 1375 * in combination with the Geneve encapsulation protocol, where it 1376 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1377 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1378 * the eBPF program. This allows for full customization of these 1379 * headers. 1380 * Return 1381 * The size of the option data retrieved. 1382 * 1383 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1384 * Description 1385 * Set tunnel options metadata for the packet associated to *skb* 1386 * to the option data contained in the raw buffer *opt* of *size*. 1387 * 1388 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1389 * helper for additional information. 1390 * Return 1391 * 0 on success, or a negative error in case of failure. 1392 * 1393 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1394 * Description 1395 * Change the protocol of the *skb* to *proto*. Currently 1396 * supported are transition from IPv4 to IPv6, and from IPv6 to 1397 * IPv4. The helper takes care of the groundwork for the 1398 * transition, including resizing the socket buffer. The eBPF 1399 * program is expected to fill the new headers, if any, via 1400 * **skb_store_bytes**\ () and to recompute the checksums with 1401 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1402 * (). The main case for this helper is to perform NAT64 1403 * operations out of an eBPF program. 1404 * 1405 * Internally, the GSO type is marked as dodgy so that headers are 1406 * checked and segments are recalculated by the GSO/GRO engine. 1407 * The size for GSO target is adapted as well. 1408 * 1409 * All values for *flags* are reserved for future usage, and must 1410 * be left at zero. 1411 * 1412 * A call to this helper is susceptible to change the underlying 1413 * packet buffer. Therefore, at load time, all checks on pointers 1414 * previously done by the verifier are invalidated and must be 1415 * performed again, if the helper is used in combination with 1416 * direct packet access. 1417 * Return 1418 * 0 on success, or a negative error in case of failure. 1419 * 1420 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 1421 * Description 1422 * Change the packet type for the packet associated to *skb*. This 1423 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1424 * the eBPF program does not have a write access to *skb*\ 1425 * **->pkt_type** beside this helper. Using a helper here allows 1426 * for graceful handling of errors. 1427 * 1428 * The major use case is to change incoming *skb*s to 1429 * **PACKET_HOST** in a programmatic way instead of having to 1430 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1431 * example. 1432 * 1433 * Note that *type* only allows certain values. At this time, they 1434 * are: 1435 * 1436 * **PACKET_HOST** 1437 * Packet is for us. 1438 * **PACKET_BROADCAST** 1439 * Send packet to all. 1440 * **PACKET_MULTICAST** 1441 * Send packet to group. 1442 * **PACKET_OTHERHOST** 1443 * Send packet to someone else. 1444 * Return 1445 * 0 on success, or a negative error in case of failure. 1446 * 1447 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1448 * Description 1449 * Check whether *skb* is a descendant of the cgroup2 held by 1450 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1451 * Return 1452 * The return value depends on the result of the test, and can be: 1453 * 1454 * * 0, if the *skb* failed the cgroup2 descendant test. 1455 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1456 * * A negative error code, if an error occurred. 1457 * 1458 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1459 * Description 1460 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1461 * not set, in particular if the hash was cleared due to mangling, 1462 * recompute this hash. Later accesses to the hash can be done 1463 * directly with *skb*\ **->hash**. 1464 * 1465 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1466 * prototype with **bpf_skb_change_proto**\ (), or calling 1467 * **bpf_skb_store_bytes**\ () with the 1468 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1469 * the hash and to trigger a new computation for the next call to 1470 * **bpf_get_hash_recalc**\ (). 1471 * Return 1472 * The 32-bit hash. 1473 * 1474 * u64 bpf_get_current_task(void) 1475 * Return 1476 * A pointer to the current task struct. 1477 * 1478 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 1479 * Description 1480 * Attempt in a safe way to write *len* bytes from the buffer 1481 * *src* to *dst* in memory. It only works for threads that are in 1482 * user context, and *dst* must be a valid user space address. 1483 * 1484 * This helper should not be used to implement any kind of 1485 * security mechanism because of TOC-TOU attacks, but rather to 1486 * debug, divert, and manipulate execution of semi-cooperative 1487 * processes. 1488 * 1489 * Keep in mind that this feature is meant for experiments, and it 1490 * has a risk of crashing the system and running programs. 1491 * Therefore, when an eBPF program using this helper is attached, 1492 * a warning including PID and process name is printed to kernel 1493 * logs. 1494 * Return 1495 * 0 on success, or a negative error in case of failure. 1496 * 1497 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1498 * Description 1499 * Check whether the probe is being run is the context of a given 1500 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1501 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1502 * Return 1503 * The return value depends on the result of the test, and can be: 1504 * 1505 * * 0, if current task belongs to the cgroup2. 1506 * * 1, if current task does not belong to the cgroup2. 1507 * * A negative error code, if an error occurred. 1508 * 1509 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1510 * Description 1511 * Resize (trim or grow) the packet associated to *skb* to the 1512 * new *len*. The *flags* are reserved for future usage, and must 1513 * be left at zero. 1514 * 1515 * The basic idea is that the helper performs the needed work to 1516 * change the size of the packet, then the eBPF program rewrites 1517 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1518 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1519 * and others. This helper is a slow path utility intended for 1520 * replies with control messages. And because it is targeted for 1521 * slow path, the helper itself can afford to be slow: it 1522 * implicitly linearizes, unclones and drops offloads from the 1523 * *skb*. 1524 * 1525 * A call to this helper is susceptible to change the underlying 1526 * packet buffer. Therefore, at load time, all checks on pointers 1527 * previously done by the verifier are invalidated and must be 1528 * performed again, if the helper is used in combination with 1529 * direct packet access. 1530 * Return 1531 * 0 on success, or a negative error in case of failure. 1532 * 1533 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1534 * Description 1535 * Pull in non-linear data in case the *skb* is non-linear and not 1536 * all of *len* are part of the linear section. Make *len* bytes 1537 * from *skb* readable and writable. If a zero value is passed for 1538 * *len*, then the whole length of the *skb* is pulled. 1539 * 1540 * This helper is only needed for reading and writing with direct 1541 * packet access. 1542 * 1543 * For direct packet access, testing that offsets to access 1544 * are within packet boundaries (test on *skb*\ **->data_end**) is 1545 * susceptible to fail if offsets are invalid, or if the requested 1546 * data is in non-linear parts of the *skb*. On failure the 1547 * program can just bail out, or in the case of a non-linear 1548 * buffer, use a helper to make the data available. The 1549 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1550 * the data. Another one consists in using **bpf_skb_pull_data** 1551 * to pull in once the non-linear parts, then retesting and 1552 * eventually access the data. 1553 * 1554 * At the same time, this also makes sure the *skb* is uncloned, 1555 * which is a necessary condition for direct write. As this needs 1556 * to be an invariant for the write part only, the verifier 1557 * detects writes and adds a prologue that is calling 1558 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1559 * the very beginning in case it is indeed cloned. 1560 * 1561 * A call to this helper is susceptible to change the underlying 1562 * packet buffer. Therefore, at load time, all checks on pointers 1563 * previously done by the verifier are invalidated and must be 1564 * performed again, if the helper is used in combination with 1565 * direct packet access. 1566 * Return 1567 * 0 on success, or a negative error in case of failure. 1568 * 1569 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1570 * Description 1571 * Add the checksum *csum* into *skb*\ **->csum** in case the 1572 * driver has supplied a checksum for the entire packet into that 1573 * field. Return an error otherwise. This helper is intended to be 1574 * used in combination with **bpf_csum_diff**\ (), in particular 1575 * when the checksum needs to be updated after data has been 1576 * written into the packet through direct packet access. 1577 * Return 1578 * The checksum on success, or a negative error code in case of 1579 * failure. 1580 * 1581 * void bpf_set_hash_invalid(struct sk_buff *skb) 1582 * Description 1583 * Invalidate the current *skb*\ **->hash**. It can be used after 1584 * mangling on headers through direct packet access, in order to 1585 * indicate that the hash is outdated and to trigger a 1586 * recalculation the next time the kernel tries to access this 1587 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1588 * 1589 * long bpf_get_numa_node_id(void) 1590 * Description 1591 * Return the id of the current NUMA node. The primary use case 1592 * for this helper is the selection of sockets for the local NUMA 1593 * node, when the program is attached to sockets using the 1594 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1595 * but the helper is also available to other eBPF program types, 1596 * similarly to **bpf_get_smp_processor_id**\ (). 1597 * Return 1598 * The id of current NUMA node. 1599 * 1600 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1601 * Description 1602 * Grows headroom of packet associated to *skb* and adjusts the 1603 * offset of the MAC header accordingly, adding *len* bytes of 1604 * space. It automatically extends and reallocates memory as 1605 * required. 1606 * 1607 * This helper can be used on a layer 3 *skb* to push a MAC header 1608 * for redirection into a layer 2 device. 1609 * 1610 * All values for *flags* are reserved for future usage, and must 1611 * be left at zero. 1612 * 1613 * A call to this helper is susceptible to change the underlying 1614 * packet buffer. Therefore, at load time, all checks on pointers 1615 * previously done by the verifier are invalidated and must be 1616 * performed again, if the helper is used in combination with 1617 * direct packet access. 1618 * Return 1619 * 0 on success, or a negative error in case of failure. 1620 * 1621 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1622 * Description 1623 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1624 * it is possible to use a negative value for *delta*. This helper 1625 * can be used to prepare the packet for pushing or popping 1626 * headers. 1627 * 1628 * A call to this helper is susceptible to change the underlying 1629 * packet buffer. Therefore, at load time, all checks on pointers 1630 * previously done by the verifier are invalidated and must be 1631 * performed again, if the helper is used in combination with 1632 * direct packet access. 1633 * Return 1634 * 0 on success, or a negative error in case of failure. 1635 * 1636 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1637 * Description 1638 * Copy a NUL terminated string from an unsafe kernel address 1639 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1640 * more details. 1641 * 1642 * Generally, use **bpf_probe_read_user_str**\ () or 1643 * **bpf_probe_read_kernel_str**\ () instead. 1644 * Return 1645 * On success, the strictly positive length of the string, 1646 * including the trailing NUL character. On error, a negative 1647 * value. 1648 * 1649 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1650 * Description 1651 * If the **struct sk_buff** pointed by *skb* has a known socket, 1652 * retrieve the cookie (generated by the kernel) of this socket. 1653 * If no cookie has been set yet, generate a new cookie. Once 1654 * generated, the socket cookie remains stable for the life of the 1655 * socket. This helper can be useful for monitoring per socket 1656 * networking traffic statistics as it provides a global socket 1657 * identifier that can be assumed unique. 1658 * Return 1659 * A 8-byte long non-decreasing number on success, or 0 if the 1660 * socket field is missing inside *skb*. 1661 * 1662 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1663 * Description 1664 * Equivalent to bpf_get_socket_cookie() helper that accepts 1665 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1666 * Return 1667 * A 8-byte long non-decreasing number. 1668 * 1669 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1670 * Description 1671 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 1672 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1673 * Return 1674 * A 8-byte long non-decreasing number. 1675 * 1676 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1677 * Return 1678 * The owner UID of the socket associated to *skb*. If the socket 1679 * is **NULL**, or if it is not a full socket (i.e. if it is a 1680 * time-wait or a request socket instead), **overflowuid** value 1681 * is returned (note that **overflowuid** might also be the actual 1682 * UID value for the socket). 1683 * 1684 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 1685 * Description 1686 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1687 * to value *hash*. 1688 * Return 1689 * 0 1690 * 1691 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1692 * Description 1693 * Emulate a call to **setsockopt()** on the socket associated to 1694 * *bpf_socket*, which must be a full socket. The *level* at 1695 * which the option resides and the name *optname* of the option 1696 * must be specified, see **setsockopt(2)** for more information. 1697 * The option value of length *optlen* is pointed by *optval*. 1698 * 1699 * *bpf_socket* should be one of the following: 1700 * 1701 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1702 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1703 * and **BPF_CGROUP_INET6_CONNECT**. 1704 * 1705 * This helper actually implements a subset of **setsockopt()**. 1706 * It supports the following *level*\ s: 1707 * 1708 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1709 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1710 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1711 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 1712 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1713 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1714 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1715 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1716 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 1717 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1718 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1719 * Return 1720 * 0 on success, or a negative error in case of failure. 1721 * 1722 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1723 * Description 1724 * Grow or shrink the room for data in the packet associated to 1725 * *skb* by *len_diff*, and according to the selected *mode*. 1726 * 1727 * By default, the helper will reset any offloaded checksum 1728 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1729 * by the following flag: 1730 * 1731 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1732 * checksum data of the skb to CHECKSUM_NONE. 1733 * 1734 * There are two supported modes at this time: 1735 * 1736 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1737 * (room space is added or removed below the layer 2 header). 1738 * 1739 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1740 * (room space is added or removed below the layer 3 header). 1741 * 1742 * The following flags are supported at this time: 1743 * 1744 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1745 * Adjusting mss in this way is not allowed for datagrams. 1746 * 1747 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1748 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1749 * Any new space is reserved to hold a tunnel header. 1750 * Configure skb offsets and other fields accordingly. 1751 * 1752 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1753 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1754 * Use with ENCAP_L3 flags to further specify the tunnel type. 1755 * 1756 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1757 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1758 * type; *len* is the length of the inner MAC header. 1759 * 1760 * A call to this helper is susceptible to change the underlying 1761 * packet buffer. Therefore, at load time, all checks on pointers 1762 * previously done by the verifier are invalidated and must be 1763 * performed again, if the helper is used in combination with 1764 * direct packet access. 1765 * Return 1766 * 0 on success, or a negative error in case of failure. 1767 * 1768 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1769 * Description 1770 * Redirect the packet to the endpoint referenced by *map* at 1771 * index *key*. Depending on its type, this *map* can contain 1772 * references to net devices (for forwarding packets through other 1773 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1774 * but this is only implemented for native XDP (with driver 1775 * support) as of this writing). 1776 * 1777 * The lower two bits of *flags* are used as the return code if 1778 * the map lookup fails. This is so that the return value can be 1779 * one of the XDP program return codes up to **XDP_TX**, as chosen 1780 * by the caller. Any higher bits in the *flags* argument must be 1781 * unset. 1782 * 1783 * See also **bpf_redirect**\ (), which only supports redirecting 1784 * to an ifindex, but doesn't require a map to do so. 1785 * Return 1786 * **XDP_REDIRECT** on success, or the value of the two lower bits 1787 * of the *flags* argument on error. 1788 * 1789 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1790 * Description 1791 * Redirect the packet to the socket referenced by *map* (of type 1792 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1793 * egress interfaces can be used for redirection. The 1794 * **BPF_F_INGRESS** value in *flags* is used to make the 1795 * distinction (ingress path is selected if the flag is present, 1796 * egress path otherwise). This is the only flag supported for now. 1797 * Return 1798 * **SK_PASS** on success, or **SK_DROP** on error. 1799 * 1800 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1801 * Description 1802 * Add an entry to, or update a *map* referencing sockets. The 1803 * *skops* is used as a new value for the entry associated to 1804 * *key*. *flags* is one of: 1805 * 1806 * **BPF_NOEXIST** 1807 * The entry for *key* must not exist in the map. 1808 * **BPF_EXIST** 1809 * The entry for *key* must already exist in the map. 1810 * **BPF_ANY** 1811 * No condition on the existence of the entry for *key*. 1812 * 1813 * If the *map* has eBPF programs (parser and verdict), those will 1814 * be inherited by the socket being added. If the socket is 1815 * already attached to eBPF programs, this results in an error. 1816 * Return 1817 * 0 on success, or a negative error in case of failure. 1818 * 1819 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1820 * Description 1821 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1822 * *delta* (which can be positive or negative). Note that this 1823 * operation modifies the address stored in *xdp_md*\ **->data**, 1824 * so the latter must be loaded only after the helper has been 1825 * called. 1826 * 1827 * The use of *xdp_md*\ **->data_meta** is optional and programs 1828 * are not required to use it. The rationale is that when the 1829 * packet is processed with XDP (e.g. as DoS filter), it is 1830 * possible to push further meta data along with it before passing 1831 * to the stack, and to give the guarantee that an ingress eBPF 1832 * program attached as a TC classifier on the same device can pick 1833 * this up for further post-processing. Since TC works with socket 1834 * buffers, it remains possible to set from XDP the **mark** or 1835 * **priority** pointers, or other pointers for the socket buffer. 1836 * Having this scratch space generic and programmable allows for 1837 * more flexibility as the user is free to store whatever meta 1838 * data they need. 1839 * 1840 * A call to this helper is susceptible to change the underlying 1841 * packet buffer. Therefore, at load time, all checks on pointers 1842 * previously done by the verifier are invalidated and must be 1843 * performed again, if the helper is used in combination with 1844 * direct packet access. 1845 * Return 1846 * 0 on success, or a negative error in case of failure. 1847 * 1848 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1849 * Description 1850 * Read the value of a perf event counter, and store it into *buf* 1851 * of size *buf_size*. This helper relies on a *map* of type 1852 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1853 * counter is selected when *map* is updated with perf event file 1854 * descriptors. The *map* is an array whose size is the number of 1855 * available CPUs, and each cell contains a value relative to one 1856 * CPU. The value to retrieve is indicated by *flags*, that 1857 * contains the index of the CPU to look up, masked with 1858 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1859 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1860 * current CPU should be retrieved. 1861 * 1862 * This helper behaves in a way close to 1863 * **bpf_perf_event_read**\ () helper, save that instead of 1864 * just returning the value observed, it fills the *buf* 1865 * structure. This allows for additional data to be retrieved: in 1866 * particular, the enabled and running times (in *buf*\ 1867 * **->enabled** and *buf*\ **->running**, respectively) are 1868 * copied. In general, **bpf_perf_event_read_value**\ () is 1869 * recommended over **bpf_perf_event_read**\ (), which has some 1870 * ABI issues and provides fewer functionalities. 1871 * 1872 * These values are interesting, because hardware PMU (Performance 1873 * Monitoring Unit) counters are limited resources. When there are 1874 * more PMU based perf events opened than available counters, 1875 * kernel will multiplex these events so each event gets certain 1876 * percentage (but not all) of the PMU time. In case that 1877 * multiplexing happens, the number of samples or counter value 1878 * will not reflect the case compared to when no multiplexing 1879 * occurs. This makes comparison between different runs difficult. 1880 * Typically, the counter value should be normalized before 1881 * comparing to other experiments. The usual normalization is done 1882 * as follows. 1883 * 1884 * :: 1885 * 1886 * normalized_counter = counter * t_enabled / t_running 1887 * 1888 * Where t_enabled is the time enabled for event and t_running is 1889 * the time running for event since last normalization. The 1890 * enabled and running times are accumulated since the perf event 1891 * open. To achieve scaling factor between two invocations of an 1892 * eBPF program, users can use CPU id as the key (which is 1893 * typical for perf array usage model) to remember the previous 1894 * value and do the calculation inside the eBPF program. 1895 * Return 1896 * 0 on success, or a negative error in case of failure. 1897 * 1898 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1899 * Description 1900 * For en eBPF program attached to a perf event, retrieve the 1901 * value of the event counter associated to *ctx* and store it in 1902 * the structure pointed by *buf* and of size *buf_size*. Enabled 1903 * and running times are also stored in the structure (see 1904 * description of helper **bpf_perf_event_read_value**\ () for 1905 * more details). 1906 * Return 1907 * 0 on success, or a negative error in case of failure. 1908 * 1909 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1910 * Description 1911 * Emulate a call to **getsockopt()** on the socket associated to 1912 * *bpf_socket*, which must be a full socket. The *level* at 1913 * which the option resides and the name *optname* of the option 1914 * must be specified, see **getsockopt(2)** for more information. 1915 * The retrieved value is stored in the structure pointed by 1916 * *opval* and of length *optlen*. 1917 * 1918 * *bpf_socket* should be one of the following: 1919 * 1920 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1921 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1922 * and **BPF_CGROUP_INET6_CONNECT**. 1923 * 1924 * This helper actually implements a subset of **getsockopt()**. 1925 * It supports the following *level*\ s: 1926 * 1927 * * **IPPROTO_TCP**, which supports *optname* 1928 * **TCP_CONGESTION**. 1929 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1930 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1931 * Return 1932 * 0 on success, or a negative error in case of failure. 1933 * 1934 * long bpf_override_return(struct pt_regs *regs, u64 rc) 1935 * Description 1936 * Used for error injection, this helper uses kprobes to override 1937 * the return value of the probed function, and to set it to *rc*. 1938 * The first argument is the context *regs* on which the kprobe 1939 * works. 1940 * 1941 * This helper works by setting the PC (program counter) 1942 * to an override function which is run in place of the original 1943 * probed function. This means the probed function is not run at 1944 * all. The replacement function just returns with the required 1945 * value. 1946 * 1947 * This helper has security implications, and thus is subject to 1948 * restrictions. It is only available if the kernel was compiled 1949 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1950 * option, and in this case it only works on functions tagged with 1951 * **ALLOW_ERROR_INJECTION** in the kernel code. 1952 * 1953 * Also, the helper is only available for the architectures having 1954 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1955 * x86 architecture is the only one to support this feature. 1956 * Return 1957 * 0 1958 * 1959 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1960 * Description 1961 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1962 * for the full TCP socket associated to *bpf_sock_ops* to 1963 * *argval*. 1964 * 1965 * The primary use of this field is to determine if there should 1966 * be calls to eBPF programs of type 1967 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1968 * code. A program of the same type can change its value, per 1969 * connection and as necessary, when the connection is 1970 * established. This field is directly accessible for reading, but 1971 * this helper must be used for updates in order to return an 1972 * error if an eBPF program tries to set a callback that is not 1973 * supported in the current kernel. 1974 * 1975 * *argval* is a flag array which can combine these flags: 1976 * 1977 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1978 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1979 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1980 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1981 * 1982 * Therefore, this function can be used to clear a callback flag by 1983 * setting the appropriate bit to zero. e.g. to disable the RTO 1984 * callback: 1985 * 1986 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1987 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1988 * 1989 * Here are some examples of where one could call such eBPF 1990 * program: 1991 * 1992 * * When RTO fires. 1993 * * When a packet is retransmitted. 1994 * * When the connection terminates. 1995 * * When a packet is sent. 1996 * * When a packet is received. 1997 * Return 1998 * Code **-EINVAL** if the socket is not a full TCP socket; 1999 * otherwise, a positive number containing the bits that could not 2000 * be set is returned (which comes down to 0 if all bits were set 2001 * as required). 2002 * 2003 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2004 * Description 2005 * This helper is used in programs implementing policies at the 2006 * socket level. If the message *msg* is allowed to pass (i.e. if 2007 * the verdict eBPF program returns **SK_PASS**), redirect it to 2008 * the socket referenced by *map* (of type 2009 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2010 * egress interfaces can be used for redirection. The 2011 * **BPF_F_INGRESS** value in *flags* is used to make the 2012 * distinction (ingress path is selected if the flag is present, 2013 * egress path otherwise). This is the only flag supported for now. 2014 * Return 2015 * **SK_PASS** on success, or **SK_DROP** on error. 2016 * 2017 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2018 * Description 2019 * For socket policies, apply the verdict of the eBPF program to 2020 * the next *bytes* (number of bytes) of message *msg*. 2021 * 2022 * For example, this helper can be used in the following cases: 2023 * 2024 * * A single **sendmsg**\ () or **sendfile**\ () system call 2025 * contains multiple logical messages that the eBPF program is 2026 * supposed to read and for which it should apply a verdict. 2027 * * An eBPF program only cares to read the first *bytes* of a 2028 * *msg*. If the message has a large payload, then setting up 2029 * and calling the eBPF program repeatedly for all bytes, even 2030 * though the verdict is already known, would create unnecessary 2031 * overhead. 2032 * 2033 * When called from within an eBPF program, the helper sets a 2034 * counter internal to the BPF infrastructure, that is used to 2035 * apply the last verdict to the next *bytes*. If *bytes* is 2036 * smaller than the current data being processed from a 2037 * **sendmsg**\ () or **sendfile**\ () system call, the first 2038 * *bytes* will be sent and the eBPF program will be re-run with 2039 * the pointer for start of data pointing to byte number *bytes* 2040 * **+ 1**. If *bytes* is larger than the current data being 2041 * processed, then the eBPF verdict will be applied to multiple 2042 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2043 * consumed. 2044 * 2045 * Note that if a socket closes with the internal counter holding 2046 * a non-zero value, this is not a problem because data is not 2047 * being buffered for *bytes* and is sent as it is received. 2048 * Return 2049 * 0 2050 * 2051 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2052 * Description 2053 * For socket policies, prevent the execution of the verdict eBPF 2054 * program for message *msg* until *bytes* (byte number) have been 2055 * accumulated. 2056 * 2057 * This can be used when one needs a specific number of bytes 2058 * before a verdict can be assigned, even if the data spans 2059 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2060 * case would be a user calling **sendmsg**\ () repeatedly with 2061 * 1-byte long message segments. Obviously, this is bad for 2062 * performance, but it is still valid. If the eBPF program needs 2063 * *bytes* bytes to validate a header, this helper can be used to 2064 * prevent the eBPF program to be called again until *bytes* have 2065 * been accumulated. 2066 * Return 2067 * 0 2068 * 2069 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2070 * Description 2071 * For socket policies, pull in non-linear data from user space 2072 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2073 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2074 * respectively. 2075 * 2076 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2077 * *msg* it can only parse data that the (**data**, **data_end**) 2078 * pointers have already consumed. For **sendmsg**\ () hooks this 2079 * is likely the first scatterlist element. But for calls relying 2080 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2081 * be the range (**0**, **0**) because the data is shared with 2082 * user space and by default the objective is to avoid allowing 2083 * user space to modify data while (or after) eBPF verdict is 2084 * being decided. This helper can be used to pull in data and to 2085 * set the start and end pointer to given values. Data will be 2086 * copied if necessary (i.e. if data was not linear and if start 2087 * and end pointers do not point to the same chunk). 2088 * 2089 * A call to this helper is susceptible to change the underlying 2090 * packet buffer. Therefore, at load time, all checks on pointers 2091 * previously done by the verifier are invalidated and must be 2092 * performed again, if the helper is used in combination with 2093 * direct packet access. 2094 * 2095 * All values for *flags* are reserved for future usage, and must 2096 * be left at zero. 2097 * Return 2098 * 0 on success, or a negative error in case of failure. 2099 * 2100 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2101 * Description 2102 * Bind the socket associated to *ctx* to the address pointed by 2103 * *addr*, of length *addr_len*. This allows for making outgoing 2104 * connection from the desired IP address, which can be useful for 2105 * example when all processes inside a cgroup should use one 2106 * single IP address on a host that has multiple IP configured. 2107 * 2108 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2109 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2110 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2111 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2112 * behavior and lets the kernel efficiently pick up an unused 2113 * port as long as 4-tuple is unique. Passing non-zero port might 2114 * lead to degraded performance. 2115 * Return 2116 * 0 on success, or a negative error in case of failure. 2117 * 2118 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2119 * Description 2120 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2121 * possible to both shrink and grow the packet tail. 2122 * Shrink done via *delta* being a negative integer. 2123 * 2124 * A call to this helper is susceptible to change the underlying 2125 * packet buffer. Therefore, at load time, all checks on pointers 2126 * previously done by the verifier are invalidated and must be 2127 * performed again, if the helper is used in combination with 2128 * direct packet access. 2129 * Return 2130 * 0 on success, or a negative error in case of failure. 2131 * 2132 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2133 * Description 2134 * Retrieve the XFRM state (IP transform framework, see also 2135 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2136 * 2137 * The retrieved value is stored in the **struct bpf_xfrm_state** 2138 * pointed by *xfrm_state* and of length *size*. 2139 * 2140 * All values for *flags* are reserved for future usage, and must 2141 * be left at zero. 2142 * 2143 * This helper is available only if the kernel was compiled with 2144 * **CONFIG_XFRM** configuration option. 2145 * Return 2146 * 0 on success, or a negative error in case of failure. 2147 * 2148 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2149 * Description 2150 * Return a user or a kernel stack in bpf program provided buffer. 2151 * To achieve this, the helper needs *ctx*, which is a pointer 2152 * to the context on which the tracing program is executed. 2153 * To store the stacktrace, the bpf program provides *buf* with 2154 * a nonnegative *size*. 2155 * 2156 * The last argument, *flags*, holds the number of stack frames to 2157 * skip (from 0 to 255), masked with 2158 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2159 * the following flags: 2160 * 2161 * **BPF_F_USER_STACK** 2162 * Collect a user space stack instead of a kernel stack. 2163 * **BPF_F_USER_BUILD_ID** 2164 * Collect buildid+offset instead of ips for user stack, 2165 * only valid if **BPF_F_USER_STACK** is also specified. 2166 * 2167 * **bpf_get_stack**\ () can collect up to 2168 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2169 * to sufficient large buffer size. Note that 2170 * this limit can be controlled with the **sysctl** program, and 2171 * that it should be manually increased in order to profile long 2172 * user stacks (such as stacks for Java programs). To do so, use: 2173 * 2174 * :: 2175 * 2176 * # sysctl kernel.perf_event_max_stack=<new value> 2177 * Return 2178 * A non-negative value equal to or less than *size* on success, 2179 * or a negative error in case of failure. 2180 * 2181 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2182 * Description 2183 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2184 * it provides an easy way to load *len* bytes from *offset* 2185 * from the packet associated to *skb*, into the buffer pointed 2186 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2187 * a fifth argument *start_header* exists in order to select a 2188 * base offset to start from. *start_header* can be one of: 2189 * 2190 * **BPF_HDR_START_MAC** 2191 * Base offset to load data from is *skb*'s mac header. 2192 * **BPF_HDR_START_NET** 2193 * Base offset to load data from is *skb*'s network header. 2194 * 2195 * In general, "direct packet access" is the preferred method to 2196 * access packet data, however, this helper is in particular useful 2197 * in socket filters where *skb*\ **->data** does not always point 2198 * to the start of the mac header and where "direct packet access" 2199 * is not available. 2200 * Return 2201 * 0 on success, or a negative error in case of failure. 2202 * 2203 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2204 * Description 2205 * Do FIB lookup in kernel tables using parameters in *params*. 2206 * If lookup is successful and result shows packet is to be 2207 * forwarded, the neighbor tables are searched for the nexthop. 2208 * If successful (ie., FIB lookup shows forwarding and nexthop 2209 * is resolved), the nexthop address is returned in ipv4_dst 2210 * or ipv6_dst based on family, smac is set to mac address of 2211 * egress device, dmac is set to nexthop mac address, rt_metric 2212 * is set to metric from route (IPv4/IPv6 only), and ifindex 2213 * is set to the device index of the nexthop from the FIB lookup. 2214 * 2215 * *plen* argument is the size of the passed in struct. 2216 * *flags* argument can be a combination of one or more of the 2217 * following values: 2218 * 2219 * **BPF_FIB_LOOKUP_DIRECT** 2220 * Do a direct table lookup vs full lookup using FIB 2221 * rules. 2222 * **BPF_FIB_LOOKUP_OUTPUT** 2223 * Perform lookup from an egress perspective (default is 2224 * ingress). 2225 * 2226 * *ctx* is either **struct xdp_md** for XDP programs or 2227 * **struct sk_buff** tc cls_act programs. 2228 * Return 2229 * * < 0 if any input argument is invalid 2230 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2231 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2232 * packet is not forwarded or needs assist from full stack 2233 * 2234 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2235 * Description 2236 * Add an entry to, or update a sockhash *map* referencing sockets. 2237 * The *skops* is used as a new value for the entry associated to 2238 * *key*. *flags* is one of: 2239 * 2240 * **BPF_NOEXIST** 2241 * The entry for *key* must not exist in the map. 2242 * **BPF_EXIST** 2243 * The entry for *key* must already exist in the map. 2244 * **BPF_ANY** 2245 * No condition on the existence of the entry for *key*. 2246 * 2247 * If the *map* has eBPF programs (parser and verdict), those will 2248 * be inherited by the socket being added. If the socket is 2249 * already attached to eBPF programs, this results in an error. 2250 * Return 2251 * 0 on success, or a negative error in case of failure. 2252 * 2253 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2254 * Description 2255 * This helper is used in programs implementing policies at the 2256 * socket level. If the message *msg* is allowed to pass (i.e. if 2257 * the verdict eBPF program returns **SK_PASS**), redirect it to 2258 * the socket referenced by *map* (of type 2259 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2260 * egress interfaces can be used for redirection. The 2261 * **BPF_F_INGRESS** value in *flags* is used to make the 2262 * distinction (ingress path is selected if the flag is present, 2263 * egress path otherwise). This is the only flag supported for now. 2264 * Return 2265 * **SK_PASS** on success, or **SK_DROP** on error. 2266 * 2267 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2268 * Description 2269 * This helper is used in programs implementing policies at the 2270 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2271 * if the verdict eBPF program returns **SK_PASS**), redirect it 2272 * to the socket referenced by *map* (of type 2273 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2274 * egress interfaces can be used for redirection. The 2275 * **BPF_F_INGRESS** value in *flags* is used to make the 2276 * distinction (ingress path is selected if the flag is present, 2277 * egress otherwise). This is the only flag supported for now. 2278 * Return 2279 * **SK_PASS** on success, or **SK_DROP** on error. 2280 * 2281 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2282 * Description 2283 * Encapsulate the packet associated to *skb* within a Layer 3 2284 * protocol header. This header is provided in the buffer at 2285 * address *hdr*, with *len* its size in bytes. *type* indicates 2286 * the protocol of the header and can be one of: 2287 * 2288 * **BPF_LWT_ENCAP_SEG6** 2289 * IPv6 encapsulation with Segment Routing Header 2290 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2291 * the IPv6 header is computed by the kernel. 2292 * **BPF_LWT_ENCAP_SEG6_INLINE** 2293 * Only works if *skb* contains an IPv6 packet. Insert a 2294 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2295 * the IPv6 header. 2296 * **BPF_LWT_ENCAP_IP** 2297 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2298 * must be IPv4 or IPv6, followed by zero or more 2299 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2300 * total bytes in all prepended headers. Please note that 2301 * if **skb_is_gso**\ (*skb*) is true, no more than two 2302 * headers can be prepended, and the inner header, if 2303 * present, should be either GRE or UDP/GUE. 2304 * 2305 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2306 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2307 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2308 * **BPF_PROG_TYPE_LWT_XMIT**. 2309 * 2310 * A call to this helper is susceptible to change the underlying 2311 * packet buffer. Therefore, at load time, all checks on pointers 2312 * previously done by the verifier are invalidated and must be 2313 * performed again, if the helper is used in combination with 2314 * direct packet access. 2315 * Return 2316 * 0 on success, or a negative error in case of failure. 2317 * 2318 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2319 * Description 2320 * Store *len* bytes from address *from* into the packet 2321 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2322 * inside the outermost IPv6 Segment Routing Header can be 2323 * modified through this helper. 2324 * 2325 * A call to this helper is susceptible to change the underlying 2326 * packet buffer. Therefore, at load time, all checks on pointers 2327 * previously done by the verifier are invalidated and must be 2328 * performed again, if the helper is used in combination with 2329 * direct packet access. 2330 * Return 2331 * 0 on success, or a negative error in case of failure. 2332 * 2333 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2334 * Description 2335 * Adjust the size allocated to TLVs in the outermost IPv6 2336 * Segment Routing Header contained in the packet associated to 2337 * *skb*, at position *offset* by *delta* bytes. Only offsets 2338 * after the segments are accepted. *delta* can be as well 2339 * positive (growing) as negative (shrinking). 2340 * 2341 * A call to this helper is susceptible to change the underlying 2342 * packet buffer. Therefore, at load time, all checks on pointers 2343 * previously done by the verifier are invalidated and must be 2344 * performed again, if the helper is used in combination with 2345 * direct packet access. 2346 * Return 2347 * 0 on success, or a negative error in case of failure. 2348 * 2349 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2350 * Description 2351 * Apply an IPv6 Segment Routing action of type *action* to the 2352 * packet associated to *skb*. Each action takes a parameter 2353 * contained at address *param*, and of length *param_len* bytes. 2354 * *action* can be one of: 2355 * 2356 * **SEG6_LOCAL_ACTION_END_X** 2357 * End.X action: Endpoint with Layer-3 cross-connect. 2358 * Type of *param*: **struct in6_addr**. 2359 * **SEG6_LOCAL_ACTION_END_T** 2360 * End.T action: Endpoint with specific IPv6 table lookup. 2361 * Type of *param*: **int**. 2362 * **SEG6_LOCAL_ACTION_END_B6** 2363 * End.B6 action: Endpoint bound to an SRv6 policy. 2364 * Type of *param*: **struct ipv6_sr_hdr**. 2365 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2366 * End.B6.Encap action: Endpoint bound to an SRv6 2367 * encapsulation policy. 2368 * Type of *param*: **struct ipv6_sr_hdr**. 2369 * 2370 * A call to this helper is susceptible to change the underlying 2371 * packet buffer. Therefore, at load time, all checks on pointers 2372 * previously done by the verifier are invalidated and must be 2373 * performed again, if the helper is used in combination with 2374 * direct packet access. 2375 * Return 2376 * 0 on success, or a negative error in case of failure. 2377 * 2378 * long bpf_rc_repeat(void *ctx) 2379 * Description 2380 * This helper is used in programs implementing IR decoding, to 2381 * report a successfully decoded repeat key message. This delays 2382 * the generation of a key up event for previously generated 2383 * key down event. 2384 * 2385 * Some IR protocols like NEC have a special IR message for 2386 * repeating last button, for when a button is held down. 2387 * 2388 * The *ctx* should point to the lirc sample as passed into 2389 * the program. 2390 * 2391 * This helper is only available is the kernel was compiled with 2392 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2393 * "**y**". 2394 * Return 2395 * 0 2396 * 2397 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2398 * Description 2399 * This helper is used in programs implementing IR decoding, to 2400 * report a successfully decoded key press with *scancode*, 2401 * *toggle* value in the given *protocol*. The scancode will be 2402 * translated to a keycode using the rc keymap, and reported as 2403 * an input key down event. After a period a key up event is 2404 * generated. This period can be extended by calling either 2405 * **bpf_rc_keydown**\ () again with the same values, or calling 2406 * **bpf_rc_repeat**\ (). 2407 * 2408 * Some protocols include a toggle bit, in case the button was 2409 * released and pressed again between consecutive scancodes. 2410 * 2411 * The *ctx* should point to the lirc sample as passed into 2412 * the program. 2413 * 2414 * The *protocol* is the decoded protocol number (see 2415 * **enum rc_proto** for some predefined values). 2416 * 2417 * This helper is only available is the kernel was compiled with 2418 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2419 * "**y**". 2420 * Return 2421 * 0 2422 * 2423 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2424 * Description 2425 * Return the cgroup v2 id of the socket associated with the *skb*. 2426 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2427 * helper for cgroup v1 by providing a tag resp. identifier that 2428 * can be matched on or used for map lookups e.g. to implement 2429 * policy. The cgroup v2 id of a given path in the hierarchy is 2430 * exposed in user space through the f_handle API in order to get 2431 * to the same 64-bit id. 2432 * 2433 * This helper can be used on TC egress path, but not on ingress, 2434 * and is available only if the kernel was compiled with the 2435 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2436 * Return 2437 * The id is returned or 0 in case the id could not be retrieved. 2438 * 2439 * u64 bpf_get_current_cgroup_id(void) 2440 * Return 2441 * A 64-bit integer containing the current cgroup id based 2442 * on the cgroup within which the current task is running. 2443 * 2444 * void *bpf_get_local_storage(void *map, u64 flags) 2445 * Description 2446 * Get the pointer to the local storage area. 2447 * The type and the size of the local storage is defined 2448 * by the *map* argument. 2449 * The *flags* meaning is specific for each map type, 2450 * and has to be 0 for cgroup local storage. 2451 * 2452 * Depending on the BPF program type, a local storage area 2453 * can be shared between multiple instances of the BPF program, 2454 * running simultaneously. 2455 * 2456 * A user should care about the synchronization by himself. 2457 * For example, by using the **BPF_ATOMIC** instructions to alter 2458 * the shared data. 2459 * Return 2460 * A pointer to the local storage area. 2461 * 2462 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2463 * Description 2464 * Select a **SO_REUSEPORT** socket from a 2465 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2466 * It checks the selected socket is matching the incoming 2467 * request in the socket buffer. 2468 * Return 2469 * 0 on success, or a negative error in case of failure. 2470 * 2471 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2472 * Description 2473 * Return id of cgroup v2 that is ancestor of cgroup associated 2474 * with the *skb* at the *ancestor_level*. The root cgroup is at 2475 * *ancestor_level* zero and each step down the hierarchy 2476 * increments the level. If *ancestor_level* == level of cgroup 2477 * associated with *skb*, then return value will be same as that 2478 * of **bpf_skb_cgroup_id**\ (). 2479 * 2480 * The helper is useful to implement policies based on cgroups 2481 * that are upper in hierarchy than immediate cgroup associated 2482 * with *skb*. 2483 * 2484 * The format of returned id and helper limitations are same as in 2485 * **bpf_skb_cgroup_id**\ (). 2486 * Return 2487 * The id is returned or 0 in case the id could not be retrieved. 2488 * 2489 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2490 * Description 2491 * Look for TCP socket matching *tuple*, optionally in a child 2492 * network namespace *netns*. The return value must be checked, 2493 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2494 * 2495 * The *ctx* should point to the context of the program, such as 2496 * the skb or socket (depending on the hook in use). This is used 2497 * to determine the base network namespace for the lookup. 2498 * 2499 * *tuple_size* must be one of: 2500 * 2501 * **sizeof**\ (*tuple*\ **->ipv4**) 2502 * Look for an IPv4 socket. 2503 * **sizeof**\ (*tuple*\ **->ipv6**) 2504 * Look for an IPv6 socket. 2505 * 2506 * If the *netns* is a negative signed 32-bit integer, then the 2507 * socket lookup table in the netns associated with the *ctx* 2508 * will be used. For the TC hooks, this is the netns of the device 2509 * in the skb. For socket hooks, this is the netns of the socket. 2510 * If *netns* is any other signed 32-bit value greater than or 2511 * equal to zero then it specifies the ID of the netns relative to 2512 * the netns associated with the *ctx*. *netns* values beyond the 2513 * range of 32-bit integers are reserved for future use. 2514 * 2515 * All values for *flags* are reserved for future usage, and must 2516 * be left at zero. 2517 * 2518 * This helper is available only if the kernel was compiled with 2519 * **CONFIG_NET** configuration option. 2520 * Return 2521 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2522 * For sockets with reuseport option, the **struct bpf_sock** 2523 * result is from *reuse*\ **->socks**\ [] using the hash of the 2524 * tuple. 2525 * 2526 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2527 * Description 2528 * Look for UDP socket matching *tuple*, optionally in a child 2529 * network namespace *netns*. The return value must be checked, 2530 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2531 * 2532 * The *ctx* should point to the context of the program, such as 2533 * the skb or socket (depending on the hook in use). This is used 2534 * to determine the base network namespace for the lookup. 2535 * 2536 * *tuple_size* must be one of: 2537 * 2538 * **sizeof**\ (*tuple*\ **->ipv4**) 2539 * Look for an IPv4 socket. 2540 * **sizeof**\ (*tuple*\ **->ipv6**) 2541 * Look for an IPv6 socket. 2542 * 2543 * If the *netns* is a negative signed 32-bit integer, then the 2544 * socket lookup table in the netns associated with the *ctx* 2545 * will be used. For the TC hooks, this is the netns of the device 2546 * in the skb. For socket hooks, this is the netns of the socket. 2547 * If *netns* is any other signed 32-bit value greater than or 2548 * equal to zero then it specifies the ID of the netns relative to 2549 * the netns associated with the *ctx*. *netns* values beyond the 2550 * range of 32-bit integers are reserved for future use. 2551 * 2552 * All values for *flags* are reserved for future usage, and must 2553 * be left at zero. 2554 * 2555 * This helper is available only if the kernel was compiled with 2556 * **CONFIG_NET** configuration option. 2557 * Return 2558 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2559 * For sockets with reuseport option, the **struct bpf_sock** 2560 * result is from *reuse*\ **->socks**\ [] using the hash of the 2561 * tuple. 2562 * 2563 * long bpf_sk_release(void *sock) 2564 * Description 2565 * Release the reference held by *sock*. *sock* must be a 2566 * non-**NULL** pointer that was returned from 2567 * **bpf_sk_lookup_xxx**\ (). 2568 * Return 2569 * 0 on success, or a negative error in case of failure. 2570 * 2571 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2572 * Description 2573 * Push an element *value* in *map*. *flags* is one of: 2574 * 2575 * **BPF_EXIST** 2576 * If the queue/stack is full, the oldest element is 2577 * removed to make room for this. 2578 * Return 2579 * 0 on success, or a negative error in case of failure. 2580 * 2581 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 2582 * Description 2583 * Pop an element from *map*. 2584 * Return 2585 * 0 on success, or a negative error in case of failure. 2586 * 2587 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 2588 * Description 2589 * Get an element from *map* without removing it. 2590 * Return 2591 * 0 on success, or a negative error in case of failure. 2592 * 2593 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2594 * Description 2595 * For socket policies, insert *len* bytes into *msg* at offset 2596 * *start*. 2597 * 2598 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2599 * *msg* it may want to insert metadata or options into the *msg*. 2600 * This can later be read and used by any of the lower layer BPF 2601 * hooks. 2602 * 2603 * This helper may fail if under memory pressure (a malloc 2604 * fails) in these cases BPF programs will get an appropriate 2605 * error and BPF programs will need to handle them. 2606 * Return 2607 * 0 on success, or a negative error in case of failure. 2608 * 2609 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2610 * Description 2611 * Will remove *len* bytes from a *msg* starting at byte *start*. 2612 * This may result in **ENOMEM** errors under certain situations if 2613 * an allocation and copy are required due to a full ring buffer. 2614 * However, the helper will try to avoid doing the allocation 2615 * if possible. Other errors can occur if input parameters are 2616 * invalid either due to *start* byte not being valid part of *msg* 2617 * payload and/or *pop* value being to large. 2618 * Return 2619 * 0 on success, or a negative error in case of failure. 2620 * 2621 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2622 * Description 2623 * This helper is used in programs implementing IR decoding, to 2624 * report a successfully decoded pointer movement. 2625 * 2626 * The *ctx* should point to the lirc sample as passed into 2627 * the program. 2628 * 2629 * This helper is only available is the kernel was compiled with 2630 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2631 * "**y**". 2632 * Return 2633 * 0 2634 * 2635 * long bpf_spin_lock(struct bpf_spin_lock *lock) 2636 * Description 2637 * Acquire a spinlock represented by the pointer *lock*, which is 2638 * stored as part of a value of a map. Taking the lock allows to 2639 * safely update the rest of the fields in that value. The 2640 * spinlock can (and must) later be released with a call to 2641 * **bpf_spin_unlock**\ (\ *lock*\ ). 2642 * 2643 * Spinlocks in BPF programs come with a number of restrictions 2644 * and constraints: 2645 * 2646 * * **bpf_spin_lock** objects are only allowed inside maps of 2647 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2648 * list could be extended in the future). 2649 * * BTF description of the map is mandatory. 2650 * * The BPF program can take ONE lock at a time, since taking two 2651 * or more could cause dead locks. 2652 * * Only one **struct bpf_spin_lock** is allowed per map element. 2653 * * When the lock is taken, calls (either BPF to BPF or helpers) 2654 * are not allowed. 2655 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2656 * allowed inside a spinlock-ed region. 2657 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2658 * the lock, on all execution paths, before it returns. 2659 * * The BPF program can access **struct bpf_spin_lock** only via 2660 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2661 * helpers. Loading or storing data into the **struct 2662 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2663 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2664 * of the map value must be a struct and have **struct 2665 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2666 * Nested lock inside another struct is not allowed. 2667 * * The **struct bpf_spin_lock** *lock* field in a map value must 2668 * be aligned on a multiple of 4 bytes in that value. 2669 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2670 * the **bpf_spin_lock** field to user space. 2671 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2672 * a BPF program, do not update the **bpf_spin_lock** field. 2673 * * **bpf_spin_lock** cannot be on the stack or inside a 2674 * networking packet (it can only be inside of a map values). 2675 * * **bpf_spin_lock** is available to root only. 2676 * * Tracing programs and socket filter programs cannot use 2677 * **bpf_spin_lock**\ () due to insufficient preemption checks 2678 * (but this may change in the future). 2679 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2680 * Return 2681 * 0 2682 * 2683 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 2684 * Description 2685 * Release the *lock* previously locked by a call to 2686 * **bpf_spin_lock**\ (\ *lock*\ ). 2687 * Return 2688 * 0 2689 * 2690 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2691 * Description 2692 * This helper gets a **struct bpf_sock** pointer such 2693 * that all the fields in this **bpf_sock** can be accessed. 2694 * Return 2695 * A **struct bpf_sock** pointer on success, or **NULL** in 2696 * case of failure. 2697 * 2698 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2699 * Description 2700 * This helper gets a **struct bpf_tcp_sock** pointer from a 2701 * **struct bpf_sock** pointer. 2702 * Return 2703 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2704 * case of failure. 2705 * 2706 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 2707 * Description 2708 * Set ECN (Explicit Congestion Notification) field of IP header 2709 * to **CE** (Congestion Encountered) if current value is **ECT** 2710 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2711 * and IPv4. 2712 * Return 2713 * 1 if the **CE** flag is set (either by the current helper call 2714 * or because it was already present), 0 if it is not set. 2715 * 2716 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2717 * Description 2718 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2719 * **bpf_sk_release**\ () is unnecessary and not allowed. 2720 * Return 2721 * A **struct bpf_sock** pointer on success, or **NULL** in 2722 * case of failure. 2723 * 2724 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2725 * Description 2726 * Look for TCP socket matching *tuple*, optionally in a child 2727 * network namespace *netns*. The return value must be checked, 2728 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2729 * 2730 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2731 * that it also returns timewait or request sockets. Use 2732 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2733 * full structure. 2734 * 2735 * This helper is available only if the kernel was compiled with 2736 * **CONFIG_NET** configuration option. 2737 * Return 2738 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2739 * For sockets with reuseport option, the **struct bpf_sock** 2740 * result is from *reuse*\ **->socks**\ [] using the hash of the 2741 * tuple. 2742 * 2743 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2744 * Description 2745 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2746 * the listening socket in *sk*. 2747 * 2748 * *iph* points to the start of the IPv4 or IPv6 header, while 2749 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2750 * **sizeof**\ (**struct ip6hdr**). 2751 * 2752 * *th* points to the start of the TCP header, while *th_len* 2753 * contains **sizeof**\ (**struct tcphdr**). 2754 * Return 2755 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2756 * error otherwise. 2757 * 2758 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2759 * Description 2760 * Get name of sysctl in /proc/sys/ and copy it into provided by 2761 * program buffer *buf* of size *buf_len*. 2762 * 2763 * The buffer is always NUL terminated, unless it's zero-sized. 2764 * 2765 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2766 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2767 * only (e.g. "tcp_mem"). 2768 * Return 2769 * Number of character copied (not including the trailing NUL). 2770 * 2771 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2772 * truncated name in this case). 2773 * 2774 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2775 * Description 2776 * Get current value of sysctl as it is presented in /proc/sys 2777 * (incl. newline, etc), and copy it as a string into provided 2778 * by program buffer *buf* of size *buf_len*. 2779 * 2780 * The whole value is copied, no matter what file position user 2781 * space issued e.g. sys_read at. 2782 * 2783 * The buffer is always NUL terminated, unless it's zero-sized. 2784 * Return 2785 * Number of character copied (not including the trailing NUL). 2786 * 2787 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2788 * truncated name in this case). 2789 * 2790 * **-EINVAL** if current value was unavailable, e.g. because 2791 * sysctl is uninitialized and read returns -EIO for it. 2792 * 2793 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2794 * Description 2795 * Get new value being written by user space to sysctl (before 2796 * the actual write happens) and copy it as a string into 2797 * provided by program buffer *buf* of size *buf_len*. 2798 * 2799 * User space may write new value at file position > 0. 2800 * 2801 * The buffer is always NUL terminated, unless it's zero-sized. 2802 * Return 2803 * Number of character copied (not including the trailing NUL). 2804 * 2805 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2806 * truncated name in this case). 2807 * 2808 * **-EINVAL** if sysctl is being read. 2809 * 2810 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2811 * Description 2812 * Override new value being written by user space to sysctl with 2813 * value provided by program in buffer *buf* of size *buf_len*. 2814 * 2815 * *buf* should contain a string in same form as provided by user 2816 * space on sysctl write. 2817 * 2818 * User space may write new value at file position > 0. To override 2819 * the whole sysctl value file position should be set to zero. 2820 * Return 2821 * 0 on success. 2822 * 2823 * **-E2BIG** if the *buf_len* is too big. 2824 * 2825 * **-EINVAL** if sysctl is being read. 2826 * 2827 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2828 * Description 2829 * Convert the initial part of the string from buffer *buf* of 2830 * size *buf_len* to a long integer according to the given base 2831 * and save the result in *res*. 2832 * 2833 * The string may begin with an arbitrary amount of white space 2834 * (as determined by **isspace**\ (3)) followed by a single 2835 * optional '**-**' sign. 2836 * 2837 * Five least significant bits of *flags* encode base, other bits 2838 * are currently unused. 2839 * 2840 * Base must be either 8, 10, 16 or 0 to detect it automatically 2841 * similar to user space **strtol**\ (3). 2842 * Return 2843 * Number of characters consumed on success. Must be positive but 2844 * no more than *buf_len*. 2845 * 2846 * **-EINVAL** if no valid digits were found or unsupported base 2847 * was provided. 2848 * 2849 * **-ERANGE** if resulting value was out of range. 2850 * 2851 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2852 * Description 2853 * Convert the initial part of the string from buffer *buf* of 2854 * size *buf_len* to an unsigned long integer according to the 2855 * given base and save the result in *res*. 2856 * 2857 * The string may begin with an arbitrary amount of white space 2858 * (as determined by **isspace**\ (3)). 2859 * 2860 * Five least significant bits of *flags* encode base, other bits 2861 * are currently unused. 2862 * 2863 * Base must be either 8, 10, 16 or 0 to detect it automatically 2864 * similar to user space **strtoul**\ (3). 2865 * Return 2866 * Number of characters consumed on success. Must be positive but 2867 * no more than *buf_len*. 2868 * 2869 * **-EINVAL** if no valid digits were found or unsupported base 2870 * was provided. 2871 * 2872 * **-ERANGE** if resulting value was out of range. 2873 * 2874 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 2875 * Description 2876 * Get a bpf-local-storage from a *sk*. 2877 * 2878 * Logically, it could be thought of getting the value from 2879 * a *map* with *sk* as the **key**. From this 2880 * perspective, the usage is not much different from 2881 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2882 * helper enforces the key must be a full socket and the map must 2883 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2884 * 2885 * Underneath, the value is stored locally at *sk* instead of 2886 * the *map*. The *map* is used as the bpf-local-storage 2887 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2888 * searched against all bpf-local-storages residing at *sk*. 2889 * 2890 * *sk* is a kernel **struct sock** pointer for LSM program. 2891 * *sk* is a **struct bpf_sock** pointer for other program types. 2892 * 2893 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2894 * used such that a new bpf-local-storage will be 2895 * created if one does not exist. *value* can be used 2896 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2897 * the initial value of a bpf-local-storage. If *value* is 2898 * **NULL**, the new bpf-local-storage will be zero initialized. 2899 * Return 2900 * A bpf-local-storage pointer is returned on success. 2901 * 2902 * **NULL** if not found or there was an error in adding 2903 * a new bpf-local-storage. 2904 * 2905 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 2906 * Description 2907 * Delete a bpf-local-storage from a *sk*. 2908 * Return 2909 * 0 on success. 2910 * 2911 * **-ENOENT** if the bpf-local-storage cannot be found. 2912 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 2913 * 2914 * long bpf_send_signal(u32 sig) 2915 * Description 2916 * Send signal *sig* to the process of the current task. 2917 * The signal may be delivered to any of this process's threads. 2918 * Return 2919 * 0 on success or successfully queued. 2920 * 2921 * **-EBUSY** if work queue under nmi is full. 2922 * 2923 * **-EINVAL** if *sig* is invalid. 2924 * 2925 * **-EPERM** if no permission to send the *sig*. 2926 * 2927 * **-EAGAIN** if bpf program can try again. 2928 * 2929 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2930 * Description 2931 * Try to issue a SYN cookie for the packet with corresponding 2932 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2933 * 2934 * *iph* points to the start of the IPv4 or IPv6 header, while 2935 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2936 * **sizeof**\ (**struct ip6hdr**). 2937 * 2938 * *th* points to the start of the TCP header, while *th_len* 2939 * contains the length of the TCP header. 2940 * Return 2941 * On success, lower 32 bits hold the generated SYN cookie in 2942 * followed by 16 bits which hold the MSS value for that cookie, 2943 * and the top 16 bits are unused. 2944 * 2945 * On failure, the returned value is one of the following: 2946 * 2947 * **-EINVAL** SYN cookie cannot be issued due to error 2948 * 2949 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2950 * 2951 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2952 * 2953 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2954 * 2955 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2956 * Description 2957 * Write raw *data* blob into a special BPF perf event held by 2958 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2959 * event must have the following attributes: **PERF_SAMPLE_RAW** 2960 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2961 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2962 * 2963 * The *flags* are used to indicate the index in *map* for which 2964 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2965 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2966 * to indicate that the index of the current CPU core should be 2967 * used. 2968 * 2969 * The value to write, of *size*, is passed through eBPF stack and 2970 * pointed by *data*. 2971 * 2972 * *ctx* is a pointer to in-kernel struct sk_buff. 2973 * 2974 * This helper is similar to **bpf_perf_event_output**\ () but 2975 * restricted to raw_tracepoint bpf programs. 2976 * Return 2977 * 0 on success, or a negative error in case of failure. 2978 * 2979 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2980 * Description 2981 * Safely attempt to read *size* bytes from user space address 2982 * *unsafe_ptr* and store the data in *dst*. 2983 * Return 2984 * 0 on success, or a negative error in case of failure. 2985 * 2986 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2987 * Description 2988 * Safely attempt to read *size* bytes from kernel space address 2989 * *unsafe_ptr* and store the data in *dst*. 2990 * Return 2991 * 0 on success, or a negative error in case of failure. 2992 * 2993 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2994 * Description 2995 * Copy a NUL terminated string from an unsafe user address 2996 * *unsafe_ptr* to *dst*. The *size* should include the 2997 * terminating NUL byte. In case the string length is smaller than 2998 * *size*, the target is not padded with further NUL bytes. If the 2999 * string length is larger than *size*, just *size*-1 bytes are 3000 * copied and the last byte is set to NUL. 3001 * 3002 * On success, returns the number of bytes that were written, 3003 * including the terminal NUL. This makes this helper useful in 3004 * tracing programs for reading strings, and more importantly to 3005 * get its length at runtime. See the following snippet: 3006 * 3007 * :: 3008 * 3009 * SEC("kprobe/sys_open") 3010 * void bpf_sys_open(struct pt_regs *ctx) 3011 * { 3012 * char buf[PATHLEN]; // PATHLEN is defined to 256 3013 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3014 * ctx->di); 3015 * 3016 * // Consume buf, for example push it to 3017 * // userspace via bpf_perf_event_output(); we 3018 * // can use res (the string length) as event 3019 * // size, after checking its boundaries. 3020 * } 3021 * 3022 * In comparison, using **bpf_probe_read_user**\ () helper here 3023 * instead to read the string would require to estimate the length 3024 * at compile time, and would often result in copying more memory 3025 * than necessary. 3026 * 3027 * Another useful use case is when parsing individual process 3028 * arguments or individual environment variables navigating 3029 * *current*\ **->mm->arg_start** and *current*\ 3030 * **->mm->env_start**: using this helper and the return value, 3031 * one can quickly iterate at the right offset of the memory area. 3032 * Return 3033 * On success, the strictly positive length of the output string, 3034 * including the trailing NUL character. On error, a negative 3035 * value. 3036 * 3037 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3038 * Description 3039 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3040 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3041 * Return 3042 * On success, the strictly positive length of the string, including 3043 * the trailing NUL character. On error, a negative value. 3044 * 3045 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3046 * Description 3047 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3048 * *rcv_nxt* is the ack_seq to be sent out. 3049 * Return 3050 * 0 on success, or a negative error in case of failure. 3051 * 3052 * long bpf_send_signal_thread(u32 sig) 3053 * Description 3054 * Send signal *sig* to the thread corresponding to the current task. 3055 * Return 3056 * 0 on success or successfully queued. 3057 * 3058 * **-EBUSY** if work queue under nmi is full. 3059 * 3060 * **-EINVAL** if *sig* is invalid. 3061 * 3062 * **-EPERM** if no permission to send the *sig*. 3063 * 3064 * **-EAGAIN** if bpf program can try again. 3065 * 3066 * u64 bpf_jiffies64(void) 3067 * Description 3068 * Obtain the 64bit jiffies 3069 * Return 3070 * The 64 bit jiffies 3071 * 3072 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3073 * Description 3074 * For an eBPF program attached to a perf event, retrieve the 3075 * branch records (**struct perf_branch_entry**) associated to *ctx* 3076 * and store it in the buffer pointed by *buf* up to size 3077 * *size* bytes. 3078 * Return 3079 * On success, number of bytes written to *buf*. On error, a 3080 * negative value. 3081 * 3082 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3083 * instead return the number of bytes required to store all the 3084 * branch entries. If this flag is set, *buf* may be NULL. 3085 * 3086 * **-EINVAL** if arguments invalid or **size** not a multiple 3087 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3088 * 3089 * **-ENOENT** if architecture does not support branch records. 3090 * 3091 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3092 * Description 3093 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3094 * *namespace* will be returned in *nsdata*. 3095 * Return 3096 * 0 on success, or one of the following in case of failure: 3097 * 3098 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3099 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3100 * 3101 * **-ENOENT** if pidns does not exists for the current task. 3102 * 3103 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3104 * Description 3105 * Write raw *data* blob into a special BPF perf event held by 3106 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3107 * event must have the following attributes: **PERF_SAMPLE_RAW** 3108 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3109 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3110 * 3111 * The *flags* are used to indicate the index in *map* for which 3112 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3113 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3114 * to indicate that the index of the current CPU core should be 3115 * used. 3116 * 3117 * The value to write, of *size*, is passed through eBPF stack and 3118 * pointed by *data*. 3119 * 3120 * *ctx* is a pointer to in-kernel struct xdp_buff. 3121 * 3122 * This helper is similar to **bpf_perf_eventoutput**\ () but 3123 * restricted to raw_tracepoint bpf programs. 3124 * Return 3125 * 0 on success, or a negative error in case of failure. 3126 * 3127 * u64 bpf_get_netns_cookie(void *ctx) 3128 * Description 3129 * Retrieve the cookie (generated by the kernel) of the network 3130 * namespace the input *ctx* is associated with. The network 3131 * namespace cookie remains stable for its lifetime and provides 3132 * a global identifier that can be assumed unique. If *ctx* is 3133 * NULL, then the helper returns the cookie for the initial 3134 * network namespace. The cookie itself is very similar to that 3135 * of **bpf_get_socket_cookie**\ () helper, but for network 3136 * namespaces instead of sockets. 3137 * Return 3138 * A 8-byte long opaque number. 3139 * 3140 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3141 * Description 3142 * Return id of cgroup v2 that is ancestor of the cgroup associated 3143 * with the current task at the *ancestor_level*. The root cgroup 3144 * is at *ancestor_level* zero and each step down the hierarchy 3145 * increments the level. If *ancestor_level* == level of cgroup 3146 * associated with the current task, then return value will be the 3147 * same as that of **bpf_get_current_cgroup_id**\ (). 3148 * 3149 * The helper is useful to implement policies based on cgroups 3150 * that are upper in hierarchy than immediate cgroup associated 3151 * with the current task. 3152 * 3153 * The format of returned id and helper limitations are same as in 3154 * **bpf_get_current_cgroup_id**\ (). 3155 * Return 3156 * The id is returned or 0 in case the id could not be retrieved. 3157 * 3158 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3159 * Description 3160 * Helper is overloaded depending on BPF program type. This 3161 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3162 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3163 * 3164 * Assign the *sk* to the *skb*. When combined with appropriate 3165 * routing configuration to receive the packet towards the socket, 3166 * will cause *skb* to be delivered to the specified socket. 3167 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3168 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3169 * interfere with successful delivery to the socket. 3170 * 3171 * This operation is only valid from TC ingress path. 3172 * 3173 * The *flags* argument must be zero. 3174 * Return 3175 * 0 on success, or a negative error in case of failure: 3176 * 3177 * **-EINVAL** if specified *flags* are not supported. 3178 * 3179 * **-ENOENT** if the socket is unavailable for assignment. 3180 * 3181 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3182 * 3183 * **-EOPNOTSUPP** if the operation is not supported, for example 3184 * a call from outside of TC ingress. 3185 * 3186 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3187 * (reuseport). 3188 * 3189 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3190 * Description 3191 * Helper is overloaded depending on BPF program type. This 3192 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3193 * 3194 * Select the *sk* as a result of a socket lookup. 3195 * 3196 * For the operation to succeed passed socket must be compatible 3197 * with the packet description provided by the *ctx* object. 3198 * 3199 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3200 * be an exact match. While IP family (**AF_INET** or 3201 * **AF_INET6**) must be compatible, that is IPv6 sockets 3202 * that are not v6-only can be selected for IPv4 packets. 3203 * 3204 * Only TCP listeners and UDP unconnected sockets can be 3205 * selected. *sk* can also be NULL to reset any previous 3206 * selection. 3207 * 3208 * *flags* argument can combination of following values: 3209 * 3210 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3211 * socket selection, potentially done by a BPF program 3212 * that ran before us. 3213 * 3214 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3215 * load-balancing within reuseport group for the socket 3216 * being selected. 3217 * 3218 * On success *ctx->sk* will point to the selected socket. 3219 * 3220 * Return 3221 * 0 on success, or a negative errno in case of failure. 3222 * 3223 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3224 * not compatible with packet family (*ctx->family*). 3225 * 3226 * * **-EEXIST** if socket has been already selected, 3227 * potentially by another program, and 3228 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3229 * 3230 * * **-EINVAL** if unsupported flags were specified. 3231 * 3232 * * **-EPROTOTYPE** if socket L4 protocol 3233 * (*sk->protocol*) doesn't match packet protocol 3234 * (*ctx->protocol*). 3235 * 3236 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3237 * state (TCP listening or UDP unconnected). 3238 * 3239 * u64 bpf_ktime_get_boot_ns(void) 3240 * Description 3241 * Return the time elapsed since system boot, in nanoseconds. 3242 * Does include the time the system was suspended. 3243 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3244 * Return 3245 * Current *ktime*. 3246 * 3247 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3248 * Description 3249 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3250 * out the format string. 3251 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3252 * the format string itself. The *data* and *data_len* are format string 3253 * arguments. The *data* are a **u64** array and corresponding format string 3254 * values are stored in the array. For strings and pointers where pointees 3255 * are accessed, only the pointer values are stored in the *data* array. 3256 * The *data_len* is the size of *data* in bytes. 3257 * 3258 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3259 * Reading kernel memory may fail due to either invalid address or 3260 * valid address but requiring a major memory fault. If reading kernel memory 3261 * fails, the string for **%s** will be an empty string, and the ip 3262 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3263 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3264 * Return 3265 * 0 on success, or a negative error in case of failure: 3266 * 3267 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3268 * by returning 1 from bpf program. 3269 * 3270 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3271 * 3272 * **-E2BIG** if *fmt* contains too many format specifiers. 3273 * 3274 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3275 * 3276 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 3277 * Description 3278 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3279 * The *m* represents the seq_file. The *data* and *len* represent the 3280 * data to write in bytes. 3281 * Return 3282 * 0 on success, or a negative error in case of failure: 3283 * 3284 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3285 * 3286 * u64 bpf_sk_cgroup_id(void *sk) 3287 * Description 3288 * Return the cgroup v2 id of the socket *sk*. 3289 * 3290 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 3291 * returned from **bpf_sk_lookup_xxx**\ (), 3292 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3293 * same as in **bpf_skb_cgroup_id**\ (). 3294 * 3295 * This helper is available only if the kernel was compiled with 3296 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3297 * Return 3298 * The id is returned or 0 in case the id could not be retrieved. 3299 * 3300 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 3301 * Description 3302 * Return id of cgroup v2 that is ancestor of cgroup associated 3303 * with the *sk* at the *ancestor_level*. The root cgroup is at 3304 * *ancestor_level* zero and each step down the hierarchy 3305 * increments the level. If *ancestor_level* == level of cgroup 3306 * associated with *sk*, then return value will be same as that 3307 * of **bpf_sk_cgroup_id**\ (). 3308 * 3309 * The helper is useful to implement policies based on cgroups 3310 * that are upper in hierarchy than immediate cgroup associated 3311 * with *sk*. 3312 * 3313 * The format of returned id and helper limitations are same as in 3314 * **bpf_sk_cgroup_id**\ (). 3315 * Return 3316 * The id is returned or 0 in case the id could not be retrieved. 3317 * 3318 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 3319 * Description 3320 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3321 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3322 * of new data availability is sent. 3323 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3324 * of new data availability is sent unconditionally. 3325 * Return 3326 * 0 on success, or a negative error in case of failure. 3327 * 3328 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 3329 * Description 3330 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3331 * Return 3332 * Valid pointer with *size* bytes of memory available; NULL, 3333 * otherwise. 3334 * 3335 * void bpf_ringbuf_submit(void *data, u64 flags) 3336 * Description 3337 * Submit reserved ring buffer sample, pointed to by *data*. 3338 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3339 * of new data availability is sent. 3340 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3341 * of new data availability is sent unconditionally. 3342 * Return 3343 * Nothing. Always succeeds. 3344 * 3345 * void bpf_ringbuf_discard(void *data, u64 flags) 3346 * Description 3347 * Discard reserved ring buffer sample, pointed to by *data*. 3348 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3349 * of new data availability is sent. 3350 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3351 * of new data availability is sent unconditionally. 3352 * Return 3353 * Nothing. Always succeeds. 3354 * 3355 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 3356 * Description 3357 * Query various characteristics of provided ring buffer. What 3358 * exactly is queries is determined by *flags*: 3359 * 3360 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3361 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3362 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3363 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3364 * 3365 * Data returned is just a momentary snapshot of actual values 3366 * and could be inaccurate, so this facility should be used to 3367 * power heuristics and for reporting, not to make 100% correct 3368 * calculation. 3369 * Return 3370 * Requested value, or 0, if *flags* are not recognized. 3371 * 3372 * long bpf_csum_level(struct sk_buff *skb, u64 level) 3373 * Description 3374 * Change the skbs checksum level by one layer up or down, or 3375 * reset it entirely to none in order to have the stack perform 3376 * checksum validation. The level is applicable to the following 3377 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3378 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3379 * through **bpf_skb_adjust_room**\ () helper with passing in 3380 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3381 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3382 * the UDP header is removed. Similarly, an encap of the latter 3383 * into the former could be accompanied by a helper call to 3384 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3385 * skb is still intended to be processed in higher layers of the 3386 * stack instead of just egressing at tc. 3387 * 3388 * There are three supported level settings at this time: 3389 * 3390 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3391 * with CHECKSUM_UNNECESSARY. 3392 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3393 * with CHECKSUM_UNNECESSARY. 3394 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3395 * sets CHECKSUM_NONE to force checksum validation by the stack. 3396 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3397 * skb->csum_level. 3398 * Return 3399 * 0 on success, or a negative error in case of failure. In the 3400 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3401 * is returned or the error code -EACCES in case the skb is not 3402 * subject to CHECKSUM_UNNECESSARY. 3403 * 3404 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 3405 * Description 3406 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3407 * Return 3408 * *sk* if casting is valid, or **NULL** otherwise. 3409 * 3410 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 3411 * Description 3412 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3413 * Return 3414 * *sk* if casting is valid, or **NULL** otherwise. 3415 * 3416 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 3417 * Description 3418 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3419 * Return 3420 * *sk* if casting is valid, or **NULL** otherwise. 3421 * 3422 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 3423 * Description 3424 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3425 * Return 3426 * *sk* if casting is valid, or **NULL** otherwise. 3427 * 3428 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 3429 * Description 3430 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3431 * Return 3432 * *sk* if casting is valid, or **NULL** otherwise. 3433 * 3434 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 3435 * Description 3436 * Return a user or a kernel stack in bpf program provided buffer. 3437 * To achieve this, the helper needs *task*, which is a valid 3438 * pointer to **struct task_struct**. To store the stacktrace, the 3439 * bpf program provides *buf* with a nonnegative *size*. 3440 * 3441 * The last argument, *flags*, holds the number of stack frames to 3442 * skip (from 0 to 255), masked with 3443 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3444 * the following flags: 3445 * 3446 * **BPF_F_USER_STACK** 3447 * Collect a user space stack instead of a kernel stack. 3448 * **BPF_F_USER_BUILD_ID** 3449 * Collect buildid+offset instead of ips for user stack, 3450 * only valid if **BPF_F_USER_STACK** is also specified. 3451 * 3452 * **bpf_get_task_stack**\ () can collect up to 3453 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3454 * to sufficient large buffer size. Note that 3455 * this limit can be controlled with the **sysctl** program, and 3456 * that it should be manually increased in order to profile long 3457 * user stacks (such as stacks for Java programs). To do so, use: 3458 * 3459 * :: 3460 * 3461 * # sysctl kernel.perf_event_max_stack=<new value> 3462 * Return 3463 * A non-negative value equal to or less than *size* on success, 3464 * or a negative error in case of failure. 3465 * 3466 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 3467 * Description 3468 * Load header option. Support reading a particular TCP header 3469 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 3470 * 3471 * If *flags* is 0, it will search the option from the 3472 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 3473 * has details on what skb_data contains under different 3474 * *skops*\ **->op**. 3475 * 3476 * The first byte of the *searchby_res* specifies the 3477 * kind that it wants to search. 3478 * 3479 * If the searching kind is an experimental kind 3480 * (i.e. 253 or 254 according to RFC6994). It also 3481 * needs to specify the "magic" which is either 3482 * 2 bytes or 4 bytes. It then also needs to 3483 * specify the size of the magic by using 3484 * the 2nd byte which is "kind-length" of a TCP 3485 * header option and the "kind-length" also 3486 * includes the first 2 bytes "kind" and "kind-length" 3487 * itself as a normal TCP header option also does. 3488 * 3489 * For example, to search experimental kind 254 with 3490 * 2 byte magic 0xeB9F, the searchby_res should be 3491 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 3492 * 3493 * To search for the standard window scale option (3), 3494 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 3495 * Note, kind-length must be 0 for regular option. 3496 * 3497 * Searching for No-Op (0) and End-of-Option-List (1) are 3498 * not supported. 3499 * 3500 * *len* must be at least 2 bytes which is the minimal size 3501 * of a header option. 3502 * 3503 * Supported flags: 3504 * 3505 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 3506 * saved_syn packet or the just-received syn packet. 3507 * 3508 * Return 3509 * > 0 when found, the header option is copied to *searchby_res*. 3510 * The return value is the total length copied. On failure, a 3511 * negative error code is returned: 3512 * 3513 * **-EINVAL** if a parameter is invalid. 3514 * 3515 * **-ENOMSG** if the option is not found. 3516 * 3517 * **-ENOENT** if no syn packet is available when 3518 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 3519 * 3520 * **-ENOSPC** if there is not enough space. Only *len* number of 3521 * bytes are copied. 3522 * 3523 * **-EFAULT** on failure to parse the header options in the 3524 * packet. 3525 * 3526 * **-EPERM** if the helper cannot be used under the current 3527 * *skops*\ **->op**. 3528 * 3529 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 3530 * Description 3531 * Store header option. The data will be copied 3532 * from buffer *from* with length *len* to the TCP header. 3533 * 3534 * The buffer *from* should have the whole option that 3535 * includes the kind, kind-length, and the actual 3536 * option data. The *len* must be at least kind-length 3537 * long. The kind-length does not have to be 4 byte 3538 * aligned. The kernel will take care of the padding 3539 * and setting the 4 bytes aligned value to th->doff. 3540 * 3541 * This helper will check for duplicated option 3542 * by searching the same option in the outgoing skb. 3543 * 3544 * This helper can only be called during 3545 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3546 * 3547 * Return 3548 * 0 on success, or negative error in case of failure: 3549 * 3550 * **-EINVAL** If param is invalid. 3551 * 3552 * **-ENOSPC** if there is not enough space in the header. 3553 * Nothing has been written 3554 * 3555 * **-EEXIST** if the option already exists. 3556 * 3557 * **-EFAULT** on failrue to parse the existing header options. 3558 * 3559 * **-EPERM** if the helper cannot be used under the current 3560 * *skops*\ **->op**. 3561 * 3562 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 3563 * Description 3564 * Reserve *len* bytes for the bpf header option. The 3565 * space will be used by **bpf_store_hdr_opt**\ () later in 3566 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3567 * 3568 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 3569 * the total number of bytes will be reserved. 3570 * 3571 * This helper can only be called during 3572 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 3573 * 3574 * Return 3575 * 0 on success, or negative error in case of failure: 3576 * 3577 * **-EINVAL** if a parameter is invalid. 3578 * 3579 * **-ENOSPC** if there is not enough space in the header. 3580 * 3581 * **-EPERM** if the helper cannot be used under the current 3582 * *skops*\ **->op**. 3583 * 3584 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 3585 * Description 3586 * Get a bpf_local_storage from an *inode*. 3587 * 3588 * Logically, it could be thought of as getting the value from 3589 * a *map* with *inode* as the **key**. From this 3590 * perspective, the usage is not much different from 3591 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 3592 * helper enforces the key must be an inode and the map must also 3593 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 3594 * 3595 * Underneath, the value is stored locally at *inode* instead of 3596 * the *map*. The *map* is used as the bpf-local-storage 3597 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3598 * searched against all bpf_local_storage residing at *inode*. 3599 * 3600 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3601 * used such that a new bpf_local_storage will be 3602 * created if one does not exist. *value* can be used 3603 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3604 * the initial value of a bpf_local_storage. If *value* is 3605 * **NULL**, the new bpf_local_storage will be zero initialized. 3606 * Return 3607 * A bpf_local_storage pointer is returned on success. 3608 * 3609 * **NULL** if not found or there was an error in adding 3610 * a new bpf_local_storage. 3611 * 3612 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 3613 * Description 3614 * Delete a bpf_local_storage from an *inode*. 3615 * Return 3616 * 0 on success. 3617 * 3618 * **-ENOENT** if the bpf_local_storage cannot be found. 3619 * 3620 * long bpf_d_path(struct path *path, char *buf, u32 sz) 3621 * Description 3622 * Return full path for given **struct path** object, which 3623 * needs to be the kernel BTF *path* object. The path is 3624 * returned in the provided buffer *buf* of size *sz* and 3625 * is zero terminated. 3626 * 3627 * Return 3628 * On success, the strictly positive length of the string, 3629 * including the trailing NUL character. On error, a negative 3630 * value. 3631 * 3632 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 3633 * Description 3634 * Read *size* bytes from user space address *user_ptr* and store 3635 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 3636 * Return 3637 * 0 on success, or a negative error in case of failure. 3638 * 3639 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 3640 * Description 3641 * Use BTF to store a string representation of *ptr*->ptr in *str*, 3642 * using *ptr*->type_id. This value should specify the type 3643 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 3644 * can be used to look up vmlinux BTF type ids. Traversing the 3645 * data structure using BTF, the type information and values are 3646 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 3647 * the pointer data is carried out to avoid kernel crashes during 3648 * operation. Smaller types can use string space on the stack; 3649 * larger programs can use map data to store the string 3650 * representation. 3651 * 3652 * The string can be subsequently shared with userspace via 3653 * bpf_perf_event_output() or ring buffer interfaces. 3654 * bpf_trace_printk() is to be avoided as it places too small 3655 * a limit on string size to be useful. 3656 * 3657 * *flags* is a combination of 3658 * 3659 * **BTF_F_COMPACT** 3660 * no formatting around type information 3661 * **BTF_F_NONAME** 3662 * no struct/union member names/types 3663 * **BTF_F_PTR_RAW** 3664 * show raw (unobfuscated) pointer values; 3665 * equivalent to printk specifier %px. 3666 * **BTF_F_ZERO** 3667 * show zero-valued struct/union members; they 3668 * are not displayed by default 3669 * 3670 * Return 3671 * The number of bytes that were written (or would have been 3672 * written if output had to be truncated due to string size), 3673 * or a negative error in cases of failure. 3674 * 3675 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 3676 * Description 3677 * Use BTF to write to seq_write a string representation of 3678 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 3679 * *flags* are identical to those used for bpf_snprintf_btf. 3680 * Return 3681 * 0 on success or a negative error in case of failure. 3682 * 3683 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 3684 * Description 3685 * See **bpf_get_cgroup_classid**\ () for the main description. 3686 * This helper differs from **bpf_get_cgroup_classid**\ () in that 3687 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 3688 * associated socket instead of the current process. 3689 * Return 3690 * The id is returned or 0 in case the id could not be retrieved. 3691 * 3692 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 3693 * Description 3694 * Redirect the packet to another net device of index *ifindex* 3695 * and fill in L2 addresses from neighboring subsystem. This helper 3696 * is somewhat similar to **bpf_redirect**\ (), except that it 3697 * populates L2 addresses as well, meaning, internally, the helper 3698 * relies on the neighbor lookup for the L2 address of the nexthop. 3699 * 3700 * The helper will perform a FIB lookup based on the skb's 3701 * networking header to get the address of the next hop, unless 3702 * this is supplied by the caller in the *params* argument. The 3703 * *plen* argument indicates the len of *params* and should be set 3704 * to 0 if *params* is NULL. 3705 * 3706 * The *flags* argument is reserved and must be 0. The helper is 3707 * currently only supported for tc BPF program types, and enabled 3708 * for IPv4 and IPv6 protocols. 3709 * Return 3710 * The helper returns **TC_ACT_REDIRECT** on success or 3711 * **TC_ACT_SHOT** on error. 3712 * 3713 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 3714 * Description 3715 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3716 * pointer to the percpu kernel variable on *cpu*. A ksym is an 3717 * extern variable decorated with '__ksym'. For ksym, there is a 3718 * global var (either static or global) defined of the same name 3719 * in the kernel. The ksym is percpu if the global var is percpu. 3720 * The returned pointer points to the global percpu var on *cpu*. 3721 * 3722 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 3723 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 3724 * happens if *cpu* is larger than nr_cpu_ids. The caller of 3725 * bpf_per_cpu_ptr() must check the returned value. 3726 * Return 3727 * A pointer pointing to the kernel percpu variable on *cpu*, or 3728 * NULL, if *cpu* is invalid. 3729 * 3730 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 3731 * Description 3732 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3733 * pointer to the percpu kernel variable on this cpu. See the 3734 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 3735 * 3736 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 3737 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 3738 * never return NULL. 3739 * Return 3740 * A pointer pointing to the kernel percpu variable on this cpu. 3741 * 3742 * long bpf_redirect_peer(u32 ifindex, u64 flags) 3743 * Description 3744 * Redirect the packet to another net device of index *ifindex*. 3745 * This helper is somewhat similar to **bpf_redirect**\ (), except 3746 * that the redirection happens to the *ifindex*' peer device and 3747 * the netns switch takes place from ingress to ingress without 3748 * going through the CPU's backlog queue. 3749 * 3750 * The *flags* argument is reserved and must be 0. The helper is 3751 * currently only supported for tc BPF program types at the ingress 3752 * hook and for veth device types. The peer device must reside in a 3753 * different network namespace. 3754 * Return 3755 * The helper returns **TC_ACT_REDIRECT** on success or 3756 * **TC_ACT_SHOT** on error. 3757 * 3758 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 3759 * Description 3760 * Get a bpf_local_storage from the *task*. 3761 * 3762 * Logically, it could be thought of as getting the value from 3763 * a *map* with *task* as the **key**. From this 3764 * perspective, the usage is not much different from 3765 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 3766 * helper enforces the key must be an task_struct and the map must also 3767 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 3768 * 3769 * Underneath, the value is stored locally at *task* instead of 3770 * the *map*. The *map* is used as the bpf-local-storage 3771 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3772 * searched against all bpf_local_storage residing at *task*. 3773 * 3774 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3775 * used such that a new bpf_local_storage will be 3776 * created if one does not exist. *value* can be used 3777 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3778 * the initial value of a bpf_local_storage. If *value* is 3779 * **NULL**, the new bpf_local_storage will be zero initialized. 3780 * Return 3781 * A bpf_local_storage pointer is returned on success. 3782 * 3783 * **NULL** if not found or there was an error in adding 3784 * a new bpf_local_storage. 3785 * 3786 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 3787 * Description 3788 * Delete a bpf_local_storage from a *task*. 3789 * Return 3790 * 0 on success. 3791 * 3792 * **-ENOENT** if the bpf_local_storage cannot be found. 3793 * 3794 * struct task_struct *bpf_get_current_task_btf(void) 3795 * Description 3796 * Return a BTF pointer to the "current" task. 3797 * This pointer can also be used in helpers that accept an 3798 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 3799 * Return 3800 * Pointer to the current task. 3801 * 3802 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 3803 * Description 3804 * Set or clear certain options on *bprm*: 3805 * 3806 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 3807 * which sets the **AT_SECURE** auxv for glibc. The bit 3808 * is cleared if the flag is not specified. 3809 * Return 3810 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 3811 * 3812 * u64 bpf_ktime_get_coarse_ns(void) 3813 * Description 3814 * Return a coarse-grained version of the time elapsed since 3815 * system boot, in nanoseconds. Does not include time the system 3816 * was suspended. 3817 * 3818 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 3819 * Return 3820 * Current *ktime*. 3821 * 3822 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 3823 * Description 3824 * Returns the stored IMA hash of the *inode* (if it's avaialable). 3825 * If the hash is larger than *size*, then only *size* 3826 * bytes will be copied to *dst* 3827 * Return 3828 * The **hash_algo** is returned on success, 3829 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 3830 * invalid arguments are passed. 3831 * 3832 * struct socket *bpf_sock_from_file(struct file *file) 3833 * Description 3834 * If the given file represents a socket, returns the associated 3835 * socket. 3836 * Return 3837 * A pointer to a struct socket on success or NULL if the file is 3838 * not a socket. 3839 */ 3840 #define __BPF_FUNC_MAPPER(FN) \ 3841 FN(unspec), \ 3842 FN(map_lookup_elem), \ 3843 FN(map_update_elem), \ 3844 FN(map_delete_elem), \ 3845 FN(probe_read), \ 3846 FN(ktime_get_ns), \ 3847 FN(trace_printk), \ 3848 FN(get_prandom_u32), \ 3849 FN(get_smp_processor_id), \ 3850 FN(skb_store_bytes), \ 3851 FN(l3_csum_replace), \ 3852 FN(l4_csum_replace), \ 3853 FN(tail_call), \ 3854 FN(clone_redirect), \ 3855 FN(get_current_pid_tgid), \ 3856 FN(get_current_uid_gid), \ 3857 FN(get_current_comm), \ 3858 FN(get_cgroup_classid), \ 3859 FN(skb_vlan_push), \ 3860 FN(skb_vlan_pop), \ 3861 FN(skb_get_tunnel_key), \ 3862 FN(skb_set_tunnel_key), \ 3863 FN(perf_event_read), \ 3864 FN(redirect), \ 3865 FN(get_route_realm), \ 3866 FN(perf_event_output), \ 3867 FN(skb_load_bytes), \ 3868 FN(get_stackid), \ 3869 FN(csum_diff), \ 3870 FN(skb_get_tunnel_opt), \ 3871 FN(skb_set_tunnel_opt), \ 3872 FN(skb_change_proto), \ 3873 FN(skb_change_type), \ 3874 FN(skb_under_cgroup), \ 3875 FN(get_hash_recalc), \ 3876 FN(get_current_task), \ 3877 FN(probe_write_user), \ 3878 FN(current_task_under_cgroup), \ 3879 FN(skb_change_tail), \ 3880 FN(skb_pull_data), \ 3881 FN(csum_update), \ 3882 FN(set_hash_invalid), \ 3883 FN(get_numa_node_id), \ 3884 FN(skb_change_head), \ 3885 FN(xdp_adjust_head), \ 3886 FN(probe_read_str), \ 3887 FN(get_socket_cookie), \ 3888 FN(get_socket_uid), \ 3889 FN(set_hash), \ 3890 FN(setsockopt), \ 3891 FN(skb_adjust_room), \ 3892 FN(redirect_map), \ 3893 FN(sk_redirect_map), \ 3894 FN(sock_map_update), \ 3895 FN(xdp_adjust_meta), \ 3896 FN(perf_event_read_value), \ 3897 FN(perf_prog_read_value), \ 3898 FN(getsockopt), \ 3899 FN(override_return), \ 3900 FN(sock_ops_cb_flags_set), \ 3901 FN(msg_redirect_map), \ 3902 FN(msg_apply_bytes), \ 3903 FN(msg_cork_bytes), \ 3904 FN(msg_pull_data), \ 3905 FN(bind), \ 3906 FN(xdp_adjust_tail), \ 3907 FN(skb_get_xfrm_state), \ 3908 FN(get_stack), \ 3909 FN(skb_load_bytes_relative), \ 3910 FN(fib_lookup), \ 3911 FN(sock_hash_update), \ 3912 FN(msg_redirect_hash), \ 3913 FN(sk_redirect_hash), \ 3914 FN(lwt_push_encap), \ 3915 FN(lwt_seg6_store_bytes), \ 3916 FN(lwt_seg6_adjust_srh), \ 3917 FN(lwt_seg6_action), \ 3918 FN(rc_repeat), \ 3919 FN(rc_keydown), \ 3920 FN(skb_cgroup_id), \ 3921 FN(get_current_cgroup_id), \ 3922 FN(get_local_storage), \ 3923 FN(sk_select_reuseport), \ 3924 FN(skb_ancestor_cgroup_id), \ 3925 FN(sk_lookup_tcp), \ 3926 FN(sk_lookup_udp), \ 3927 FN(sk_release), \ 3928 FN(map_push_elem), \ 3929 FN(map_pop_elem), \ 3930 FN(map_peek_elem), \ 3931 FN(msg_push_data), \ 3932 FN(msg_pop_data), \ 3933 FN(rc_pointer_rel), \ 3934 FN(spin_lock), \ 3935 FN(spin_unlock), \ 3936 FN(sk_fullsock), \ 3937 FN(tcp_sock), \ 3938 FN(skb_ecn_set_ce), \ 3939 FN(get_listener_sock), \ 3940 FN(skc_lookup_tcp), \ 3941 FN(tcp_check_syncookie), \ 3942 FN(sysctl_get_name), \ 3943 FN(sysctl_get_current_value), \ 3944 FN(sysctl_get_new_value), \ 3945 FN(sysctl_set_new_value), \ 3946 FN(strtol), \ 3947 FN(strtoul), \ 3948 FN(sk_storage_get), \ 3949 FN(sk_storage_delete), \ 3950 FN(send_signal), \ 3951 FN(tcp_gen_syncookie), \ 3952 FN(skb_output), \ 3953 FN(probe_read_user), \ 3954 FN(probe_read_kernel), \ 3955 FN(probe_read_user_str), \ 3956 FN(probe_read_kernel_str), \ 3957 FN(tcp_send_ack), \ 3958 FN(send_signal_thread), \ 3959 FN(jiffies64), \ 3960 FN(read_branch_records), \ 3961 FN(get_ns_current_pid_tgid), \ 3962 FN(xdp_output), \ 3963 FN(get_netns_cookie), \ 3964 FN(get_current_ancestor_cgroup_id), \ 3965 FN(sk_assign), \ 3966 FN(ktime_get_boot_ns), \ 3967 FN(seq_printf), \ 3968 FN(seq_write), \ 3969 FN(sk_cgroup_id), \ 3970 FN(sk_ancestor_cgroup_id), \ 3971 FN(ringbuf_output), \ 3972 FN(ringbuf_reserve), \ 3973 FN(ringbuf_submit), \ 3974 FN(ringbuf_discard), \ 3975 FN(ringbuf_query), \ 3976 FN(csum_level), \ 3977 FN(skc_to_tcp6_sock), \ 3978 FN(skc_to_tcp_sock), \ 3979 FN(skc_to_tcp_timewait_sock), \ 3980 FN(skc_to_tcp_request_sock), \ 3981 FN(skc_to_udp6_sock), \ 3982 FN(get_task_stack), \ 3983 FN(load_hdr_opt), \ 3984 FN(store_hdr_opt), \ 3985 FN(reserve_hdr_opt), \ 3986 FN(inode_storage_get), \ 3987 FN(inode_storage_delete), \ 3988 FN(d_path), \ 3989 FN(copy_from_user), \ 3990 FN(snprintf_btf), \ 3991 FN(seq_printf_btf), \ 3992 FN(skb_cgroup_classid), \ 3993 FN(redirect_neigh), \ 3994 FN(per_cpu_ptr), \ 3995 FN(this_cpu_ptr), \ 3996 FN(redirect_peer), \ 3997 FN(task_storage_get), \ 3998 FN(task_storage_delete), \ 3999 FN(get_current_task_btf), \ 4000 FN(bprm_opts_set), \ 4001 FN(ktime_get_coarse_ns), \ 4002 FN(ima_inode_hash), \ 4003 FN(sock_from_file), \ 4004 /* */ 4005 4006 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 4007 * function eBPF program intends to call 4008 */ 4009 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 4010 enum bpf_func_id { 4011 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 4012 __BPF_FUNC_MAX_ID, 4013 }; 4014 #undef __BPF_ENUM_FN 4015 4016 /* All flags used by eBPF helper functions, placed here. */ 4017 4018 /* BPF_FUNC_skb_store_bytes flags. */ 4019 enum { 4020 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 4021 BPF_F_INVALIDATE_HASH = (1ULL << 1), 4022 }; 4023 4024 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 4025 * First 4 bits are for passing the header field size. 4026 */ 4027 enum { 4028 BPF_F_HDR_FIELD_MASK = 0xfULL, 4029 }; 4030 4031 /* BPF_FUNC_l4_csum_replace flags. */ 4032 enum { 4033 BPF_F_PSEUDO_HDR = (1ULL << 4), 4034 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 4035 BPF_F_MARK_ENFORCE = (1ULL << 6), 4036 }; 4037 4038 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 4039 enum { 4040 BPF_F_INGRESS = (1ULL << 0), 4041 }; 4042 4043 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 4044 enum { 4045 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 4046 }; 4047 4048 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 4049 enum { 4050 BPF_F_SKIP_FIELD_MASK = 0xffULL, 4051 BPF_F_USER_STACK = (1ULL << 8), 4052 /* flags used by BPF_FUNC_get_stackid only. */ 4053 BPF_F_FAST_STACK_CMP = (1ULL << 9), 4054 BPF_F_REUSE_STACKID = (1ULL << 10), 4055 /* flags used by BPF_FUNC_get_stack only. */ 4056 BPF_F_USER_BUILD_ID = (1ULL << 11), 4057 }; 4058 4059 /* BPF_FUNC_skb_set_tunnel_key flags. */ 4060 enum { 4061 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 4062 BPF_F_DONT_FRAGMENT = (1ULL << 2), 4063 BPF_F_SEQ_NUMBER = (1ULL << 3), 4064 }; 4065 4066 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 4067 * BPF_FUNC_perf_event_read_value flags. 4068 */ 4069 enum { 4070 BPF_F_INDEX_MASK = 0xffffffffULL, 4071 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 4072 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 4073 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 4074 }; 4075 4076 /* Current network namespace */ 4077 enum { 4078 BPF_F_CURRENT_NETNS = (-1L), 4079 }; 4080 4081 /* BPF_FUNC_csum_level level values. */ 4082 enum { 4083 BPF_CSUM_LEVEL_QUERY, 4084 BPF_CSUM_LEVEL_INC, 4085 BPF_CSUM_LEVEL_DEC, 4086 BPF_CSUM_LEVEL_RESET, 4087 }; 4088 4089 /* BPF_FUNC_skb_adjust_room flags. */ 4090 enum { 4091 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 4092 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 4093 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 4094 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 4095 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 4096 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 4097 }; 4098 4099 enum { 4100 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 4101 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 4102 }; 4103 4104 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 4105 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 4106 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 4107 4108 /* BPF_FUNC_sysctl_get_name flags. */ 4109 enum { 4110 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 4111 }; 4112 4113 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 4114 enum { 4115 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 4116 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 4117 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 4118 */ 4119 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 4120 }; 4121 4122 /* BPF_FUNC_read_branch_records flags. */ 4123 enum { 4124 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 4125 }; 4126 4127 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 4128 * BPF_FUNC_bpf_ringbuf_output flags. 4129 */ 4130 enum { 4131 BPF_RB_NO_WAKEUP = (1ULL << 0), 4132 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 4133 }; 4134 4135 /* BPF_FUNC_bpf_ringbuf_query flags */ 4136 enum { 4137 BPF_RB_AVAIL_DATA = 0, 4138 BPF_RB_RING_SIZE = 1, 4139 BPF_RB_CONS_POS = 2, 4140 BPF_RB_PROD_POS = 3, 4141 }; 4142 4143 /* BPF ring buffer constants */ 4144 enum { 4145 BPF_RINGBUF_BUSY_BIT = (1U << 31), 4146 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 4147 BPF_RINGBUF_HDR_SZ = 8, 4148 }; 4149 4150 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 4151 enum { 4152 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 4153 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 4154 }; 4155 4156 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 4157 enum bpf_adj_room_mode { 4158 BPF_ADJ_ROOM_NET, 4159 BPF_ADJ_ROOM_MAC, 4160 }; 4161 4162 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 4163 enum bpf_hdr_start_off { 4164 BPF_HDR_START_MAC, 4165 BPF_HDR_START_NET, 4166 }; 4167 4168 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 4169 enum bpf_lwt_encap_mode { 4170 BPF_LWT_ENCAP_SEG6, 4171 BPF_LWT_ENCAP_SEG6_INLINE, 4172 BPF_LWT_ENCAP_IP, 4173 }; 4174 4175 /* Flags for bpf_bprm_opts_set helper */ 4176 enum { 4177 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 4178 }; 4179 4180 #define __bpf_md_ptr(type, name) \ 4181 union { \ 4182 type name; \ 4183 __u64 :64; \ 4184 } __attribute__((aligned(8))) 4185 4186 /* user accessible mirror of in-kernel sk_buff. 4187 * new fields can only be added to the end of this structure 4188 */ 4189 struct __sk_buff { 4190 __u32 len; 4191 __u32 pkt_type; 4192 __u32 mark; 4193 __u32 queue_mapping; 4194 __u32 protocol; 4195 __u32 vlan_present; 4196 __u32 vlan_tci; 4197 __u32 vlan_proto; 4198 __u32 priority; 4199 __u32 ingress_ifindex; 4200 __u32 ifindex; 4201 __u32 tc_index; 4202 __u32 cb[5]; 4203 __u32 hash; 4204 __u32 tc_classid; 4205 __u32 data; 4206 __u32 data_end; 4207 __u32 napi_id; 4208 4209 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 4210 __u32 family; 4211 __u32 remote_ip4; /* Stored in network byte order */ 4212 __u32 local_ip4; /* Stored in network byte order */ 4213 __u32 remote_ip6[4]; /* Stored in network byte order */ 4214 __u32 local_ip6[4]; /* Stored in network byte order */ 4215 __u32 remote_port; /* Stored in network byte order */ 4216 __u32 local_port; /* stored in host byte order */ 4217 /* ... here. */ 4218 4219 __u32 data_meta; 4220 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 4221 __u64 tstamp; 4222 __u32 wire_len; 4223 __u32 gso_segs; 4224 __bpf_md_ptr(struct bpf_sock *, sk); 4225 __u32 gso_size; 4226 }; 4227 4228 struct bpf_tunnel_key { 4229 __u32 tunnel_id; 4230 union { 4231 __u32 remote_ipv4; 4232 __u32 remote_ipv6[4]; 4233 }; 4234 __u8 tunnel_tos; 4235 __u8 tunnel_ttl; 4236 __u16 tunnel_ext; /* Padding, future use. */ 4237 __u32 tunnel_label; 4238 }; 4239 4240 /* user accessible mirror of in-kernel xfrm_state. 4241 * new fields can only be added to the end of this structure 4242 */ 4243 struct bpf_xfrm_state { 4244 __u32 reqid; 4245 __u32 spi; /* Stored in network byte order */ 4246 __u16 family; 4247 __u16 ext; /* Padding, future use. */ 4248 union { 4249 __u32 remote_ipv4; /* Stored in network byte order */ 4250 __u32 remote_ipv6[4]; /* Stored in network byte order */ 4251 }; 4252 }; 4253 4254 /* Generic BPF return codes which all BPF program types may support. 4255 * The values are binary compatible with their TC_ACT_* counter-part to 4256 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 4257 * programs. 4258 * 4259 * XDP is handled seprately, see XDP_*. 4260 */ 4261 enum bpf_ret_code { 4262 BPF_OK = 0, 4263 /* 1 reserved */ 4264 BPF_DROP = 2, 4265 /* 3-6 reserved */ 4266 BPF_REDIRECT = 7, 4267 /* >127 are reserved for prog type specific return codes. 4268 * 4269 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 4270 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 4271 * changed and should be routed based on its new L3 header. 4272 * (This is an L3 redirect, as opposed to L2 redirect 4273 * represented by BPF_REDIRECT above). 4274 */ 4275 BPF_LWT_REROUTE = 128, 4276 }; 4277 4278 struct bpf_sock { 4279 __u32 bound_dev_if; 4280 __u32 family; 4281 __u32 type; 4282 __u32 protocol; 4283 __u32 mark; 4284 __u32 priority; 4285 /* IP address also allows 1 and 2 bytes access */ 4286 __u32 src_ip4; 4287 __u32 src_ip6[4]; 4288 __u32 src_port; /* host byte order */ 4289 __u32 dst_port; /* network byte order */ 4290 __u32 dst_ip4; 4291 __u32 dst_ip6[4]; 4292 __u32 state; 4293 __s32 rx_queue_mapping; 4294 }; 4295 4296 struct bpf_tcp_sock { 4297 __u32 snd_cwnd; /* Sending congestion window */ 4298 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 4299 __u32 rtt_min; 4300 __u32 snd_ssthresh; /* Slow start size threshold */ 4301 __u32 rcv_nxt; /* What we want to receive next */ 4302 __u32 snd_nxt; /* Next sequence we send */ 4303 __u32 snd_una; /* First byte we want an ack for */ 4304 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 4305 __u32 ecn_flags; /* ECN status bits. */ 4306 __u32 rate_delivered; /* saved rate sample: packets delivered */ 4307 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 4308 __u32 packets_out; /* Packets which are "in flight" */ 4309 __u32 retrans_out; /* Retransmitted packets out */ 4310 __u32 total_retrans; /* Total retransmits for entire connection */ 4311 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 4312 * total number of segments in. 4313 */ 4314 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 4315 * total number of data segments in. 4316 */ 4317 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 4318 * The total number of segments sent. 4319 */ 4320 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 4321 * total number of data segments sent. 4322 */ 4323 __u32 lost_out; /* Lost packets */ 4324 __u32 sacked_out; /* SACK'd packets */ 4325 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 4326 * sum(delta(rcv_nxt)), or how many bytes 4327 * were acked. 4328 */ 4329 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 4330 * sum(delta(snd_una)), or how many bytes 4331 * were acked. 4332 */ 4333 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 4334 * total number of DSACK blocks received 4335 */ 4336 __u32 delivered; /* Total data packets delivered incl. rexmits */ 4337 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 4338 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 4339 }; 4340 4341 struct bpf_sock_tuple { 4342 union { 4343 struct { 4344 __be32 saddr; 4345 __be32 daddr; 4346 __be16 sport; 4347 __be16 dport; 4348 } ipv4; 4349 struct { 4350 __be32 saddr[4]; 4351 __be32 daddr[4]; 4352 __be16 sport; 4353 __be16 dport; 4354 } ipv6; 4355 }; 4356 }; 4357 4358 struct bpf_xdp_sock { 4359 __u32 queue_id; 4360 }; 4361 4362 #define XDP_PACKET_HEADROOM 256 4363 4364 /* User return codes for XDP prog type. 4365 * A valid XDP program must return one of these defined values. All other 4366 * return codes are reserved for future use. Unknown return codes will 4367 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 4368 */ 4369 enum xdp_action { 4370 XDP_ABORTED = 0, 4371 XDP_DROP, 4372 XDP_PASS, 4373 XDP_TX, 4374 XDP_REDIRECT, 4375 }; 4376 4377 /* user accessible metadata for XDP packet hook 4378 * new fields must be added to the end of this structure 4379 */ 4380 struct xdp_md { 4381 __u32 data; 4382 __u32 data_end; 4383 __u32 data_meta; 4384 /* Below access go through struct xdp_rxq_info */ 4385 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 4386 __u32 rx_queue_index; /* rxq->queue_index */ 4387 4388 __u32 egress_ifindex; /* txq->dev->ifindex */ 4389 }; 4390 4391 /* DEVMAP map-value layout 4392 * 4393 * The struct data-layout of map-value is a configuration interface. 4394 * New members can only be added to the end of this structure. 4395 */ 4396 struct bpf_devmap_val { 4397 __u32 ifindex; /* device index */ 4398 union { 4399 int fd; /* prog fd on map write */ 4400 __u32 id; /* prog id on map read */ 4401 } bpf_prog; 4402 }; 4403 4404 /* CPUMAP map-value layout 4405 * 4406 * The struct data-layout of map-value is a configuration interface. 4407 * New members can only be added to the end of this structure. 4408 */ 4409 struct bpf_cpumap_val { 4410 __u32 qsize; /* queue size to remote target CPU */ 4411 union { 4412 int fd; /* prog fd on map write */ 4413 __u32 id; /* prog id on map read */ 4414 } bpf_prog; 4415 }; 4416 4417 enum sk_action { 4418 SK_DROP = 0, 4419 SK_PASS, 4420 }; 4421 4422 /* user accessible metadata for SK_MSG packet hook, new fields must 4423 * be added to the end of this structure 4424 */ 4425 struct sk_msg_md { 4426 __bpf_md_ptr(void *, data); 4427 __bpf_md_ptr(void *, data_end); 4428 4429 __u32 family; 4430 __u32 remote_ip4; /* Stored in network byte order */ 4431 __u32 local_ip4; /* Stored in network byte order */ 4432 __u32 remote_ip6[4]; /* Stored in network byte order */ 4433 __u32 local_ip6[4]; /* Stored in network byte order */ 4434 __u32 remote_port; /* Stored in network byte order */ 4435 __u32 local_port; /* stored in host byte order */ 4436 __u32 size; /* Total size of sk_msg */ 4437 4438 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 4439 }; 4440 4441 struct sk_reuseport_md { 4442 /* 4443 * Start of directly accessible data. It begins from 4444 * the tcp/udp header. 4445 */ 4446 __bpf_md_ptr(void *, data); 4447 /* End of directly accessible data */ 4448 __bpf_md_ptr(void *, data_end); 4449 /* 4450 * Total length of packet (starting from the tcp/udp header). 4451 * Note that the directly accessible bytes (data_end - data) 4452 * could be less than this "len". Those bytes could be 4453 * indirectly read by a helper "bpf_skb_load_bytes()". 4454 */ 4455 __u32 len; 4456 /* 4457 * Eth protocol in the mac header (network byte order). e.g. 4458 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 4459 */ 4460 __u32 eth_protocol; 4461 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 4462 __u32 bind_inany; /* Is sock bound to an INANY address? */ 4463 __u32 hash; /* A hash of the packet 4 tuples */ 4464 }; 4465 4466 #define BPF_TAG_SIZE 8 4467 4468 struct bpf_prog_info { 4469 __u32 type; 4470 __u32 id; 4471 __u8 tag[BPF_TAG_SIZE]; 4472 __u32 jited_prog_len; 4473 __u32 xlated_prog_len; 4474 __aligned_u64 jited_prog_insns; 4475 __aligned_u64 xlated_prog_insns; 4476 __u64 load_time; /* ns since boottime */ 4477 __u32 created_by_uid; 4478 __u32 nr_map_ids; 4479 __aligned_u64 map_ids; 4480 char name[BPF_OBJ_NAME_LEN]; 4481 __u32 ifindex; 4482 __u32 gpl_compatible:1; 4483 __u32 :31; /* alignment pad */ 4484 __u64 netns_dev; 4485 __u64 netns_ino; 4486 __u32 nr_jited_ksyms; 4487 __u32 nr_jited_func_lens; 4488 __aligned_u64 jited_ksyms; 4489 __aligned_u64 jited_func_lens; 4490 __u32 btf_id; 4491 __u32 func_info_rec_size; 4492 __aligned_u64 func_info; 4493 __u32 nr_func_info; 4494 __u32 nr_line_info; 4495 __aligned_u64 line_info; 4496 __aligned_u64 jited_line_info; 4497 __u32 nr_jited_line_info; 4498 __u32 line_info_rec_size; 4499 __u32 jited_line_info_rec_size; 4500 __u32 nr_prog_tags; 4501 __aligned_u64 prog_tags; 4502 __u64 run_time_ns; 4503 __u64 run_cnt; 4504 } __attribute__((aligned(8))); 4505 4506 struct bpf_map_info { 4507 __u32 type; 4508 __u32 id; 4509 __u32 key_size; 4510 __u32 value_size; 4511 __u32 max_entries; 4512 __u32 map_flags; 4513 char name[BPF_OBJ_NAME_LEN]; 4514 __u32 ifindex; 4515 __u32 btf_vmlinux_value_type_id; 4516 __u64 netns_dev; 4517 __u64 netns_ino; 4518 __u32 btf_id; 4519 __u32 btf_key_type_id; 4520 __u32 btf_value_type_id; 4521 } __attribute__((aligned(8))); 4522 4523 struct bpf_btf_info { 4524 __aligned_u64 btf; 4525 __u32 btf_size; 4526 __u32 id; 4527 __aligned_u64 name; 4528 __u32 name_len; 4529 __u32 kernel_btf; 4530 } __attribute__((aligned(8))); 4531 4532 struct bpf_link_info { 4533 __u32 type; 4534 __u32 id; 4535 __u32 prog_id; 4536 union { 4537 struct { 4538 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 4539 __u32 tp_name_len; /* in/out: tp_name buffer len */ 4540 } raw_tracepoint; 4541 struct { 4542 __u32 attach_type; 4543 } tracing; 4544 struct { 4545 __u64 cgroup_id; 4546 __u32 attach_type; 4547 } cgroup; 4548 struct { 4549 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 4550 __u32 target_name_len; /* in/out: target_name buffer len */ 4551 union { 4552 struct { 4553 __u32 map_id; 4554 } map; 4555 }; 4556 } iter; 4557 struct { 4558 __u32 netns_ino; 4559 __u32 attach_type; 4560 } netns; 4561 struct { 4562 __u32 ifindex; 4563 } xdp; 4564 }; 4565 } __attribute__((aligned(8))); 4566 4567 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 4568 * by user and intended to be used by socket (e.g. to bind to, depends on 4569 * attach type). 4570 */ 4571 struct bpf_sock_addr { 4572 __u32 user_family; /* Allows 4-byte read, but no write. */ 4573 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4574 * Stored in network byte order. 4575 */ 4576 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4577 * Stored in network byte order. 4578 */ 4579 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 4580 * Stored in network byte order 4581 */ 4582 __u32 family; /* Allows 4-byte read, but no write */ 4583 __u32 type; /* Allows 4-byte read, but no write */ 4584 __u32 protocol; /* Allows 4-byte read, but no write */ 4585 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4586 * Stored in network byte order. 4587 */ 4588 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4589 * Stored in network byte order. 4590 */ 4591 __bpf_md_ptr(struct bpf_sock *, sk); 4592 }; 4593 4594 /* User bpf_sock_ops struct to access socket values and specify request ops 4595 * and their replies. 4596 * Some of this fields are in network (bigendian) byte order and may need 4597 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 4598 * New fields can only be added at the end of this structure 4599 */ 4600 struct bpf_sock_ops { 4601 __u32 op; 4602 union { 4603 __u32 args[4]; /* Optionally passed to bpf program */ 4604 __u32 reply; /* Returned by bpf program */ 4605 __u32 replylong[4]; /* Optionally returned by bpf prog */ 4606 }; 4607 __u32 family; 4608 __u32 remote_ip4; /* Stored in network byte order */ 4609 __u32 local_ip4; /* Stored in network byte order */ 4610 __u32 remote_ip6[4]; /* Stored in network byte order */ 4611 __u32 local_ip6[4]; /* Stored in network byte order */ 4612 __u32 remote_port; /* Stored in network byte order */ 4613 __u32 local_port; /* stored in host byte order */ 4614 __u32 is_fullsock; /* Some TCP fields are only valid if 4615 * there is a full socket. If not, the 4616 * fields read as zero. 4617 */ 4618 __u32 snd_cwnd; 4619 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 4620 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 4621 __u32 state; 4622 __u32 rtt_min; 4623 __u32 snd_ssthresh; 4624 __u32 rcv_nxt; 4625 __u32 snd_nxt; 4626 __u32 snd_una; 4627 __u32 mss_cache; 4628 __u32 ecn_flags; 4629 __u32 rate_delivered; 4630 __u32 rate_interval_us; 4631 __u32 packets_out; 4632 __u32 retrans_out; 4633 __u32 total_retrans; 4634 __u32 segs_in; 4635 __u32 data_segs_in; 4636 __u32 segs_out; 4637 __u32 data_segs_out; 4638 __u32 lost_out; 4639 __u32 sacked_out; 4640 __u32 sk_txhash; 4641 __u64 bytes_received; 4642 __u64 bytes_acked; 4643 __bpf_md_ptr(struct bpf_sock *, sk); 4644 /* [skb_data, skb_data_end) covers the whole TCP header. 4645 * 4646 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 4647 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 4648 * header has not been written. 4649 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 4650 * been written so far. 4651 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 4652 * the 3WHS. 4653 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 4654 * the 3WHS. 4655 * 4656 * bpf_load_hdr_opt() can also be used to read a particular option. 4657 */ 4658 __bpf_md_ptr(void *, skb_data); 4659 __bpf_md_ptr(void *, skb_data_end); 4660 __u32 skb_len; /* The total length of a packet. 4661 * It includes the header, options, 4662 * and payload. 4663 */ 4664 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 4665 * an easy way to check for tcp_flags 4666 * without parsing skb_data. 4667 * 4668 * In particular, the skb_tcp_flags 4669 * will still be available in 4670 * BPF_SOCK_OPS_HDR_OPT_LEN even though 4671 * the outgoing header has not 4672 * been written yet. 4673 */ 4674 }; 4675 4676 /* Definitions for bpf_sock_ops_cb_flags */ 4677 enum { 4678 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 4679 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 4680 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 4681 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 4682 /* Call bpf for all received TCP headers. The bpf prog will be 4683 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4684 * 4685 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4686 * for the header option related helpers that will be useful 4687 * to the bpf programs. 4688 * 4689 * It could be used at the client/active side (i.e. connect() side) 4690 * when the server told it that the server was in syncookie 4691 * mode and required the active side to resend the bpf-written 4692 * options. The active side can keep writing the bpf-options until 4693 * it received a valid packet from the server side to confirm 4694 * the earlier packet (and options) has been received. The later 4695 * example patch is using it like this at the active side when the 4696 * server is in syncookie mode. 4697 * 4698 * The bpf prog will usually turn this off in the common cases. 4699 */ 4700 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 4701 /* Call bpf when kernel has received a header option that 4702 * the kernel cannot handle. The bpf prog will be called under 4703 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 4704 * 4705 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4706 * for the header option related helpers that will be useful 4707 * to the bpf programs. 4708 */ 4709 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 4710 /* Call bpf when the kernel is writing header options for the 4711 * outgoing packet. The bpf prog will first be called 4712 * to reserve space in a skb under 4713 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 4714 * the bpf prog will be called to write the header option(s) 4715 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4716 * 4717 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 4718 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 4719 * related helpers that will be useful to the bpf programs. 4720 * 4721 * The kernel gets its chance to reserve space and write 4722 * options first before the BPF program does. 4723 */ 4724 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 4725 /* Mask of all currently supported cb flags */ 4726 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 4727 }; 4728 4729 /* List of known BPF sock_ops operators. 4730 * New entries can only be added at the end 4731 */ 4732 enum { 4733 BPF_SOCK_OPS_VOID, 4734 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 4735 * -1 if default value should be used 4736 */ 4737 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 4738 * window (in packets) or -1 if default 4739 * value should be used 4740 */ 4741 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 4742 * active connection is initialized 4743 */ 4744 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 4745 * active connection is 4746 * established 4747 */ 4748 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 4749 * passive connection is 4750 * established 4751 */ 4752 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 4753 * needs ECN 4754 */ 4755 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 4756 * based on the path and may be 4757 * dependent on the congestion control 4758 * algorithm. In general it indicates 4759 * a congestion threshold. RTTs above 4760 * this indicate congestion 4761 */ 4762 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 4763 * Arg1: value of icsk_retransmits 4764 * Arg2: value of icsk_rto 4765 * Arg3: whether RTO has expired 4766 */ 4767 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 4768 * Arg1: sequence number of 1st byte 4769 * Arg2: # segments 4770 * Arg3: return value of 4771 * tcp_transmit_skb (0 => success) 4772 */ 4773 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 4774 * Arg1: old_state 4775 * Arg2: new_state 4776 */ 4777 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 4778 * socket transition to LISTEN state. 4779 */ 4780 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 4781 */ 4782 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 4783 * It will be called to handle 4784 * the packets received at 4785 * an already established 4786 * connection. 4787 * 4788 * sock_ops->skb_data: 4789 * Referring to the received skb. 4790 * It covers the TCP header only. 4791 * 4792 * bpf_load_hdr_opt() can also 4793 * be used to search for a 4794 * particular option. 4795 */ 4796 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 4797 * header option later in 4798 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4799 * Arg1: bool want_cookie. (in 4800 * writing SYNACK only) 4801 * 4802 * sock_ops->skb_data: 4803 * Not available because no header has 4804 * been written yet. 4805 * 4806 * sock_ops->skb_tcp_flags: 4807 * The tcp_flags of the 4808 * outgoing skb. (e.g. SYN, ACK, FIN). 4809 * 4810 * bpf_reserve_hdr_opt() should 4811 * be used to reserve space. 4812 */ 4813 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 4814 * Arg1: bool want_cookie. (in 4815 * writing SYNACK only) 4816 * 4817 * sock_ops->skb_data: 4818 * Referring to the outgoing skb. 4819 * It covers the TCP header 4820 * that has already been written 4821 * by the kernel and the 4822 * earlier bpf-progs. 4823 * 4824 * sock_ops->skb_tcp_flags: 4825 * The tcp_flags of the outgoing 4826 * skb. (e.g. SYN, ACK, FIN). 4827 * 4828 * bpf_store_hdr_opt() should 4829 * be used to write the 4830 * option. 4831 * 4832 * bpf_load_hdr_opt() can also 4833 * be used to search for a 4834 * particular option that 4835 * has already been written 4836 * by the kernel or the 4837 * earlier bpf-progs. 4838 */ 4839 }; 4840 4841 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 4842 * changes between the TCP and BPF versions. Ideally this should never happen. 4843 * If it does, we need to add code to convert them before calling 4844 * the BPF sock_ops function. 4845 */ 4846 enum { 4847 BPF_TCP_ESTABLISHED = 1, 4848 BPF_TCP_SYN_SENT, 4849 BPF_TCP_SYN_RECV, 4850 BPF_TCP_FIN_WAIT1, 4851 BPF_TCP_FIN_WAIT2, 4852 BPF_TCP_TIME_WAIT, 4853 BPF_TCP_CLOSE, 4854 BPF_TCP_CLOSE_WAIT, 4855 BPF_TCP_LAST_ACK, 4856 BPF_TCP_LISTEN, 4857 BPF_TCP_CLOSING, /* Now a valid state */ 4858 BPF_TCP_NEW_SYN_RECV, 4859 4860 BPF_TCP_MAX_STATES /* Leave at the end! */ 4861 }; 4862 4863 enum { 4864 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 4865 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 4866 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 4867 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 4868 /* Copy the SYN pkt to optval 4869 * 4870 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 4871 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 4872 * to only getting from the saved_syn. It can either get the 4873 * syn packet from: 4874 * 4875 * 1. the just-received SYN packet (only available when writing the 4876 * SYNACK). It will be useful when it is not necessary to 4877 * save the SYN packet for latter use. It is also the only way 4878 * to get the SYN during syncookie mode because the syn 4879 * packet cannot be saved during syncookie. 4880 * 4881 * OR 4882 * 4883 * 2. the earlier saved syn which was done by 4884 * bpf_setsockopt(TCP_SAVE_SYN). 4885 * 4886 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 4887 * SYN packet is obtained. 4888 * 4889 * If the bpf-prog does not need the IP[46] header, the 4890 * bpf-prog can avoid parsing the IP header by using 4891 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 4892 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 4893 * 4894 * >0: Total number of bytes copied 4895 * -ENOSPC: Not enough space in optval. Only optlen number of 4896 * bytes is copied. 4897 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 4898 * is not saved by setsockopt(TCP_SAVE_SYN). 4899 */ 4900 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 4901 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 4902 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 4903 }; 4904 4905 enum { 4906 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 4907 }; 4908 4909 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 4910 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4911 */ 4912 enum { 4913 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 4914 * total option spaces 4915 * required for an established 4916 * sk in order to calculate the 4917 * MSS. No skb is actually 4918 * sent. 4919 */ 4920 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 4921 * when sending a SYN. 4922 */ 4923 }; 4924 4925 struct bpf_perf_event_value { 4926 __u64 counter; 4927 __u64 enabled; 4928 __u64 running; 4929 }; 4930 4931 enum { 4932 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 4933 BPF_DEVCG_ACC_READ = (1ULL << 1), 4934 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 4935 }; 4936 4937 enum { 4938 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 4939 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 4940 }; 4941 4942 struct bpf_cgroup_dev_ctx { 4943 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 4944 __u32 access_type; 4945 __u32 major; 4946 __u32 minor; 4947 }; 4948 4949 struct bpf_raw_tracepoint_args { 4950 __u64 args[0]; 4951 }; 4952 4953 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 4954 * OUTPUT: Do lookup from egress perspective; default is ingress 4955 */ 4956 enum { 4957 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 4958 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 4959 }; 4960 4961 enum { 4962 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 4963 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 4964 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 4965 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 4966 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 4967 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 4968 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 4969 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 4970 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 4971 }; 4972 4973 struct bpf_fib_lookup { 4974 /* input: network family for lookup (AF_INET, AF_INET6) 4975 * output: network family of egress nexthop 4976 */ 4977 __u8 family; 4978 4979 /* set if lookup is to consider L4 data - e.g., FIB rules */ 4980 __u8 l4_protocol; 4981 __be16 sport; 4982 __be16 dport; 4983 4984 /* total length of packet from network header - used for MTU check */ 4985 __u16 tot_len; 4986 4987 /* input: L3 device index for lookup 4988 * output: device index from FIB lookup 4989 */ 4990 __u32 ifindex; 4991 4992 union { 4993 /* inputs to lookup */ 4994 __u8 tos; /* AF_INET */ 4995 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 4996 4997 /* output: metric of fib result (IPv4/IPv6 only) */ 4998 __u32 rt_metric; 4999 }; 5000 5001 union { 5002 __be32 ipv4_src; 5003 __u32 ipv6_src[4]; /* in6_addr; network order */ 5004 }; 5005 5006 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 5007 * network header. output: bpf_fib_lookup sets to gateway address 5008 * if FIB lookup returns gateway route 5009 */ 5010 union { 5011 __be32 ipv4_dst; 5012 __u32 ipv6_dst[4]; /* in6_addr; network order */ 5013 }; 5014 5015 /* output */ 5016 __be16 h_vlan_proto; 5017 __be16 h_vlan_TCI; 5018 __u8 smac[6]; /* ETH_ALEN */ 5019 __u8 dmac[6]; /* ETH_ALEN */ 5020 }; 5021 5022 struct bpf_redir_neigh { 5023 /* network family for lookup (AF_INET, AF_INET6) */ 5024 __u32 nh_family; 5025 /* network address of nexthop; skips fib lookup to find gateway */ 5026 union { 5027 __be32 ipv4_nh; 5028 __u32 ipv6_nh[4]; /* in6_addr; network order */ 5029 }; 5030 }; 5031 5032 enum bpf_task_fd_type { 5033 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 5034 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 5035 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 5036 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 5037 BPF_FD_TYPE_UPROBE, /* filename + offset */ 5038 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 5039 }; 5040 5041 enum { 5042 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 5043 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 5044 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 5045 }; 5046 5047 struct bpf_flow_keys { 5048 __u16 nhoff; 5049 __u16 thoff; 5050 __u16 addr_proto; /* ETH_P_* of valid addrs */ 5051 __u8 is_frag; 5052 __u8 is_first_frag; 5053 __u8 is_encap; 5054 __u8 ip_proto; 5055 __be16 n_proto; 5056 __be16 sport; 5057 __be16 dport; 5058 union { 5059 struct { 5060 __be32 ipv4_src; 5061 __be32 ipv4_dst; 5062 }; 5063 struct { 5064 __u32 ipv6_src[4]; /* in6_addr; network order */ 5065 __u32 ipv6_dst[4]; /* in6_addr; network order */ 5066 }; 5067 }; 5068 __u32 flags; 5069 __be32 flow_label; 5070 }; 5071 5072 struct bpf_func_info { 5073 __u32 insn_off; 5074 __u32 type_id; 5075 }; 5076 5077 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 5078 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 5079 5080 struct bpf_line_info { 5081 __u32 insn_off; 5082 __u32 file_name_off; 5083 __u32 line_off; 5084 __u32 line_col; 5085 }; 5086 5087 struct bpf_spin_lock { 5088 __u32 val; 5089 }; 5090 5091 struct bpf_sysctl { 5092 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 5093 * Allows 1,2,4-byte read, but no write. 5094 */ 5095 __u32 file_pos; /* Sysctl file position to read from, write to. 5096 * Allows 1,2,4-byte read an 4-byte write. 5097 */ 5098 }; 5099 5100 struct bpf_sockopt { 5101 __bpf_md_ptr(struct bpf_sock *, sk); 5102 __bpf_md_ptr(void *, optval); 5103 __bpf_md_ptr(void *, optval_end); 5104 5105 __s32 level; 5106 __s32 optname; 5107 __s32 optlen; 5108 __s32 retval; 5109 }; 5110 5111 struct bpf_pidns_info { 5112 __u32 pid; 5113 __u32 tgid; 5114 }; 5115 5116 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 5117 struct bpf_sk_lookup { 5118 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 5119 5120 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 5121 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 5122 __u32 remote_ip4; /* Network byte order */ 5123 __u32 remote_ip6[4]; /* Network byte order */ 5124 __u32 remote_port; /* Network byte order */ 5125 __u32 local_ip4; /* Network byte order */ 5126 __u32 local_ip6[4]; /* Network byte order */ 5127 __u32 local_port; /* Host byte order */ 5128 }; 5129 5130 /* 5131 * struct btf_ptr is used for typed pointer representation; the 5132 * type id is used to render the pointer data as the appropriate type 5133 * via the bpf_snprintf_btf() helper described above. A flags field - 5134 * potentially to specify additional details about the BTF pointer 5135 * (rather than its mode of display) - is included for future use. 5136 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 5137 */ 5138 struct btf_ptr { 5139 void *ptr; 5140 __u32 type_id; 5141 __u32 flags; /* BTF ptr flags; unused at present. */ 5142 }; 5143 5144 /* 5145 * Flags to control bpf_snprintf_btf() behaviour. 5146 * - BTF_F_COMPACT: no formatting around type information 5147 * - BTF_F_NONAME: no struct/union member names/types 5148 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 5149 * equivalent to %px. 5150 * - BTF_F_ZERO: show zero-valued struct/union members; they 5151 * are not displayed by default 5152 */ 5153 enum { 5154 BTF_F_COMPACT = (1ULL << 0), 5155 BTF_F_NONAME = (1ULL << 1), 5156 BTF_F_PTR_RAW = (1ULL << 2), 5157 BTF_F_ZERO = (1ULL << 3), 5158 }; 5159 5160 #endif /* _UAPI__LINUX_BPF_H__ */ 5161