1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 union bpf_iter_link_info { 85 struct { 86 __u32 map_fd; 87 } map; 88 }; 89 90 /* BPF syscall commands, see bpf(2) man-page for details. */ 91 enum bpf_cmd { 92 BPF_MAP_CREATE, 93 BPF_MAP_LOOKUP_ELEM, 94 BPF_MAP_UPDATE_ELEM, 95 BPF_MAP_DELETE_ELEM, 96 BPF_MAP_GET_NEXT_KEY, 97 BPF_PROG_LOAD, 98 BPF_OBJ_PIN, 99 BPF_OBJ_GET, 100 BPF_PROG_ATTACH, 101 BPF_PROG_DETACH, 102 BPF_PROG_TEST_RUN, 103 BPF_PROG_GET_NEXT_ID, 104 BPF_MAP_GET_NEXT_ID, 105 BPF_PROG_GET_FD_BY_ID, 106 BPF_MAP_GET_FD_BY_ID, 107 BPF_OBJ_GET_INFO_BY_FD, 108 BPF_PROG_QUERY, 109 BPF_RAW_TRACEPOINT_OPEN, 110 BPF_BTF_LOAD, 111 BPF_BTF_GET_FD_BY_ID, 112 BPF_TASK_FD_QUERY, 113 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 114 BPF_MAP_FREEZE, 115 BPF_BTF_GET_NEXT_ID, 116 BPF_MAP_LOOKUP_BATCH, 117 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 118 BPF_MAP_UPDATE_BATCH, 119 BPF_MAP_DELETE_BATCH, 120 BPF_LINK_CREATE, 121 BPF_LINK_UPDATE, 122 BPF_LINK_GET_FD_BY_ID, 123 BPF_LINK_GET_NEXT_ID, 124 BPF_ENABLE_STATS, 125 BPF_ITER_CREATE, 126 BPF_LINK_DETACH, 127 BPF_PROG_BIND_MAP, 128 }; 129 130 enum bpf_map_type { 131 BPF_MAP_TYPE_UNSPEC, 132 BPF_MAP_TYPE_HASH, 133 BPF_MAP_TYPE_ARRAY, 134 BPF_MAP_TYPE_PROG_ARRAY, 135 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 136 BPF_MAP_TYPE_PERCPU_HASH, 137 BPF_MAP_TYPE_PERCPU_ARRAY, 138 BPF_MAP_TYPE_STACK_TRACE, 139 BPF_MAP_TYPE_CGROUP_ARRAY, 140 BPF_MAP_TYPE_LRU_HASH, 141 BPF_MAP_TYPE_LRU_PERCPU_HASH, 142 BPF_MAP_TYPE_LPM_TRIE, 143 BPF_MAP_TYPE_ARRAY_OF_MAPS, 144 BPF_MAP_TYPE_HASH_OF_MAPS, 145 BPF_MAP_TYPE_DEVMAP, 146 BPF_MAP_TYPE_SOCKMAP, 147 BPF_MAP_TYPE_CPUMAP, 148 BPF_MAP_TYPE_XSKMAP, 149 BPF_MAP_TYPE_SOCKHASH, 150 BPF_MAP_TYPE_CGROUP_STORAGE, 151 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 152 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 153 BPF_MAP_TYPE_QUEUE, 154 BPF_MAP_TYPE_STACK, 155 BPF_MAP_TYPE_SK_STORAGE, 156 BPF_MAP_TYPE_DEVMAP_HASH, 157 BPF_MAP_TYPE_STRUCT_OPS, 158 BPF_MAP_TYPE_RINGBUF, 159 BPF_MAP_TYPE_INODE_STORAGE, 160 }; 161 162 /* Note that tracing related programs such as 163 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 164 * are not subject to a stable API since kernel internal data 165 * structures can change from release to release and may 166 * therefore break existing tracing BPF programs. Tracing BPF 167 * programs correspond to /a/ specific kernel which is to be 168 * analyzed, and not /a/ specific kernel /and/ all future ones. 169 */ 170 enum bpf_prog_type { 171 BPF_PROG_TYPE_UNSPEC, 172 BPF_PROG_TYPE_SOCKET_FILTER, 173 BPF_PROG_TYPE_KPROBE, 174 BPF_PROG_TYPE_SCHED_CLS, 175 BPF_PROG_TYPE_SCHED_ACT, 176 BPF_PROG_TYPE_TRACEPOINT, 177 BPF_PROG_TYPE_XDP, 178 BPF_PROG_TYPE_PERF_EVENT, 179 BPF_PROG_TYPE_CGROUP_SKB, 180 BPF_PROG_TYPE_CGROUP_SOCK, 181 BPF_PROG_TYPE_LWT_IN, 182 BPF_PROG_TYPE_LWT_OUT, 183 BPF_PROG_TYPE_LWT_XMIT, 184 BPF_PROG_TYPE_SOCK_OPS, 185 BPF_PROG_TYPE_SK_SKB, 186 BPF_PROG_TYPE_CGROUP_DEVICE, 187 BPF_PROG_TYPE_SK_MSG, 188 BPF_PROG_TYPE_RAW_TRACEPOINT, 189 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 190 BPF_PROG_TYPE_LWT_SEG6LOCAL, 191 BPF_PROG_TYPE_LIRC_MODE2, 192 BPF_PROG_TYPE_SK_REUSEPORT, 193 BPF_PROG_TYPE_FLOW_DISSECTOR, 194 BPF_PROG_TYPE_CGROUP_SYSCTL, 195 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 196 BPF_PROG_TYPE_CGROUP_SOCKOPT, 197 BPF_PROG_TYPE_TRACING, 198 BPF_PROG_TYPE_STRUCT_OPS, 199 BPF_PROG_TYPE_EXT, 200 BPF_PROG_TYPE_LSM, 201 BPF_PROG_TYPE_SK_LOOKUP, 202 }; 203 204 enum bpf_attach_type { 205 BPF_CGROUP_INET_INGRESS, 206 BPF_CGROUP_INET_EGRESS, 207 BPF_CGROUP_INET_SOCK_CREATE, 208 BPF_CGROUP_SOCK_OPS, 209 BPF_SK_SKB_STREAM_PARSER, 210 BPF_SK_SKB_STREAM_VERDICT, 211 BPF_CGROUP_DEVICE, 212 BPF_SK_MSG_VERDICT, 213 BPF_CGROUP_INET4_BIND, 214 BPF_CGROUP_INET6_BIND, 215 BPF_CGROUP_INET4_CONNECT, 216 BPF_CGROUP_INET6_CONNECT, 217 BPF_CGROUP_INET4_POST_BIND, 218 BPF_CGROUP_INET6_POST_BIND, 219 BPF_CGROUP_UDP4_SENDMSG, 220 BPF_CGROUP_UDP6_SENDMSG, 221 BPF_LIRC_MODE2, 222 BPF_FLOW_DISSECTOR, 223 BPF_CGROUP_SYSCTL, 224 BPF_CGROUP_UDP4_RECVMSG, 225 BPF_CGROUP_UDP6_RECVMSG, 226 BPF_CGROUP_GETSOCKOPT, 227 BPF_CGROUP_SETSOCKOPT, 228 BPF_TRACE_RAW_TP, 229 BPF_TRACE_FENTRY, 230 BPF_TRACE_FEXIT, 231 BPF_MODIFY_RETURN, 232 BPF_LSM_MAC, 233 BPF_TRACE_ITER, 234 BPF_CGROUP_INET4_GETPEERNAME, 235 BPF_CGROUP_INET6_GETPEERNAME, 236 BPF_CGROUP_INET4_GETSOCKNAME, 237 BPF_CGROUP_INET6_GETSOCKNAME, 238 BPF_XDP_DEVMAP, 239 BPF_CGROUP_INET_SOCK_RELEASE, 240 BPF_XDP_CPUMAP, 241 BPF_SK_LOOKUP, 242 BPF_XDP, 243 __MAX_BPF_ATTACH_TYPE 244 }; 245 246 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 247 248 enum bpf_link_type { 249 BPF_LINK_TYPE_UNSPEC = 0, 250 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 251 BPF_LINK_TYPE_TRACING = 2, 252 BPF_LINK_TYPE_CGROUP = 3, 253 BPF_LINK_TYPE_ITER = 4, 254 BPF_LINK_TYPE_NETNS = 5, 255 BPF_LINK_TYPE_XDP = 6, 256 257 MAX_BPF_LINK_TYPE, 258 }; 259 260 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 261 * 262 * NONE(default): No further bpf programs allowed in the subtree. 263 * 264 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 265 * the program in this cgroup yields to sub-cgroup program. 266 * 267 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 268 * that cgroup program gets run in addition to the program in this cgroup. 269 * 270 * Only one program is allowed to be attached to a cgroup with 271 * NONE or BPF_F_ALLOW_OVERRIDE flag. 272 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 273 * release old program and attach the new one. Attach flags has to match. 274 * 275 * Multiple programs are allowed to be attached to a cgroup with 276 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 277 * (those that were attached first, run first) 278 * The programs of sub-cgroup are executed first, then programs of 279 * this cgroup and then programs of parent cgroup. 280 * When children program makes decision (like picking TCP CA or sock bind) 281 * parent program has a chance to override it. 282 * 283 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 284 * programs for a cgroup. Though it's possible to replace an old program at 285 * any position by also specifying BPF_F_REPLACE flag and position itself in 286 * replace_bpf_fd attribute. Old program at this position will be released. 287 * 288 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 289 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 290 * Ex1: 291 * cgrp1 (MULTI progs A, B) -> 292 * cgrp2 (OVERRIDE prog C) -> 293 * cgrp3 (MULTI prog D) -> 294 * cgrp4 (OVERRIDE prog E) -> 295 * cgrp5 (NONE prog F) 296 * the event in cgrp5 triggers execution of F,D,A,B in that order. 297 * if prog F is detached, the execution is E,D,A,B 298 * if prog F and D are detached, the execution is E,A,B 299 * if prog F, E and D are detached, the execution is C,A,B 300 * 301 * All eligible programs are executed regardless of return code from 302 * earlier programs. 303 */ 304 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 305 #define BPF_F_ALLOW_MULTI (1U << 1) 306 #define BPF_F_REPLACE (1U << 2) 307 308 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 309 * verifier will perform strict alignment checking as if the kernel 310 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 311 * and NET_IP_ALIGN defined to 2. 312 */ 313 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 314 315 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 316 * verifier will allow any alignment whatsoever. On platforms 317 * with strict alignment requirements for loads ands stores (such 318 * as sparc and mips) the verifier validates that all loads and 319 * stores provably follow this requirement. This flag turns that 320 * checking and enforcement off. 321 * 322 * It is mostly used for testing when we want to validate the 323 * context and memory access aspects of the verifier, but because 324 * of an unaligned access the alignment check would trigger before 325 * the one we are interested in. 326 */ 327 #define BPF_F_ANY_ALIGNMENT (1U << 1) 328 329 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 330 * Verifier does sub-register def/use analysis and identifies instructions whose 331 * def only matters for low 32-bit, high 32-bit is never referenced later 332 * through implicit zero extension. Therefore verifier notifies JIT back-ends 333 * that it is safe to ignore clearing high 32-bit for these instructions. This 334 * saves some back-ends a lot of code-gen. However such optimization is not 335 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 336 * hence hasn't used verifier's analysis result. But, we really want to have a 337 * way to be able to verify the correctness of the described optimization on 338 * x86_64 on which testsuites are frequently exercised. 339 * 340 * So, this flag is introduced. Once it is set, verifier will randomize high 341 * 32-bit for those instructions who has been identified as safe to ignore them. 342 * Then, if verifier is not doing correct analysis, such randomization will 343 * regress tests to expose bugs. 344 */ 345 #define BPF_F_TEST_RND_HI32 (1U << 2) 346 347 /* The verifier internal test flag. Behavior is undefined */ 348 #define BPF_F_TEST_STATE_FREQ (1U << 3) 349 350 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 351 * restrict map and helper usage for such programs. Sleepable BPF programs can 352 * only be attached to hooks where kernel execution context allows sleeping. 353 * Such programs are allowed to use helpers that may sleep like 354 * bpf_copy_from_user(). 355 */ 356 #define BPF_F_SLEEPABLE (1U << 4) 357 358 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 359 * the following extensions: 360 * 361 * insn[0].src_reg: BPF_PSEUDO_MAP_FD 362 * insn[0].imm: map fd 363 * insn[1].imm: 0 364 * insn[0].off: 0 365 * insn[1].off: 0 366 * ldimm64 rewrite: address of map 367 * verifier type: CONST_PTR_TO_MAP 368 */ 369 #define BPF_PSEUDO_MAP_FD 1 370 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE 371 * insn[0].imm: map fd 372 * insn[1].imm: offset into value 373 * insn[0].off: 0 374 * insn[1].off: 0 375 * ldimm64 rewrite: address of map[0]+offset 376 * verifier type: PTR_TO_MAP_VALUE 377 */ 378 #define BPF_PSEUDO_MAP_VALUE 2 379 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 380 * insn[0].imm: kernel btd id of VAR 381 * insn[1].imm: 0 382 * insn[0].off: 0 383 * insn[1].off: 0 384 * ldimm64 rewrite: address of the kernel variable 385 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 386 * is struct/union. 387 */ 388 #define BPF_PSEUDO_BTF_ID 3 389 390 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 391 * offset to another bpf function 392 */ 393 #define BPF_PSEUDO_CALL 1 394 395 /* flags for BPF_MAP_UPDATE_ELEM command */ 396 enum { 397 BPF_ANY = 0, /* create new element or update existing */ 398 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 399 BPF_EXIST = 2, /* update existing element */ 400 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 401 }; 402 403 /* flags for BPF_MAP_CREATE command */ 404 enum { 405 BPF_F_NO_PREALLOC = (1U << 0), 406 /* Instead of having one common LRU list in the 407 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 408 * which can scale and perform better. 409 * Note, the LRU nodes (including free nodes) cannot be moved 410 * across different LRU lists. 411 */ 412 BPF_F_NO_COMMON_LRU = (1U << 1), 413 /* Specify numa node during map creation */ 414 BPF_F_NUMA_NODE = (1U << 2), 415 416 /* Flags for accessing BPF object from syscall side. */ 417 BPF_F_RDONLY = (1U << 3), 418 BPF_F_WRONLY = (1U << 4), 419 420 /* Flag for stack_map, store build_id+offset instead of pointer */ 421 BPF_F_STACK_BUILD_ID = (1U << 5), 422 423 /* Zero-initialize hash function seed. This should only be used for testing. */ 424 BPF_F_ZERO_SEED = (1U << 6), 425 426 /* Flags for accessing BPF object from program side. */ 427 BPF_F_RDONLY_PROG = (1U << 7), 428 BPF_F_WRONLY_PROG = (1U << 8), 429 430 /* Clone map from listener for newly accepted socket */ 431 BPF_F_CLONE = (1U << 9), 432 433 /* Enable memory-mapping BPF map */ 434 BPF_F_MMAPABLE = (1U << 10), 435 436 /* Share perf_event among processes */ 437 BPF_F_PRESERVE_ELEMS = (1U << 11), 438 439 /* Create a map that is suitable to be an inner map with dynamic max entries */ 440 BPF_F_INNER_MAP = (1U << 12), 441 }; 442 443 /* Flags for BPF_PROG_QUERY. */ 444 445 /* Query effective (directly attached + inherited from ancestor cgroups) 446 * programs that will be executed for events within a cgroup. 447 * attach_flags with this flag are returned only for directly attached programs. 448 */ 449 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 450 451 /* Flags for BPF_PROG_TEST_RUN */ 452 453 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 454 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 455 456 /* type for BPF_ENABLE_STATS */ 457 enum bpf_stats_type { 458 /* enabled run_time_ns and run_cnt */ 459 BPF_STATS_RUN_TIME = 0, 460 }; 461 462 enum bpf_stack_build_id_status { 463 /* user space need an empty entry to identify end of a trace */ 464 BPF_STACK_BUILD_ID_EMPTY = 0, 465 /* with valid build_id and offset */ 466 BPF_STACK_BUILD_ID_VALID = 1, 467 /* couldn't get build_id, fallback to ip */ 468 BPF_STACK_BUILD_ID_IP = 2, 469 }; 470 471 #define BPF_BUILD_ID_SIZE 20 472 struct bpf_stack_build_id { 473 __s32 status; 474 unsigned char build_id[BPF_BUILD_ID_SIZE]; 475 union { 476 __u64 offset; 477 __u64 ip; 478 }; 479 }; 480 481 #define BPF_OBJ_NAME_LEN 16U 482 483 union bpf_attr { 484 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 485 __u32 map_type; /* one of enum bpf_map_type */ 486 __u32 key_size; /* size of key in bytes */ 487 __u32 value_size; /* size of value in bytes */ 488 __u32 max_entries; /* max number of entries in a map */ 489 __u32 map_flags; /* BPF_MAP_CREATE related 490 * flags defined above. 491 */ 492 __u32 inner_map_fd; /* fd pointing to the inner map */ 493 __u32 numa_node; /* numa node (effective only if 494 * BPF_F_NUMA_NODE is set). 495 */ 496 char map_name[BPF_OBJ_NAME_LEN]; 497 __u32 map_ifindex; /* ifindex of netdev to create on */ 498 __u32 btf_fd; /* fd pointing to a BTF type data */ 499 __u32 btf_key_type_id; /* BTF type_id of the key */ 500 __u32 btf_value_type_id; /* BTF type_id of the value */ 501 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 502 * struct stored as the 503 * map value 504 */ 505 }; 506 507 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 508 __u32 map_fd; 509 __aligned_u64 key; 510 union { 511 __aligned_u64 value; 512 __aligned_u64 next_key; 513 }; 514 __u64 flags; 515 }; 516 517 struct { /* struct used by BPF_MAP_*_BATCH commands */ 518 __aligned_u64 in_batch; /* start batch, 519 * NULL to start from beginning 520 */ 521 __aligned_u64 out_batch; /* output: next start batch */ 522 __aligned_u64 keys; 523 __aligned_u64 values; 524 __u32 count; /* input/output: 525 * input: # of key/value 526 * elements 527 * output: # of filled elements 528 */ 529 __u32 map_fd; 530 __u64 elem_flags; 531 __u64 flags; 532 } batch; 533 534 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 535 __u32 prog_type; /* one of enum bpf_prog_type */ 536 __u32 insn_cnt; 537 __aligned_u64 insns; 538 __aligned_u64 license; 539 __u32 log_level; /* verbosity level of verifier */ 540 __u32 log_size; /* size of user buffer */ 541 __aligned_u64 log_buf; /* user supplied buffer */ 542 __u32 kern_version; /* not used */ 543 __u32 prog_flags; 544 char prog_name[BPF_OBJ_NAME_LEN]; 545 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 546 /* For some prog types expected attach type must be known at 547 * load time to verify attach type specific parts of prog 548 * (context accesses, allowed helpers, etc). 549 */ 550 __u32 expected_attach_type; 551 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 552 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 553 __aligned_u64 func_info; /* func info */ 554 __u32 func_info_cnt; /* number of bpf_func_info records */ 555 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 556 __aligned_u64 line_info; /* line info */ 557 __u32 line_info_cnt; /* number of bpf_line_info records */ 558 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 559 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 560 }; 561 562 struct { /* anonymous struct used by BPF_OBJ_* commands */ 563 __aligned_u64 pathname; 564 __u32 bpf_fd; 565 __u32 file_flags; 566 }; 567 568 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 569 __u32 target_fd; /* container object to attach to */ 570 __u32 attach_bpf_fd; /* eBPF program to attach */ 571 __u32 attach_type; 572 __u32 attach_flags; 573 __u32 replace_bpf_fd; /* previously attached eBPF 574 * program to replace if 575 * BPF_F_REPLACE is used 576 */ 577 }; 578 579 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 580 __u32 prog_fd; 581 __u32 retval; 582 __u32 data_size_in; /* input: len of data_in */ 583 __u32 data_size_out; /* input/output: len of data_out 584 * returns ENOSPC if data_out 585 * is too small. 586 */ 587 __aligned_u64 data_in; 588 __aligned_u64 data_out; 589 __u32 repeat; 590 __u32 duration; 591 __u32 ctx_size_in; /* input: len of ctx_in */ 592 __u32 ctx_size_out; /* input/output: len of ctx_out 593 * returns ENOSPC if ctx_out 594 * is too small. 595 */ 596 __aligned_u64 ctx_in; 597 __aligned_u64 ctx_out; 598 __u32 flags; 599 __u32 cpu; 600 } test; 601 602 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 603 union { 604 __u32 start_id; 605 __u32 prog_id; 606 __u32 map_id; 607 __u32 btf_id; 608 __u32 link_id; 609 }; 610 __u32 next_id; 611 __u32 open_flags; 612 }; 613 614 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 615 __u32 bpf_fd; 616 __u32 info_len; 617 __aligned_u64 info; 618 } info; 619 620 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 621 __u32 target_fd; /* container object to query */ 622 __u32 attach_type; 623 __u32 query_flags; 624 __u32 attach_flags; 625 __aligned_u64 prog_ids; 626 __u32 prog_cnt; 627 } query; 628 629 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 630 __u64 name; 631 __u32 prog_fd; 632 } raw_tracepoint; 633 634 struct { /* anonymous struct for BPF_BTF_LOAD */ 635 __aligned_u64 btf; 636 __aligned_u64 btf_log_buf; 637 __u32 btf_size; 638 __u32 btf_log_size; 639 __u32 btf_log_level; 640 }; 641 642 struct { 643 __u32 pid; /* input: pid */ 644 __u32 fd; /* input: fd */ 645 __u32 flags; /* input: flags */ 646 __u32 buf_len; /* input/output: buf len */ 647 __aligned_u64 buf; /* input/output: 648 * tp_name for tracepoint 649 * symbol for kprobe 650 * filename for uprobe 651 */ 652 __u32 prog_id; /* output: prod_id */ 653 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 654 __u64 probe_offset; /* output: probe_offset */ 655 __u64 probe_addr; /* output: probe_addr */ 656 } task_fd_query; 657 658 struct { /* struct used by BPF_LINK_CREATE command */ 659 __u32 prog_fd; /* eBPF program to attach */ 660 union { 661 __u32 target_fd; /* object to attach to */ 662 __u32 target_ifindex; /* target ifindex */ 663 }; 664 __u32 attach_type; /* attach type */ 665 __u32 flags; /* extra flags */ 666 union { 667 __u32 target_btf_id; /* btf_id of target to attach to */ 668 struct { 669 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 670 __u32 iter_info_len; /* iter_info length */ 671 }; 672 }; 673 } link_create; 674 675 struct { /* struct used by BPF_LINK_UPDATE command */ 676 __u32 link_fd; /* link fd */ 677 /* new program fd to update link with */ 678 __u32 new_prog_fd; 679 __u32 flags; /* extra flags */ 680 /* expected link's program fd; is specified only if 681 * BPF_F_REPLACE flag is set in flags */ 682 __u32 old_prog_fd; 683 } link_update; 684 685 struct { 686 __u32 link_fd; 687 } link_detach; 688 689 struct { /* struct used by BPF_ENABLE_STATS command */ 690 __u32 type; 691 } enable_stats; 692 693 struct { /* struct used by BPF_ITER_CREATE command */ 694 __u32 link_fd; 695 __u32 flags; 696 } iter_create; 697 698 struct { /* struct used by BPF_PROG_BIND_MAP command */ 699 __u32 prog_fd; 700 __u32 map_fd; 701 __u32 flags; /* extra flags */ 702 } prog_bind_map; 703 704 } __attribute__((aligned(8))); 705 706 /* The description below is an attempt at providing documentation to eBPF 707 * developers about the multiple available eBPF helper functions. It can be 708 * parsed and used to produce a manual page. The workflow is the following, 709 * and requires the rst2man utility: 710 * 711 * $ ./scripts/bpf_helpers_doc.py \ 712 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 713 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 714 * $ man /tmp/bpf-helpers.7 715 * 716 * Note that in order to produce this external documentation, some RST 717 * formatting is used in the descriptions to get "bold" and "italics" in 718 * manual pages. Also note that the few trailing white spaces are 719 * intentional, removing them would break paragraphs for rst2man. 720 * 721 * Start of BPF helper function descriptions: 722 * 723 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 724 * Description 725 * Perform a lookup in *map* for an entry associated to *key*. 726 * Return 727 * Map value associated to *key*, or **NULL** if no entry was 728 * found. 729 * 730 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 731 * Description 732 * Add or update the value of the entry associated to *key* in 733 * *map* with *value*. *flags* is one of: 734 * 735 * **BPF_NOEXIST** 736 * The entry for *key* must not exist in the map. 737 * **BPF_EXIST** 738 * The entry for *key* must already exist in the map. 739 * **BPF_ANY** 740 * No condition on the existence of the entry for *key*. 741 * 742 * Flag value **BPF_NOEXIST** cannot be used for maps of types 743 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 744 * elements always exist), the helper would return an error. 745 * Return 746 * 0 on success, or a negative error in case of failure. 747 * 748 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 749 * Description 750 * Delete entry with *key* from *map*. 751 * Return 752 * 0 on success, or a negative error in case of failure. 753 * 754 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 755 * Description 756 * For tracing programs, safely attempt to read *size* bytes from 757 * kernel space address *unsafe_ptr* and store the data in *dst*. 758 * 759 * Generally, use **bpf_probe_read_user**\ () or 760 * **bpf_probe_read_kernel**\ () instead. 761 * Return 762 * 0 on success, or a negative error in case of failure. 763 * 764 * u64 bpf_ktime_get_ns(void) 765 * Description 766 * Return the time elapsed since system boot, in nanoseconds. 767 * Does not include time the system was suspended. 768 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 769 * Return 770 * Current *ktime*. 771 * 772 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 773 * Description 774 * This helper is a "printk()-like" facility for debugging. It 775 * prints a message defined by format *fmt* (of size *fmt_size*) 776 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 777 * available. It can take up to three additional **u64** 778 * arguments (as an eBPF helpers, the total number of arguments is 779 * limited to five). 780 * 781 * Each time the helper is called, it appends a line to the trace. 782 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 783 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 784 * The format of the trace is customizable, and the exact output 785 * one will get depends on the options set in 786 * *\/sys/kernel/debug/tracing/trace_options* (see also the 787 * *README* file under the same directory). However, it usually 788 * defaults to something like: 789 * 790 * :: 791 * 792 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 793 * 794 * In the above: 795 * 796 * * ``telnet`` is the name of the current task. 797 * * ``470`` is the PID of the current task. 798 * * ``001`` is the CPU number on which the task is 799 * running. 800 * * In ``.N..``, each character refers to a set of 801 * options (whether irqs are enabled, scheduling 802 * options, whether hard/softirqs are running, level of 803 * preempt_disabled respectively). **N** means that 804 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 805 * are set. 806 * * ``419421.045894`` is a timestamp. 807 * * ``0x00000001`` is a fake value used by BPF for the 808 * instruction pointer register. 809 * * ``<formatted msg>`` is the message formatted with 810 * *fmt*. 811 * 812 * The conversion specifiers supported by *fmt* are similar, but 813 * more limited than for printk(). They are **%d**, **%i**, 814 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 815 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 816 * of field, padding with zeroes, etc.) is available, and the 817 * helper will return **-EINVAL** (but print nothing) if it 818 * encounters an unknown specifier. 819 * 820 * Also, note that **bpf_trace_printk**\ () is slow, and should 821 * only be used for debugging purposes. For this reason, a notice 822 * block (spanning several lines) is printed to kernel logs and 823 * states that the helper should not be used "for production use" 824 * the first time this helper is used (or more precisely, when 825 * **trace_printk**\ () buffers are allocated). For passing values 826 * to user space, perf events should be preferred. 827 * Return 828 * The number of bytes written to the buffer, or a negative error 829 * in case of failure. 830 * 831 * u32 bpf_get_prandom_u32(void) 832 * Description 833 * Get a pseudo-random number. 834 * 835 * From a security point of view, this helper uses its own 836 * pseudo-random internal state, and cannot be used to infer the 837 * seed of other random functions in the kernel. However, it is 838 * essential to note that the generator used by the helper is not 839 * cryptographically secure. 840 * Return 841 * A random 32-bit unsigned value. 842 * 843 * u32 bpf_get_smp_processor_id(void) 844 * Description 845 * Get the SMP (symmetric multiprocessing) processor id. Note that 846 * all programs run with preemption disabled, which means that the 847 * SMP processor id is stable during all the execution of the 848 * program. 849 * Return 850 * The SMP id of the processor running the program. 851 * 852 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 853 * Description 854 * Store *len* bytes from address *from* into the packet 855 * associated to *skb*, at *offset*. *flags* are a combination of 856 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 857 * checksum for the packet after storing the bytes) and 858 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 859 * **->swhash** and *skb*\ **->l4hash** to 0). 860 * 861 * A call to this helper is susceptible to change the underlying 862 * packet buffer. Therefore, at load time, all checks on pointers 863 * previously done by the verifier are invalidated and must be 864 * performed again, if the helper is used in combination with 865 * direct packet access. 866 * Return 867 * 0 on success, or a negative error in case of failure. 868 * 869 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 870 * Description 871 * Recompute the layer 3 (e.g. IP) checksum for the packet 872 * associated to *skb*. Computation is incremental, so the helper 873 * must know the former value of the header field that was 874 * modified (*from*), the new value of this field (*to*), and the 875 * number of bytes (2 or 4) for this field, stored in *size*. 876 * Alternatively, it is possible to store the difference between 877 * the previous and the new values of the header field in *to*, by 878 * setting *from* and *size* to 0. For both methods, *offset* 879 * indicates the location of the IP checksum within the packet. 880 * 881 * This helper works in combination with **bpf_csum_diff**\ (), 882 * which does not update the checksum in-place, but offers more 883 * flexibility and can handle sizes larger than 2 or 4 for the 884 * checksum to update. 885 * 886 * A call to this helper is susceptible to change the underlying 887 * packet buffer. Therefore, at load time, all checks on pointers 888 * previously done by the verifier are invalidated and must be 889 * performed again, if the helper is used in combination with 890 * direct packet access. 891 * Return 892 * 0 on success, or a negative error in case of failure. 893 * 894 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 895 * Description 896 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 897 * packet associated to *skb*. Computation is incremental, so the 898 * helper must know the former value of the header field that was 899 * modified (*from*), the new value of this field (*to*), and the 900 * number of bytes (2 or 4) for this field, stored on the lowest 901 * four bits of *flags*. Alternatively, it is possible to store 902 * the difference between the previous and the new values of the 903 * header field in *to*, by setting *from* and the four lowest 904 * bits of *flags* to 0. For both methods, *offset* indicates the 905 * location of the IP checksum within the packet. In addition to 906 * the size of the field, *flags* can be added (bitwise OR) actual 907 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 908 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 909 * for updates resulting in a null checksum the value is set to 910 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 911 * the checksum is to be computed against a pseudo-header. 912 * 913 * This helper works in combination with **bpf_csum_diff**\ (), 914 * which does not update the checksum in-place, but offers more 915 * flexibility and can handle sizes larger than 2 or 4 for the 916 * checksum to update. 917 * 918 * A call to this helper is susceptible to change the underlying 919 * packet buffer. Therefore, at load time, all checks on pointers 920 * previously done by the verifier are invalidated and must be 921 * performed again, if the helper is used in combination with 922 * direct packet access. 923 * Return 924 * 0 on success, or a negative error in case of failure. 925 * 926 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 927 * Description 928 * This special helper is used to trigger a "tail call", or in 929 * other words, to jump into another eBPF program. The same stack 930 * frame is used (but values on stack and in registers for the 931 * caller are not accessible to the callee). This mechanism allows 932 * for program chaining, either for raising the maximum number of 933 * available eBPF instructions, or to execute given programs in 934 * conditional blocks. For security reasons, there is an upper 935 * limit to the number of successive tail calls that can be 936 * performed. 937 * 938 * Upon call of this helper, the program attempts to jump into a 939 * program referenced at index *index* in *prog_array_map*, a 940 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 941 * *ctx*, a pointer to the context. 942 * 943 * If the call succeeds, the kernel immediately runs the first 944 * instruction of the new program. This is not a function call, 945 * and it never returns to the previous program. If the call 946 * fails, then the helper has no effect, and the caller continues 947 * to run its subsequent instructions. A call can fail if the 948 * destination program for the jump does not exist (i.e. *index* 949 * is superior to the number of entries in *prog_array_map*), or 950 * if the maximum number of tail calls has been reached for this 951 * chain of programs. This limit is defined in the kernel by the 952 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 953 * which is currently set to 32. 954 * Return 955 * 0 on success, or a negative error in case of failure. 956 * 957 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 958 * Description 959 * Clone and redirect the packet associated to *skb* to another 960 * net device of index *ifindex*. Both ingress and egress 961 * interfaces can be used for redirection. The **BPF_F_INGRESS** 962 * value in *flags* is used to make the distinction (ingress path 963 * is selected if the flag is present, egress path otherwise). 964 * This is the only flag supported for now. 965 * 966 * In comparison with **bpf_redirect**\ () helper, 967 * **bpf_clone_redirect**\ () has the associated cost of 968 * duplicating the packet buffer, but this can be executed out of 969 * the eBPF program. Conversely, **bpf_redirect**\ () is more 970 * efficient, but it is handled through an action code where the 971 * redirection happens only after the eBPF program has returned. 972 * 973 * A call to this helper is susceptible to change the underlying 974 * packet buffer. Therefore, at load time, all checks on pointers 975 * previously done by the verifier are invalidated and must be 976 * performed again, if the helper is used in combination with 977 * direct packet access. 978 * Return 979 * 0 on success, or a negative error in case of failure. Positive 980 * error indicates a potential drop or congestion in the target 981 * device. The particular positive error codes are not defined. 982 * 983 * u64 bpf_get_current_pid_tgid(void) 984 * Return 985 * A 64-bit integer containing the current tgid and pid, and 986 * created as such: 987 * *current_task*\ **->tgid << 32 \|** 988 * *current_task*\ **->pid**. 989 * 990 * u64 bpf_get_current_uid_gid(void) 991 * Return 992 * A 64-bit integer containing the current GID and UID, and 993 * created as such: *current_gid* **<< 32 \|** *current_uid*. 994 * 995 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 996 * Description 997 * Copy the **comm** attribute of the current task into *buf* of 998 * *size_of_buf*. The **comm** attribute contains the name of 999 * the executable (excluding the path) for the current task. The 1000 * *size_of_buf* must be strictly positive. On success, the 1001 * helper makes sure that the *buf* is NUL-terminated. On failure, 1002 * it is filled with zeroes. 1003 * Return 1004 * 0 on success, or a negative error in case of failure. 1005 * 1006 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1007 * Description 1008 * Retrieve the classid for the current task, i.e. for the net_cls 1009 * cgroup to which *skb* belongs. 1010 * 1011 * This helper can be used on TC egress path, but not on ingress. 1012 * 1013 * The net_cls cgroup provides an interface to tag network packets 1014 * based on a user-provided identifier for all traffic coming from 1015 * the tasks belonging to the related cgroup. See also the related 1016 * kernel documentation, available from the Linux sources in file 1017 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1018 * 1019 * The Linux kernel has two versions for cgroups: there are 1020 * cgroups v1 and cgroups v2. Both are available to users, who can 1021 * use a mixture of them, but note that the net_cls cgroup is for 1022 * cgroup v1 only. This makes it incompatible with BPF programs 1023 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1024 * only hold data for one version of cgroups at a time). 1025 * 1026 * This helper is only available is the kernel was compiled with 1027 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1028 * "**y**" or to "**m**". 1029 * Return 1030 * The classid, or 0 for the default unconfigured classid. 1031 * 1032 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1033 * Description 1034 * Push a *vlan_tci* (VLAN tag control information) of protocol 1035 * *vlan_proto* to the packet associated to *skb*, then update 1036 * the checksum. Note that if *vlan_proto* is different from 1037 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1038 * be **ETH_P_8021Q**. 1039 * 1040 * A call to this helper is susceptible to change the underlying 1041 * packet buffer. Therefore, at load time, all checks on pointers 1042 * previously done by the verifier are invalidated and must be 1043 * performed again, if the helper is used in combination with 1044 * direct packet access. 1045 * Return 1046 * 0 on success, or a negative error in case of failure. 1047 * 1048 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1049 * Description 1050 * Pop a VLAN header from the packet associated to *skb*. 1051 * 1052 * A call to this helper is susceptible to change the underlying 1053 * packet buffer. Therefore, at load time, all checks on pointers 1054 * previously done by the verifier are invalidated and must be 1055 * performed again, if the helper is used in combination with 1056 * direct packet access. 1057 * Return 1058 * 0 on success, or a negative error in case of failure. 1059 * 1060 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1061 * Description 1062 * Get tunnel metadata. This helper takes a pointer *key* to an 1063 * empty **struct bpf_tunnel_key** of **size**, that will be 1064 * filled with tunnel metadata for the packet associated to *skb*. 1065 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1066 * indicates that the tunnel is based on IPv6 protocol instead of 1067 * IPv4. 1068 * 1069 * The **struct bpf_tunnel_key** is an object that generalizes the 1070 * principal parameters used by various tunneling protocols into a 1071 * single struct. This way, it can be used to easily make a 1072 * decision based on the contents of the encapsulation header, 1073 * "summarized" in this struct. In particular, it holds the IP 1074 * address of the remote end (IPv4 or IPv6, depending on the case) 1075 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1076 * this struct exposes the *key*\ **->tunnel_id**, which is 1077 * generally mapped to a VNI (Virtual Network Identifier), making 1078 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1079 * () helper. 1080 * 1081 * Let's imagine that the following code is part of a program 1082 * attached to the TC ingress interface, on one end of a GRE 1083 * tunnel, and is supposed to filter out all messages coming from 1084 * remote ends with IPv4 address other than 10.0.0.1: 1085 * 1086 * :: 1087 * 1088 * int ret; 1089 * struct bpf_tunnel_key key = {}; 1090 * 1091 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1092 * if (ret < 0) 1093 * return TC_ACT_SHOT; // drop packet 1094 * 1095 * if (key.remote_ipv4 != 0x0a000001) 1096 * return TC_ACT_SHOT; // drop packet 1097 * 1098 * return TC_ACT_OK; // accept packet 1099 * 1100 * This interface can also be used with all encapsulation devices 1101 * that can operate in "collect metadata" mode: instead of having 1102 * one network device per specific configuration, the "collect 1103 * metadata" mode only requires a single device where the 1104 * configuration can be extracted from this helper. 1105 * 1106 * This can be used together with various tunnels such as VXLan, 1107 * Geneve, GRE or IP in IP (IPIP). 1108 * Return 1109 * 0 on success, or a negative error in case of failure. 1110 * 1111 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1112 * Description 1113 * Populate tunnel metadata for packet associated to *skb.* The 1114 * tunnel metadata is set to the contents of *key*, of *size*. The 1115 * *flags* can be set to a combination of the following values: 1116 * 1117 * **BPF_F_TUNINFO_IPV6** 1118 * Indicate that the tunnel is based on IPv6 protocol 1119 * instead of IPv4. 1120 * **BPF_F_ZERO_CSUM_TX** 1121 * For IPv4 packets, add a flag to tunnel metadata 1122 * indicating that checksum computation should be skipped 1123 * and checksum set to zeroes. 1124 * **BPF_F_DONT_FRAGMENT** 1125 * Add a flag to tunnel metadata indicating that the 1126 * packet should not be fragmented. 1127 * **BPF_F_SEQ_NUMBER** 1128 * Add a flag to tunnel metadata indicating that a 1129 * sequence number should be added to tunnel header before 1130 * sending the packet. This flag was added for GRE 1131 * encapsulation, but might be used with other protocols 1132 * as well in the future. 1133 * 1134 * Here is a typical usage on the transmit path: 1135 * 1136 * :: 1137 * 1138 * struct bpf_tunnel_key key; 1139 * populate key ... 1140 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1141 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1142 * 1143 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1144 * helper for additional information. 1145 * Return 1146 * 0 on success, or a negative error in case of failure. 1147 * 1148 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1149 * Description 1150 * Read the value of a perf event counter. This helper relies on a 1151 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1152 * the perf event counter is selected when *map* is updated with 1153 * perf event file descriptors. The *map* is an array whose size 1154 * is the number of available CPUs, and each cell contains a value 1155 * relative to one CPU. The value to retrieve is indicated by 1156 * *flags*, that contains the index of the CPU to look up, masked 1157 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1158 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1159 * current CPU should be retrieved. 1160 * 1161 * Note that before Linux 4.13, only hardware perf event can be 1162 * retrieved. 1163 * 1164 * Also, be aware that the newer helper 1165 * **bpf_perf_event_read_value**\ () is recommended over 1166 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1167 * quirks where error and counter value are used as a return code 1168 * (which is wrong to do since ranges may overlap). This issue is 1169 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1170 * time provides more features over the **bpf_perf_event_read**\ 1171 * () interface. Please refer to the description of 1172 * **bpf_perf_event_read_value**\ () for details. 1173 * Return 1174 * The value of the perf event counter read from the map, or a 1175 * negative error code in case of failure. 1176 * 1177 * long bpf_redirect(u32 ifindex, u64 flags) 1178 * Description 1179 * Redirect the packet to another net device of index *ifindex*. 1180 * This helper is somewhat similar to **bpf_clone_redirect**\ 1181 * (), except that the packet is not cloned, which provides 1182 * increased performance. 1183 * 1184 * Except for XDP, both ingress and egress interfaces can be used 1185 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1186 * to make the distinction (ingress path is selected if the flag 1187 * is present, egress path otherwise). Currently, XDP only 1188 * supports redirection to the egress interface, and accepts no 1189 * flag at all. 1190 * 1191 * The same effect can also be attained with the more generic 1192 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1193 * redirect target instead of providing it directly to the helper. 1194 * Return 1195 * For XDP, the helper returns **XDP_REDIRECT** on success or 1196 * **XDP_ABORTED** on error. For other program types, the values 1197 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1198 * error. 1199 * 1200 * u32 bpf_get_route_realm(struct sk_buff *skb) 1201 * Description 1202 * Retrieve the realm or the route, that is to say the 1203 * **tclassid** field of the destination for the *skb*. The 1204 * identifier retrieved is a user-provided tag, similar to the 1205 * one used with the net_cls cgroup (see description for 1206 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1207 * held by a route (a destination entry), not by a task. 1208 * 1209 * Retrieving this identifier works with the clsact TC egress hook 1210 * (see also **tc-bpf(8)**), or alternatively on conventional 1211 * classful egress qdiscs, but not on TC ingress path. In case of 1212 * clsact TC egress hook, this has the advantage that, internally, 1213 * the destination entry has not been dropped yet in the transmit 1214 * path. Therefore, the destination entry does not need to be 1215 * artificially held via **netif_keep_dst**\ () for a classful 1216 * qdisc until the *skb* is freed. 1217 * 1218 * This helper is available only if the kernel was compiled with 1219 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1220 * Return 1221 * The realm of the route for the packet associated to *skb*, or 0 1222 * if none was found. 1223 * 1224 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1225 * Description 1226 * Write raw *data* blob into a special BPF perf event held by 1227 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1228 * event must have the following attributes: **PERF_SAMPLE_RAW** 1229 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1230 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1231 * 1232 * The *flags* are used to indicate the index in *map* for which 1233 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1234 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1235 * to indicate that the index of the current CPU core should be 1236 * used. 1237 * 1238 * The value to write, of *size*, is passed through eBPF stack and 1239 * pointed by *data*. 1240 * 1241 * The context of the program *ctx* needs also be passed to the 1242 * helper. 1243 * 1244 * On user space, a program willing to read the values needs to 1245 * call **perf_event_open**\ () on the perf event (either for 1246 * one or for all CPUs) and to store the file descriptor into the 1247 * *map*. This must be done before the eBPF program can send data 1248 * into it. An example is available in file 1249 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1250 * tree (the eBPF program counterpart is in 1251 * *samples/bpf/trace_output_kern.c*). 1252 * 1253 * **bpf_perf_event_output**\ () achieves better performance 1254 * than **bpf_trace_printk**\ () for sharing data with user 1255 * space, and is much better suitable for streaming data from eBPF 1256 * programs. 1257 * 1258 * Note that this helper is not restricted to tracing use cases 1259 * and can be used with programs attached to TC or XDP as well, 1260 * where it allows for passing data to user space listeners. Data 1261 * can be: 1262 * 1263 * * Only custom structs, 1264 * * Only the packet payload, or 1265 * * A combination of both. 1266 * Return 1267 * 0 on success, or a negative error in case of failure. 1268 * 1269 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1270 * Description 1271 * This helper was provided as an easy way to load data from a 1272 * packet. It can be used to load *len* bytes from *offset* from 1273 * the packet associated to *skb*, into the buffer pointed by 1274 * *to*. 1275 * 1276 * Since Linux 4.7, usage of this helper has mostly been replaced 1277 * by "direct packet access", enabling packet data to be 1278 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1279 * pointing respectively to the first byte of packet data and to 1280 * the byte after the last byte of packet data. However, it 1281 * remains useful if one wishes to read large quantities of data 1282 * at once from a packet into the eBPF stack. 1283 * Return 1284 * 0 on success, or a negative error in case of failure. 1285 * 1286 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1287 * Description 1288 * Walk a user or a kernel stack and return its id. To achieve 1289 * this, the helper needs *ctx*, which is a pointer to the context 1290 * on which the tracing program is executed, and a pointer to a 1291 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1292 * 1293 * The last argument, *flags*, holds the number of stack frames to 1294 * skip (from 0 to 255), masked with 1295 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1296 * a combination of the following flags: 1297 * 1298 * **BPF_F_USER_STACK** 1299 * Collect a user space stack instead of a kernel stack. 1300 * **BPF_F_FAST_STACK_CMP** 1301 * Compare stacks by hash only. 1302 * **BPF_F_REUSE_STACKID** 1303 * If two different stacks hash into the same *stackid*, 1304 * discard the old one. 1305 * 1306 * The stack id retrieved is a 32 bit long integer handle which 1307 * can be further combined with other data (including other stack 1308 * ids) and used as a key into maps. This can be useful for 1309 * generating a variety of graphs (such as flame graphs or off-cpu 1310 * graphs). 1311 * 1312 * For walking a stack, this helper is an improvement over 1313 * **bpf_probe_read**\ (), which can be used with unrolled loops 1314 * but is not efficient and consumes a lot of eBPF instructions. 1315 * Instead, **bpf_get_stackid**\ () can collect up to 1316 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1317 * this limit can be controlled with the **sysctl** program, and 1318 * that it should be manually increased in order to profile long 1319 * user stacks (such as stacks for Java programs). To do so, use: 1320 * 1321 * :: 1322 * 1323 * # sysctl kernel.perf_event_max_stack=<new value> 1324 * Return 1325 * The positive or null stack id on success, or a negative error 1326 * in case of failure. 1327 * 1328 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1329 * Description 1330 * Compute a checksum difference, from the raw buffer pointed by 1331 * *from*, of length *from_size* (that must be a multiple of 4), 1332 * towards the raw buffer pointed by *to*, of size *to_size* 1333 * (same remark). An optional *seed* can be added to the value 1334 * (this can be cascaded, the seed may come from a previous call 1335 * to the helper). 1336 * 1337 * This is flexible enough to be used in several ways: 1338 * 1339 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1340 * checksum, it can be used when pushing new data. 1341 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1342 * checksum, it can be used when removing data from a packet. 1343 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1344 * can be used to compute a diff. Note that *from_size* and 1345 * *to_size* do not need to be equal. 1346 * 1347 * This helper can be used in combination with 1348 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1349 * which one can feed in the difference computed with 1350 * **bpf_csum_diff**\ (). 1351 * Return 1352 * The checksum result, or a negative error code in case of 1353 * failure. 1354 * 1355 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1356 * Description 1357 * Retrieve tunnel options metadata for the packet associated to 1358 * *skb*, and store the raw tunnel option data to the buffer *opt* 1359 * of *size*. 1360 * 1361 * This helper can be used with encapsulation devices that can 1362 * operate in "collect metadata" mode (please refer to the related 1363 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1364 * more details). A particular example where this can be used is 1365 * in combination with the Geneve encapsulation protocol, where it 1366 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1367 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1368 * the eBPF program. This allows for full customization of these 1369 * headers. 1370 * Return 1371 * The size of the option data retrieved. 1372 * 1373 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1374 * Description 1375 * Set tunnel options metadata for the packet associated to *skb* 1376 * to the option data contained in the raw buffer *opt* of *size*. 1377 * 1378 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1379 * helper for additional information. 1380 * Return 1381 * 0 on success, or a negative error in case of failure. 1382 * 1383 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1384 * Description 1385 * Change the protocol of the *skb* to *proto*. Currently 1386 * supported are transition from IPv4 to IPv6, and from IPv6 to 1387 * IPv4. The helper takes care of the groundwork for the 1388 * transition, including resizing the socket buffer. The eBPF 1389 * program is expected to fill the new headers, if any, via 1390 * **skb_store_bytes**\ () and to recompute the checksums with 1391 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1392 * (). The main case for this helper is to perform NAT64 1393 * operations out of an eBPF program. 1394 * 1395 * Internally, the GSO type is marked as dodgy so that headers are 1396 * checked and segments are recalculated by the GSO/GRO engine. 1397 * The size for GSO target is adapted as well. 1398 * 1399 * All values for *flags* are reserved for future usage, and must 1400 * be left at zero. 1401 * 1402 * A call to this helper is susceptible to change the underlying 1403 * packet buffer. Therefore, at load time, all checks on pointers 1404 * previously done by the verifier are invalidated and must be 1405 * performed again, if the helper is used in combination with 1406 * direct packet access. 1407 * Return 1408 * 0 on success, or a negative error in case of failure. 1409 * 1410 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 1411 * Description 1412 * Change the packet type for the packet associated to *skb*. This 1413 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1414 * the eBPF program does not have a write access to *skb*\ 1415 * **->pkt_type** beside this helper. Using a helper here allows 1416 * for graceful handling of errors. 1417 * 1418 * The major use case is to change incoming *skb*s to 1419 * **PACKET_HOST** in a programmatic way instead of having to 1420 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1421 * example. 1422 * 1423 * Note that *type* only allows certain values. At this time, they 1424 * are: 1425 * 1426 * **PACKET_HOST** 1427 * Packet is for us. 1428 * **PACKET_BROADCAST** 1429 * Send packet to all. 1430 * **PACKET_MULTICAST** 1431 * Send packet to group. 1432 * **PACKET_OTHERHOST** 1433 * Send packet to someone else. 1434 * Return 1435 * 0 on success, or a negative error in case of failure. 1436 * 1437 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1438 * Description 1439 * Check whether *skb* is a descendant of the cgroup2 held by 1440 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1441 * Return 1442 * The return value depends on the result of the test, and can be: 1443 * 1444 * * 0, if the *skb* failed the cgroup2 descendant test. 1445 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1446 * * A negative error code, if an error occurred. 1447 * 1448 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1449 * Description 1450 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1451 * not set, in particular if the hash was cleared due to mangling, 1452 * recompute this hash. Later accesses to the hash can be done 1453 * directly with *skb*\ **->hash**. 1454 * 1455 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1456 * prototype with **bpf_skb_change_proto**\ (), or calling 1457 * **bpf_skb_store_bytes**\ () with the 1458 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1459 * the hash and to trigger a new computation for the next call to 1460 * **bpf_get_hash_recalc**\ (). 1461 * Return 1462 * The 32-bit hash. 1463 * 1464 * u64 bpf_get_current_task(void) 1465 * Return 1466 * A pointer to the current task struct. 1467 * 1468 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 1469 * Description 1470 * Attempt in a safe way to write *len* bytes from the buffer 1471 * *src* to *dst* in memory. It only works for threads that are in 1472 * user context, and *dst* must be a valid user space address. 1473 * 1474 * This helper should not be used to implement any kind of 1475 * security mechanism because of TOC-TOU attacks, but rather to 1476 * debug, divert, and manipulate execution of semi-cooperative 1477 * processes. 1478 * 1479 * Keep in mind that this feature is meant for experiments, and it 1480 * has a risk of crashing the system and running programs. 1481 * Therefore, when an eBPF program using this helper is attached, 1482 * a warning including PID and process name is printed to kernel 1483 * logs. 1484 * Return 1485 * 0 on success, or a negative error in case of failure. 1486 * 1487 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1488 * Description 1489 * Check whether the probe is being run is the context of a given 1490 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1491 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1492 * Return 1493 * The return value depends on the result of the test, and can be: 1494 * 1495 * * 1, if current task belongs to the cgroup2. 1496 * * 0, if current task does not belong to the cgroup2. 1497 * * A negative error code, if an error occurred. 1498 * 1499 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1500 * Description 1501 * Resize (trim or grow) the packet associated to *skb* to the 1502 * new *len*. The *flags* are reserved for future usage, and must 1503 * be left at zero. 1504 * 1505 * The basic idea is that the helper performs the needed work to 1506 * change the size of the packet, then the eBPF program rewrites 1507 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1508 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1509 * and others. This helper is a slow path utility intended for 1510 * replies with control messages. And because it is targeted for 1511 * slow path, the helper itself can afford to be slow: it 1512 * implicitly linearizes, unclones and drops offloads from the 1513 * *skb*. 1514 * 1515 * A call to this helper is susceptible to change the underlying 1516 * packet buffer. Therefore, at load time, all checks on pointers 1517 * previously done by the verifier are invalidated and must be 1518 * performed again, if the helper is used in combination with 1519 * direct packet access. 1520 * Return 1521 * 0 on success, or a negative error in case of failure. 1522 * 1523 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1524 * Description 1525 * Pull in non-linear data in case the *skb* is non-linear and not 1526 * all of *len* are part of the linear section. Make *len* bytes 1527 * from *skb* readable and writable. If a zero value is passed for 1528 * *len*, then the whole length of the *skb* is pulled. 1529 * 1530 * This helper is only needed for reading and writing with direct 1531 * packet access. 1532 * 1533 * For direct packet access, testing that offsets to access 1534 * are within packet boundaries (test on *skb*\ **->data_end**) is 1535 * susceptible to fail if offsets are invalid, or if the requested 1536 * data is in non-linear parts of the *skb*. On failure the 1537 * program can just bail out, or in the case of a non-linear 1538 * buffer, use a helper to make the data available. The 1539 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1540 * the data. Another one consists in using **bpf_skb_pull_data** 1541 * to pull in once the non-linear parts, then retesting and 1542 * eventually access the data. 1543 * 1544 * At the same time, this also makes sure the *skb* is uncloned, 1545 * which is a necessary condition for direct write. As this needs 1546 * to be an invariant for the write part only, the verifier 1547 * detects writes and adds a prologue that is calling 1548 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1549 * the very beginning in case it is indeed cloned. 1550 * 1551 * A call to this helper is susceptible to change the underlying 1552 * packet buffer. Therefore, at load time, all checks on pointers 1553 * previously done by the verifier are invalidated and must be 1554 * performed again, if the helper is used in combination with 1555 * direct packet access. 1556 * Return 1557 * 0 on success, or a negative error in case of failure. 1558 * 1559 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1560 * Description 1561 * Add the checksum *csum* into *skb*\ **->csum** in case the 1562 * driver has supplied a checksum for the entire packet into that 1563 * field. Return an error otherwise. This helper is intended to be 1564 * used in combination with **bpf_csum_diff**\ (), in particular 1565 * when the checksum needs to be updated after data has been 1566 * written into the packet through direct packet access. 1567 * Return 1568 * The checksum on success, or a negative error code in case of 1569 * failure. 1570 * 1571 * void bpf_set_hash_invalid(struct sk_buff *skb) 1572 * Description 1573 * Invalidate the current *skb*\ **->hash**. It can be used after 1574 * mangling on headers through direct packet access, in order to 1575 * indicate that the hash is outdated and to trigger a 1576 * recalculation the next time the kernel tries to access this 1577 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1578 * 1579 * long bpf_get_numa_node_id(void) 1580 * Description 1581 * Return the id of the current NUMA node. The primary use case 1582 * for this helper is the selection of sockets for the local NUMA 1583 * node, when the program is attached to sockets using the 1584 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1585 * but the helper is also available to other eBPF program types, 1586 * similarly to **bpf_get_smp_processor_id**\ (). 1587 * Return 1588 * The id of current NUMA node. 1589 * 1590 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1591 * Description 1592 * Grows headroom of packet associated to *skb* and adjusts the 1593 * offset of the MAC header accordingly, adding *len* bytes of 1594 * space. It automatically extends and reallocates memory as 1595 * required. 1596 * 1597 * This helper can be used on a layer 3 *skb* to push a MAC header 1598 * for redirection into a layer 2 device. 1599 * 1600 * All values for *flags* are reserved for future usage, and must 1601 * be left at zero. 1602 * 1603 * A call to this helper is susceptible to change the underlying 1604 * packet buffer. Therefore, at load time, all checks on pointers 1605 * previously done by the verifier are invalidated and must be 1606 * performed again, if the helper is used in combination with 1607 * direct packet access. 1608 * Return 1609 * 0 on success, or a negative error in case of failure. 1610 * 1611 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1612 * Description 1613 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1614 * it is possible to use a negative value for *delta*. This helper 1615 * can be used to prepare the packet for pushing or popping 1616 * headers. 1617 * 1618 * A call to this helper is susceptible to change the underlying 1619 * packet buffer. Therefore, at load time, all checks on pointers 1620 * previously done by the verifier are invalidated and must be 1621 * performed again, if the helper is used in combination with 1622 * direct packet access. 1623 * Return 1624 * 0 on success, or a negative error in case of failure. 1625 * 1626 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1627 * Description 1628 * Copy a NUL terminated string from an unsafe kernel address 1629 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1630 * more details. 1631 * 1632 * Generally, use **bpf_probe_read_user_str**\ () or 1633 * **bpf_probe_read_kernel_str**\ () instead. 1634 * Return 1635 * On success, the strictly positive length of the string, 1636 * including the trailing NUL character. On error, a negative 1637 * value. 1638 * 1639 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1640 * Description 1641 * If the **struct sk_buff** pointed by *skb* has a known socket, 1642 * retrieve the cookie (generated by the kernel) of this socket. 1643 * If no cookie has been set yet, generate a new cookie. Once 1644 * generated, the socket cookie remains stable for the life of the 1645 * socket. This helper can be useful for monitoring per socket 1646 * networking traffic statistics as it provides a global socket 1647 * identifier that can be assumed unique. 1648 * Return 1649 * A 8-byte long non-decreasing number on success, or 0 if the 1650 * socket field is missing inside *skb*. 1651 * 1652 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1653 * Description 1654 * Equivalent to bpf_get_socket_cookie() helper that accepts 1655 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1656 * Return 1657 * A 8-byte long non-decreasing number. 1658 * 1659 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1660 * Description 1661 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 1662 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1663 * Return 1664 * A 8-byte long non-decreasing number. 1665 * 1666 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1667 * Return 1668 * The owner UID of the socket associated to *skb*. If the socket 1669 * is **NULL**, or if it is not a full socket (i.e. if it is a 1670 * time-wait or a request socket instead), **overflowuid** value 1671 * is returned (note that **overflowuid** might also be the actual 1672 * UID value for the socket). 1673 * 1674 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 1675 * Description 1676 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1677 * to value *hash*. 1678 * Return 1679 * 0 1680 * 1681 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1682 * Description 1683 * Emulate a call to **setsockopt()** on the socket associated to 1684 * *bpf_socket*, which must be a full socket. The *level* at 1685 * which the option resides and the name *optname* of the option 1686 * must be specified, see **setsockopt(2)** for more information. 1687 * The option value of length *optlen* is pointed by *optval*. 1688 * 1689 * *bpf_socket* should be one of the following: 1690 * 1691 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1692 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1693 * and **BPF_CGROUP_INET6_CONNECT**. 1694 * 1695 * This helper actually implements a subset of **setsockopt()**. 1696 * It supports the following *level*\ s: 1697 * 1698 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1699 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1700 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1701 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 1702 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1703 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1704 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1705 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1706 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 1707 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1708 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1709 * Return 1710 * 0 on success, or a negative error in case of failure. 1711 * 1712 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1713 * Description 1714 * Grow or shrink the room for data in the packet associated to 1715 * *skb* by *len_diff*, and according to the selected *mode*. 1716 * 1717 * By default, the helper will reset any offloaded checksum 1718 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1719 * by the following flag: 1720 * 1721 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1722 * checksum data of the skb to CHECKSUM_NONE. 1723 * 1724 * There are two supported modes at this time: 1725 * 1726 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1727 * (room space is added or removed below the layer 2 header). 1728 * 1729 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1730 * (room space is added or removed below the layer 3 header). 1731 * 1732 * The following flags are supported at this time: 1733 * 1734 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1735 * Adjusting mss in this way is not allowed for datagrams. 1736 * 1737 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1738 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1739 * Any new space is reserved to hold a tunnel header. 1740 * Configure skb offsets and other fields accordingly. 1741 * 1742 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1743 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1744 * Use with ENCAP_L3 flags to further specify the tunnel type. 1745 * 1746 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1747 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1748 * type; *len* is the length of the inner MAC header. 1749 * 1750 * A call to this helper is susceptible to change the underlying 1751 * packet buffer. Therefore, at load time, all checks on pointers 1752 * previously done by the verifier are invalidated and must be 1753 * performed again, if the helper is used in combination with 1754 * direct packet access. 1755 * Return 1756 * 0 on success, or a negative error in case of failure. 1757 * 1758 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1759 * Description 1760 * Redirect the packet to the endpoint referenced by *map* at 1761 * index *key*. Depending on its type, this *map* can contain 1762 * references to net devices (for forwarding packets through other 1763 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1764 * but this is only implemented for native XDP (with driver 1765 * support) as of this writing). 1766 * 1767 * The lower two bits of *flags* are used as the return code if 1768 * the map lookup fails. This is so that the return value can be 1769 * one of the XDP program return codes up to **XDP_TX**, as chosen 1770 * by the caller. Any higher bits in the *flags* argument must be 1771 * unset. 1772 * 1773 * See also **bpf_redirect**\ (), which only supports redirecting 1774 * to an ifindex, but doesn't require a map to do so. 1775 * Return 1776 * **XDP_REDIRECT** on success, or the value of the two lower bits 1777 * of the *flags* argument on error. 1778 * 1779 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1780 * Description 1781 * Redirect the packet to the socket referenced by *map* (of type 1782 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1783 * egress interfaces can be used for redirection. The 1784 * **BPF_F_INGRESS** value in *flags* is used to make the 1785 * distinction (ingress path is selected if the flag is present, 1786 * egress path otherwise). This is the only flag supported for now. 1787 * Return 1788 * **SK_PASS** on success, or **SK_DROP** on error. 1789 * 1790 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1791 * Description 1792 * Add an entry to, or update a *map* referencing sockets. The 1793 * *skops* is used as a new value for the entry associated to 1794 * *key*. *flags* is one of: 1795 * 1796 * **BPF_NOEXIST** 1797 * The entry for *key* must not exist in the map. 1798 * **BPF_EXIST** 1799 * The entry for *key* must already exist in the map. 1800 * **BPF_ANY** 1801 * No condition on the existence of the entry for *key*. 1802 * 1803 * If the *map* has eBPF programs (parser and verdict), those will 1804 * be inherited by the socket being added. If the socket is 1805 * already attached to eBPF programs, this results in an error. 1806 * Return 1807 * 0 on success, or a negative error in case of failure. 1808 * 1809 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1810 * Description 1811 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1812 * *delta* (which can be positive or negative). Note that this 1813 * operation modifies the address stored in *xdp_md*\ **->data**, 1814 * so the latter must be loaded only after the helper has been 1815 * called. 1816 * 1817 * The use of *xdp_md*\ **->data_meta** is optional and programs 1818 * are not required to use it. The rationale is that when the 1819 * packet is processed with XDP (e.g. as DoS filter), it is 1820 * possible to push further meta data along with it before passing 1821 * to the stack, and to give the guarantee that an ingress eBPF 1822 * program attached as a TC classifier on the same device can pick 1823 * this up for further post-processing. Since TC works with socket 1824 * buffers, it remains possible to set from XDP the **mark** or 1825 * **priority** pointers, or other pointers for the socket buffer. 1826 * Having this scratch space generic and programmable allows for 1827 * more flexibility as the user is free to store whatever meta 1828 * data they need. 1829 * 1830 * A call to this helper is susceptible to change the underlying 1831 * packet buffer. Therefore, at load time, all checks on pointers 1832 * previously done by the verifier are invalidated and must be 1833 * performed again, if the helper is used in combination with 1834 * direct packet access. 1835 * Return 1836 * 0 on success, or a negative error in case of failure. 1837 * 1838 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1839 * Description 1840 * Read the value of a perf event counter, and store it into *buf* 1841 * of size *buf_size*. This helper relies on a *map* of type 1842 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1843 * counter is selected when *map* is updated with perf event file 1844 * descriptors. The *map* is an array whose size is the number of 1845 * available CPUs, and each cell contains a value relative to one 1846 * CPU. The value to retrieve is indicated by *flags*, that 1847 * contains the index of the CPU to look up, masked with 1848 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1849 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1850 * current CPU should be retrieved. 1851 * 1852 * This helper behaves in a way close to 1853 * **bpf_perf_event_read**\ () helper, save that instead of 1854 * just returning the value observed, it fills the *buf* 1855 * structure. This allows for additional data to be retrieved: in 1856 * particular, the enabled and running times (in *buf*\ 1857 * **->enabled** and *buf*\ **->running**, respectively) are 1858 * copied. In general, **bpf_perf_event_read_value**\ () is 1859 * recommended over **bpf_perf_event_read**\ (), which has some 1860 * ABI issues and provides fewer functionalities. 1861 * 1862 * These values are interesting, because hardware PMU (Performance 1863 * Monitoring Unit) counters are limited resources. When there are 1864 * more PMU based perf events opened than available counters, 1865 * kernel will multiplex these events so each event gets certain 1866 * percentage (but not all) of the PMU time. In case that 1867 * multiplexing happens, the number of samples or counter value 1868 * will not reflect the case compared to when no multiplexing 1869 * occurs. This makes comparison between different runs difficult. 1870 * Typically, the counter value should be normalized before 1871 * comparing to other experiments. The usual normalization is done 1872 * as follows. 1873 * 1874 * :: 1875 * 1876 * normalized_counter = counter * t_enabled / t_running 1877 * 1878 * Where t_enabled is the time enabled for event and t_running is 1879 * the time running for event since last normalization. The 1880 * enabled and running times are accumulated since the perf event 1881 * open. To achieve scaling factor between two invocations of an 1882 * eBPF program, users can use CPU id as the key (which is 1883 * typical for perf array usage model) to remember the previous 1884 * value and do the calculation inside the eBPF program. 1885 * Return 1886 * 0 on success, or a negative error in case of failure. 1887 * 1888 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1889 * Description 1890 * For en eBPF program attached to a perf event, retrieve the 1891 * value of the event counter associated to *ctx* and store it in 1892 * the structure pointed by *buf* and of size *buf_size*. Enabled 1893 * and running times are also stored in the structure (see 1894 * description of helper **bpf_perf_event_read_value**\ () for 1895 * more details). 1896 * Return 1897 * 0 on success, or a negative error in case of failure. 1898 * 1899 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1900 * Description 1901 * Emulate a call to **getsockopt()** on the socket associated to 1902 * *bpf_socket*, which must be a full socket. The *level* at 1903 * which the option resides and the name *optname* of the option 1904 * must be specified, see **getsockopt(2)** for more information. 1905 * The retrieved value is stored in the structure pointed by 1906 * *opval* and of length *optlen*. 1907 * 1908 * *bpf_socket* should be one of the following: 1909 * 1910 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1911 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1912 * and **BPF_CGROUP_INET6_CONNECT**. 1913 * 1914 * This helper actually implements a subset of **getsockopt()**. 1915 * It supports the following *level*\ s: 1916 * 1917 * * **IPPROTO_TCP**, which supports *optname* 1918 * **TCP_CONGESTION**. 1919 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1920 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1921 * Return 1922 * 0 on success, or a negative error in case of failure. 1923 * 1924 * long bpf_override_return(struct pt_regs *regs, u64 rc) 1925 * Description 1926 * Used for error injection, this helper uses kprobes to override 1927 * the return value of the probed function, and to set it to *rc*. 1928 * The first argument is the context *regs* on which the kprobe 1929 * works. 1930 * 1931 * This helper works by setting the PC (program counter) 1932 * to an override function which is run in place of the original 1933 * probed function. This means the probed function is not run at 1934 * all. The replacement function just returns with the required 1935 * value. 1936 * 1937 * This helper has security implications, and thus is subject to 1938 * restrictions. It is only available if the kernel was compiled 1939 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1940 * option, and in this case it only works on functions tagged with 1941 * **ALLOW_ERROR_INJECTION** in the kernel code. 1942 * 1943 * Also, the helper is only available for the architectures having 1944 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1945 * x86 architecture is the only one to support this feature. 1946 * Return 1947 * 0 1948 * 1949 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1950 * Description 1951 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1952 * for the full TCP socket associated to *bpf_sock_ops* to 1953 * *argval*. 1954 * 1955 * The primary use of this field is to determine if there should 1956 * be calls to eBPF programs of type 1957 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1958 * code. A program of the same type can change its value, per 1959 * connection and as necessary, when the connection is 1960 * established. This field is directly accessible for reading, but 1961 * this helper must be used for updates in order to return an 1962 * error if an eBPF program tries to set a callback that is not 1963 * supported in the current kernel. 1964 * 1965 * *argval* is a flag array which can combine these flags: 1966 * 1967 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1968 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1969 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1970 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1971 * 1972 * Therefore, this function can be used to clear a callback flag by 1973 * setting the appropriate bit to zero. e.g. to disable the RTO 1974 * callback: 1975 * 1976 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1977 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1978 * 1979 * Here are some examples of where one could call such eBPF 1980 * program: 1981 * 1982 * * When RTO fires. 1983 * * When a packet is retransmitted. 1984 * * When the connection terminates. 1985 * * When a packet is sent. 1986 * * When a packet is received. 1987 * Return 1988 * Code **-EINVAL** if the socket is not a full TCP socket; 1989 * otherwise, a positive number containing the bits that could not 1990 * be set is returned (which comes down to 0 if all bits were set 1991 * as required). 1992 * 1993 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1994 * Description 1995 * This helper is used in programs implementing policies at the 1996 * socket level. If the message *msg* is allowed to pass (i.e. if 1997 * the verdict eBPF program returns **SK_PASS**), redirect it to 1998 * the socket referenced by *map* (of type 1999 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2000 * egress interfaces can be used for redirection. The 2001 * **BPF_F_INGRESS** value in *flags* is used to make the 2002 * distinction (ingress path is selected if the flag is present, 2003 * egress path otherwise). This is the only flag supported for now. 2004 * Return 2005 * **SK_PASS** on success, or **SK_DROP** on error. 2006 * 2007 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2008 * Description 2009 * For socket policies, apply the verdict of the eBPF program to 2010 * the next *bytes* (number of bytes) of message *msg*. 2011 * 2012 * For example, this helper can be used in the following cases: 2013 * 2014 * * A single **sendmsg**\ () or **sendfile**\ () system call 2015 * contains multiple logical messages that the eBPF program is 2016 * supposed to read and for which it should apply a verdict. 2017 * * An eBPF program only cares to read the first *bytes* of a 2018 * *msg*. If the message has a large payload, then setting up 2019 * and calling the eBPF program repeatedly for all bytes, even 2020 * though the verdict is already known, would create unnecessary 2021 * overhead. 2022 * 2023 * When called from within an eBPF program, the helper sets a 2024 * counter internal to the BPF infrastructure, that is used to 2025 * apply the last verdict to the next *bytes*. If *bytes* is 2026 * smaller than the current data being processed from a 2027 * **sendmsg**\ () or **sendfile**\ () system call, the first 2028 * *bytes* will be sent and the eBPF program will be re-run with 2029 * the pointer for start of data pointing to byte number *bytes* 2030 * **+ 1**. If *bytes* is larger than the current data being 2031 * processed, then the eBPF verdict will be applied to multiple 2032 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2033 * consumed. 2034 * 2035 * Note that if a socket closes with the internal counter holding 2036 * a non-zero value, this is not a problem because data is not 2037 * being buffered for *bytes* and is sent as it is received. 2038 * Return 2039 * 0 2040 * 2041 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2042 * Description 2043 * For socket policies, prevent the execution of the verdict eBPF 2044 * program for message *msg* until *bytes* (byte number) have been 2045 * accumulated. 2046 * 2047 * This can be used when one needs a specific number of bytes 2048 * before a verdict can be assigned, even if the data spans 2049 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2050 * case would be a user calling **sendmsg**\ () repeatedly with 2051 * 1-byte long message segments. Obviously, this is bad for 2052 * performance, but it is still valid. If the eBPF program needs 2053 * *bytes* bytes to validate a header, this helper can be used to 2054 * prevent the eBPF program to be called again until *bytes* have 2055 * been accumulated. 2056 * Return 2057 * 0 2058 * 2059 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2060 * Description 2061 * For socket policies, pull in non-linear data from user space 2062 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2063 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2064 * respectively. 2065 * 2066 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2067 * *msg* it can only parse data that the (**data**, **data_end**) 2068 * pointers have already consumed. For **sendmsg**\ () hooks this 2069 * is likely the first scatterlist element. But for calls relying 2070 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2071 * be the range (**0**, **0**) because the data is shared with 2072 * user space and by default the objective is to avoid allowing 2073 * user space to modify data while (or after) eBPF verdict is 2074 * being decided. This helper can be used to pull in data and to 2075 * set the start and end pointer to given values. Data will be 2076 * copied if necessary (i.e. if data was not linear and if start 2077 * and end pointers do not point to the same chunk). 2078 * 2079 * A call to this helper is susceptible to change the underlying 2080 * packet buffer. Therefore, at load time, all checks on pointers 2081 * previously done by the verifier are invalidated and must be 2082 * performed again, if the helper is used in combination with 2083 * direct packet access. 2084 * 2085 * All values for *flags* are reserved for future usage, and must 2086 * be left at zero. 2087 * Return 2088 * 0 on success, or a negative error in case of failure. 2089 * 2090 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2091 * Description 2092 * Bind the socket associated to *ctx* to the address pointed by 2093 * *addr*, of length *addr_len*. This allows for making outgoing 2094 * connection from the desired IP address, which can be useful for 2095 * example when all processes inside a cgroup should use one 2096 * single IP address on a host that has multiple IP configured. 2097 * 2098 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2099 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2100 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2101 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2102 * behavior and lets the kernel efficiently pick up an unused 2103 * port as long as 4-tuple is unique. Passing non-zero port might 2104 * lead to degraded performance. 2105 * Return 2106 * 0 on success, or a negative error in case of failure. 2107 * 2108 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2109 * Description 2110 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2111 * possible to both shrink and grow the packet tail. 2112 * Shrink done via *delta* being a negative integer. 2113 * 2114 * A call to this helper is susceptible to change the underlying 2115 * packet buffer. Therefore, at load time, all checks on pointers 2116 * previously done by the verifier are invalidated and must be 2117 * performed again, if the helper is used in combination with 2118 * direct packet access. 2119 * Return 2120 * 0 on success, or a negative error in case of failure. 2121 * 2122 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2123 * Description 2124 * Retrieve the XFRM state (IP transform framework, see also 2125 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2126 * 2127 * The retrieved value is stored in the **struct bpf_xfrm_state** 2128 * pointed by *xfrm_state* and of length *size*. 2129 * 2130 * All values for *flags* are reserved for future usage, and must 2131 * be left at zero. 2132 * 2133 * This helper is available only if the kernel was compiled with 2134 * **CONFIG_XFRM** configuration option. 2135 * Return 2136 * 0 on success, or a negative error in case of failure. 2137 * 2138 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2139 * Description 2140 * Return a user or a kernel stack in bpf program provided buffer. 2141 * To achieve this, the helper needs *ctx*, which is a pointer 2142 * to the context on which the tracing program is executed. 2143 * To store the stacktrace, the bpf program provides *buf* with 2144 * a nonnegative *size*. 2145 * 2146 * The last argument, *flags*, holds the number of stack frames to 2147 * skip (from 0 to 255), masked with 2148 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2149 * the following flags: 2150 * 2151 * **BPF_F_USER_STACK** 2152 * Collect a user space stack instead of a kernel stack. 2153 * **BPF_F_USER_BUILD_ID** 2154 * Collect buildid+offset instead of ips for user stack, 2155 * only valid if **BPF_F_USER_STACK** is also specified. 2156 * 2157 * **bpf_get_stack**\ () can collect up to 2158 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2159 * to sufficient large buffer size. Note that 2160 * this limit can be controlled with the **sysctl** program, and 2161 * that it should be manually increased in order to profile long 2162 * user stacks (such as stacks for Java programs). To do so, use: 2163 * 2164 * :: 2165 * 2166 * # sysctl kernel.perf_event_max_stack=<new value> 2167 * Return 2168 * The non-negative copied *buf* length equal to or less than 2169 * *size* on success, or a negative error in case of failure. 2170 * 2171 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2172 * Description 2173 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2174 * it provides an easy way to load *len* bytes from *offset* 2175 * from the packet associated to *skb*, into the buffer pointed 2176 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2177 * a fifth argument *start_header* exists in order to select a 2178 * base offset to start from. *start_header* can be one of: 2179 * 2180 * **BPF_HDR_START_MAC** 2181 * Base offset to load data from is *skb*'s mac header. 2182 * **BPF_HDR_START_NET** 2183 * Base offset to load data from is *skb*'s network header. 2184 * 2185 * In general, "direct packet access" is the preferred method to 2186 * access packet data, however, this helper is in particular useful 2187 * in socket filters where *skb*\ **->data** does not always point 2188 * to the start of the mac header and where "direct packet access" 2189 * is not available. 2190 * Return 2191 * 0 on success, or a negative error in case of failure. 2192 * 2193 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2194 * Description 2195 * Do FIB lookup in kernel tables using parameters in *params*. 2196 * If lookup is successful and result shows packet is to be 2197 * forwarded, the neighbor tables are searched for the nexthop. 2198 * If successful (ie., FIB lookup shows forwarding and nexthop 2199 * is resolved), the nexthop address is returned in ipv4_dst 2200 * or ipv6_dst based on family, smac is set to mac address of 2201 * egress device, dmac is set to nexthop mac address, rt_metric 2202 * is set to metric from route (IPv4/IPv6 only), and ifindex 2203 * is set to the device index of the nexthop from the FIB lookup. 2204 * 2205 * *plen* argument is the size of the passed in struct. 2206 * *flags* argument can be a combination of one or more of the 2207 * following values: 2208 * 2209 * **BPF_FIB_LOOKUP_DIRECT** 2210 * Do a direct table lookup vs full lookup using FIB 2211 * rules. 2212 * **BPF_FIB_LOOKUP_OUTPUT** 2213 * Perform lookup from an egress perspective (default is 2214 * ingress). 2215 * 2216 * *ctx* is either **struct xdp_md** for XDP programs or 2217 * **struct sk_buff** tc cls_act programs. 2218 * Return 2219 * * < 0 if any input argument is invalid 2220 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2221 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2222 * packet is not forwarded or needs assist from full stack 2223 * 2224 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2225 * Description 2226 * Add an entry to, or update a sockhash *map* referencing sockets. 2227 * The *skops* is used as a new value for the entry associated to 2228 * *key*. *flags* is one of: 2229 * 2230 * **BPF_NOEXIST** 2231 * The entry for *key* must not exist in the map. 2232 * **BPF_EXIST** 2233 * The entry for *key* must already exist in the map. 2234 * **BPF_ANY** 2235 * No condition on the existence of the entry for *key*. 2236 * 2237 * If the *map* has eBPF programs (parser and verdict), those will 2238 * be inherited by the socket being added. If the socket is 2239 * already attached to eBPF programs, this results in an error. 2240 * Return 2241 * 0 on success, or a negative error in case of failure. 2242 * 2243 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2244 * Description 2245 * This helper is used in programs implementing policies at the 2246 * socket level. If the message *msg* is allowed to pass (i.e. if 2247 * the verdict eBPF program returns **SK_PASS**), redirect it to 2248 * the socket referenced by *map* (of type 2249 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2250 * egress interfaces can be used for redirection. The 2251 * **BPF_F_INGRESS** value in *flags* is used to make the 2252 * distinction (ingress path is selected if the flag is present, 2253 * egress path otherwise). This is the only flag supported for now. 2254 * Return 2255 * **SK_PASS** on success, or **SK_DROP** on error. 2256 * 2257 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2258 * Description 2259 * This helper is used in programs implementing policies at the 2260 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2261 * if the verdict eBPF program returns **SK_PASS**), redirect it 2262 * to the socket referenced by *map* (of type 2263 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2264 * egress interfaces can be used for redirection. The 2265 * **BPF_F_INGRESS** value in *flags* is used to make the 2266 * distinction (ingress path is selected if the flag is present, 2267 * egress otherwise). This is the only flag supported for now. 2268 * Return 2269 * **SK_PASS** on success, or **SK_DROP** on error. 2270 * 2271 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2272 * Description 2273 * Encapsulate the packet associated to *skb* within a Layer 3 2274 * protocol header. This header is provided in the buffer at 2275 * address *hdr*, with *len* its size in bytes. *type* indicates 2276 * the protocol of the header and can be one of: 2277 * 2278 * **BPF_LWT_ENCAP_SEG6** 2279 * IPv6 encapsulation with Segment Routing Header 2280 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2281 * the IPv6 header is computed by the kernel. 2282 * **BPF_LWT_ENCAP_SEG6_INLINE** 2283 * Only works if *skb* contains an IPv6 packet. Insert a 2284 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2285 * the IPv6 header. 2286 * **BPF_LWT_ENCAP_IP** 2287 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2288 * must be IPv4 or IPv6, followed by zero or more 2289 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2290 * total bytes in all prepended headers. Please note that 2291 * if **skb_is_gso**\ (*skb*) is true, no more than two 2292 * headers can be prepended, and the inner header, if 2293 * present, should be either GRE or UDP/GUE. 2294 * 2295 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2296 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2297 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2298 * **BPF_PROG_TYPE_LWT_XMIT**. 2299 * 2300 * A call to this helper is susceptible to change the underlying 2301 * packet buffer. Therefore, at load time, all checks on pointers 2302 * previously done by the verifier are invalidated and must be 2303 * performed again, if the helper is used in combination with 2304 * direct packet access. 2305 * Return 2306 * 0 on success, or a negative error in case of failure. 2307 * 2308 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2309 * Description 2310 * Store *len* bytes from address *from* into the packet 2311 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2312 * inside the outermost IPv6 Segment Routing Header can be 2313 * modified through this helper. 2314 * 2315 * A call to this helper is susceptible to change the underlying 2316 * packet buffer. Therefore, at load time, all checks on pointers 2317 * previously done by the verifier are invalidated and must be 2318 * performed again, if the helper is used in combination with 2319 * direct packet access. 2320 * Return 2321 * 0 on success, or a negative error in case of failure. 2322 * 2323 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2324 * Description 2325 * Adjust the size allocated to TLVs in the outermost IPv6 2326 * Segment Routing Header contained in the packet associated to 2327 * *skb*, at position *offset* by *delta* bytes. Only offsets 2328 * after the segments are accepted. *delta* can be as well 2329 * positive (growing) as negative (shrinking). 2330 * 2331 * A call to this helper is susceptible to change the underlying 2332 * packet buffer. Therefore, at load time, all checks on pointers 2333 * previously done by the verifier are invalidated and must be 2334 * performed again, if the helper is used in combination with 2335 * direct packet access. 2336 * Return 2337 * 0 on success, or a negative error in case of failure. 2338 * 2339 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2340 * Description 2341 * Apply an IPv6 Segment Routing action of type *action* to the 2342 * packet associated to *skb*. Each action takes a parameter 2343 * contained at address *param*, and of length *param_len* bytes. 2344 * *action* can be one of: 2345 * 2346 * **SEG6_LOCAL_ACTION_END_X** 2347 * End.X action: Endpoint with Layer-3 cross-connect. 2348 * Type of *param*: **struct in6_addr**. 2349 * **SEG6_LOCAL_ACTION_END_T** 2350 * End.T action: Endpoint with specific IPv6 table lookup. 2351 * Type of *param*: **int**. 2352 * **SEG6_LOCAL_ACTION_END_B6** 2353 * End.B6 action: Endpoint bound to an SRv6 policy. 2354 * Type of *param*: **struct ipv6_sr_hdr**. 2355 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2356 * End.B6.Encap action: Endpoint bound to an SRv6 2357 * encapsulation policy. 2358 * Type of *param*: **struct ipv6_sr_hdr**. 2359 * 2360 * A call to this helper is susceptible to change the underlying 2361 * packet buffer. Therefore, at load time, all checks on pointers 2362 * previously done by the verifier are invalidated and must be 2363 * performed again, if the helper is used in combination with 2364 * direct packet access. 2365 * Return 2366 * 0 on success, or a negative error in case of failure. 2367 * 2368 * long bpf_rc_repeat(void *ctx) 2369 * Description 2370 * This helper is used in programs implementing IR decoding, to 2371 * report a successfully decoded repeat key message. This delays 2372 * the generation of a key up event for previously generated 2373 * key down event. 2374 * 2375 * Some IR protocols like NEC have a special IR message for 2376 * repeating last button, for when a button is held down. 2377 * 2378 * The *ctx* should point to the lirc sample as passed into 2379 * the program. 2380 * 2381 * This helper is only available is the kernel was compiled with 2382 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2383 * "**y**". 2384 * Return 2385 * 0 2386 * 2387 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2388 * Description 2389 * This helper is used in programs implementing IR decoding, to 2390 * report a successfully decoded key press with *scancode*, 2391 * *toggle* value in the given *protocol*. The scancode will be 2392 * translated to a keycode using the rc keymap, and reported as 2393 * an input key down event. After a period a key up event is 2394 * generated. This period can be extended by calling either 2395 * **bpf_rc_keydown**\ () again with the same values, or calling 2396 * **bpf_rc_repeat**\ (). 2397 * 2398 * Some protocols include a toggle bit, in case the button was 2399 * released and pressed again between consecutive scancodes. 2400 * 2401 * The *ctx* should point to the lirc sample as passed into 2402 * the program. 2403 * 2404 * The *protocol* is the decoded protocol number (see 2405 * **enum rc_proto** for some predefined values). 2406 * 2407 * This helper is only available is the kernel was compiled with 2408 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2409 * "**y**". 2410 * Return 2411 * 0 2412 * 2413 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2414 * Description 2415 * Return the cgroup v2 id of the socket associated with the *skb*. 2416 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2417 * helper for cgroup v1 by providing a tag resp. identifier that 2418 * can be matched on or used for map lookups e.g. to implement 2419 * policy. The cgroup v2 id of a given path in the hierarchy is 2420 * exposed in user space through the f_handle API in order to get 2421 * to the same 64-bit id. 2422 * 2423 * This helper can be used on TC egress path, but not on ingress, 2424 * and is available only if the kernel was compiled with the 2425 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2426 * Return 2427 * The id is returned or 0 in case the id could not be retrieved. 2428 * 2429 * u64 bpf_get_current_cgroup_id(void) 2430 * Return 2431 * A 64-bit integer containing the current cgroup id based 2432 * on the cgroup within which the current task is running. 2433 * 2434 * void *bpf_get_local_storage(void *map, u64 flags) 2435 * Description 2436 * Get the pointer to the local storage area. 2437 * The type and the size of the local storage is defined 2438 * by the *map* argument. 2439 * The *flags* meaning is specific for each map type, 2440 * and has to be 0 for cgroup local storage. 2441 * 2442 * Depending on the BPF program type, a local storage area 2443 * can be shared between multiple instances of the BPF program, 2444 * running simultaneously. 2445 * 2446 * A user should care about the synchronization by himself. 2447 * For example, by using the **BPF_STX_XADD** instruction to alter 2448 * the shared data. 2449 * Return 2450 * A pointer to the local storage area. 2451 * 2452 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2453 * Description 2454 * Select a **SO_REUSEPORT** socket from a 2455 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 2456 * It checks the selected socket is matching the incoming 2457 * request in the socket buffer. 2458 * Return 2459 * 0 on success, or a negative error in case of failure. 2460 * 2461 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2462 * Description 2463 * Return id of cgroup v2 that is ancestor of cgroup associated 2464 * with the *skb* at the *ancestor_level*. The root cgroup is at 2465 * *ancestor_level* zero and each step down the hierarchy 2466 * increments the level. If *ancestor_level* == level of cgroup 2467 * associated with *skb*, then return value will be same as that 2468 * of **bpf_skb_cgroup_id**\ (). 2469 * 2470 * The helper is useful to implement policies based on cgroups 2471 * that are upper in hierarchy than immediate cgroup associated 2472 * with *skb*. 2473 * 2474 * The format of returned id and helper limitations are same as in 2475 * **bpf_skb_cgroup_id**\ (). 2476 * Return 2477 * The id is returned or 0 in case the id could not be retrieved. 2478 * 2479 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2480 * Description 2481 * Look for TCP socket matching *tuple*, optionally in a child 2482 * network namespace *netns*. The return value must be checked, 2483 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2484 * 2485 * The *ctx* should point to the context of the program, such as 2486 * the skb or socket (depending on the hook in use). This is used 2487 * to determine the base network namespace for the lookup. 2488 * 2489 * *tuple_size* must be one of: 2490 * 2491 * **sizeof**\ (*tuple*\ **->ipv4**) 2492 * Look for an IPv4 socket. 2493 * **sizeof**\ (*tuple*\ **->ipv6**) 2494 * Look for an IPv6 socket. 2495 * 2496 * If the *netns* is a negative signed 32-bit integer, then the 2497 * socket lookup table in the netns associated with the *ctx* 2498 * will be used. For the TC hooks, this is the netns of the device 2499 * in the skb. For socket hooks, this is the netns of the socket. 2500 * If *netns* is any other signed 32-bit value greater than or 2501 * equal to zero then it specifies the ID of the netns relative to 2502 * the netns associated with the *ctx*. *netns* values beyond the 2503 * range of 32-bit integers are reserved for future use. 2504 * 2505 * All values for *flags* are reserved for future usage, and must 2506 * be left at zero. 2507 * 2508 * This helper is available only if the kernel was compiled with 2509 * **CONFIG_NET** configuration option. 2510 * Return 2511 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2512 * For sockets with reuseport option, the **struct bpf_sock** 2513 * result is from *reuse*\ **->socks**\ [] using the hash of the 2514 * tuple. 2515 * 2516 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2517 * Description 2518 * Look for UDP socket matching *tuple*, optionally in a child 2519 * network namespace *netns*. The return value must be checked, 2520 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2521 * 2522 * The *ctx* should point to the context of the program, such as 2523 * the skb or socket (depending on the hook in use). This is used 2524 * to determine the base network namespace for the lookup. 2525 * 2526 * *tuple_size* must be one of: 2527 * 2528 * **sizeof**\ (*tuple*\ **->ipv4**) 2529 * Look for an IPv4 socket. 2530 * **sizeof**\ (*tuple*\ **->ipv6**) 2531 * Look for an IPv6 socket. 2532 * 2533 * If the *netns* is a negative signed 32-bit integer, then the 2534 * socket lookup table in the netns associated with the *ctx* 2535 * will be used. For the TC hooks, this is the netns of the device 2536 * in the skb. For socket hooks, this is the netns of the socket. 2537 * If *netns* is any other signed 32-bit value greater than or 2538 * equal to zero then it specifies the ID of the netns relative to 2539 * the netns associated with the *ctx*. *netns* values beyond the 2540 * range of 32-bit integers are reserved for future use. 2541 * 2542 * All values for *flags* are reserved for future usage, and must 2543 * be left at zero. 2544 * 2545 * This helper is available only if the kernel was compiled with 2546 * **CONFIG_NET** configuration option. 2547 * Return 2548 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2549 * For sockets with reuseport option, the **struct bpf_sock** 2550 * result is from *reuse*\ **->socks**\ [] using the hash of the 2551 * tuple. 2552 * 2553 * long bpf_sk_release(void *sock) 2554 * Description 2555 * Release the reference held by *sock*. *sock* must be a 2556 * non-**NULL** pointer that was returned from 2557 * **bpf_sk_lookup_xxx**\ (). 2558 * Return 2559 * 0 on success, or a negative error in case of failure. 2560 * 2561 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2562 * Description 2563 * Push an element *value* in *map*. *flags* is one of: 2564 * 2565 * **BPF_EXIST** 2566 * If the queue/stack is full, the oldest element is 2567 * removed to make room for this. 2568 * Return 2569 * 0 on success, or a negative error in case of failure. 2570 * 2571 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 2572 * Description 2573 * Pop an element from *map*. 2574 * Return 2575 * 0 on success, or a negative error in case of failure. 2576 * 2577 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 2578 * Description 2579 * Get an element from *map* without removing it. 2580 * Return 2581 * 0 on success, or a negative error in case of failure. 2582 * 2583 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2584 * Description 2585 * For socket policies, insert *len* bytes into *msg* at offset 2586 * *start*. 2587 * 2588 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2589 * *msg* it may want to insert metadata or options into the *msg*. 2590 * This can later be read and used by any of the lower layer BPF 2591 * hooks. 2592 * 2593 * This helper may fail if under memory pressure (a malloc 2594 * fails) in these cases BPF programs will get an appropriate 2595 * error and BPF programs will need to handle them. 2596 * Return 2597 * 0 on success, or a negative error in case of failure. 2598 * 2599 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2600 * Description 2601 * Will remove *len* bytes from a *msg* starting at byte *start*. 2602 * This may result in **ENOMEM** errors under certain situations if 2603 * an allocation and copy are required due to a full ring buffer. 2604 * However, the helper will try to avoid doing the allocation 2605 * if possible. Other errors can occur if input parameters are 2606 * invalid either due to *start* byte not being valid part of *msg* 2607 * payload and/or *pop* value being to large. 2608 * Return 2609 * 0 on success, or a negative error in case of failure. 2610 * 2611 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2612 * Description 2613 * This helper is used in programs implementing IR decoding, to 2614 * report a successfully decoded pointer movement. 2615 * 2616 * The *ctx* should point to the lirc sample as passed into 2617 * the program. 2618 * 2619 * This helper is only available is the kernel was compiled with 2620 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2621 * "**y**". 2622 * Return 2623 * 0 2624 * 2625 * long bpf_spin_lock(struct bpf_spin_lock *lock) 2626 * Description 2627 * Acquire a spinlock represented by the pointer *lock*, which is 2628 * stored as part of a value of a map. Taking the lock allows to 2629 * safely update the rest of the fields in that value. The 2630 * spinlock can (and must) later be released with a call to 2631 * **bpf_spin_unlock**\ (\ *lock*\ ). 2632 * 2633 * Spinlocks in BPF programs come with a number of restrictions 2634 * and constraints: 2635 * 2636 * * **bpf_spin_lock** objects are only allowed inside maps of 2637 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2638 * list could be extended in the future). 2639 * * BTF description of the map is mandatory. 2640 * * The BPF program can take ONE lock at a time, since taking two 2641 * or more could cause dead locks. 2642 * * Only one **struct bpf_spin_lock** is allowed per map element. 2643 * * When the lock is taken, calls (either BPF to BPF or helpers) 2644 * are not allowed. 2645 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2646 * allowed inside a spinlock-ed region. 2647 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2648 * the lock, on all execution paths, before it returns. 2649 * * The BPF program can access **struct bpf_spin_lock** only via 2650 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2651 * helpers. Loading or storing data into the **struct 2652 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2653 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2654 * of the map value must be a struct and have **struct 2655 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2656 * Nested lock inside another struct is not allowed. 2657 * * The **struct bpf_spin_lock** *lock* field in a map value must 2658 * be aligned on a multiple of 4 bytes in that value. 2659 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2660 * the **bpf_spin_lock** field to user space. 2661 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2662 * a BPF program, do not update the **bpf_spin_lock** field. 2663 * * **bpf_spin_lock** cannot be on the stack or inside a 2664 * networking packet (it can only be inside of a map values). 2665 * * **bpf_spin_lock** is available to root only. 2666 * * Tracing programs and socket filter programs cannot use 2667 * **bpf_spin_lock**\ () due to insufficient preemption checks 2668 * (but this may change in the future). 2669 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2670 * Return 2671 * 0 2672 * 2673 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 2674 * Description 2675 * Release the *lock* previously locked by a call to 2676 * **bpf_spin_lock**\ (\ *lock*\ ). 2677 * Return 2678 * 0 2679 * 2680 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2681 * Description 2682 * This helper gets a **struct bpf_sock** pointer such 2683 * that all the fields in this **bpf_sock** can be accessed. 2684 * Return 2685 * A **struct bpf_sock** pointer on success, or **NULL** in 2686 * case of failure. 2687 * 2688 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2689 * Description 2690 * This helper gets a **struct bpf_tcp_sock** pointer from a 2691 * **struct bpf_sock** pointer. 2692 * Return 2693 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2694 * case of failure. 2695 * 2696 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 2697 * Description 2698 * Set ECN (Explicit Congestion Notification) field of IP header 2699 * to **CE** (Congestion Encountered) if current value is **ECT** 2700 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2701 * and IPv4. 2702 * Return 2703 * 1 if the **CE** flag is set (either by the current helper call 2704 * or because it was already present), 0 if it is not set. 2705 * 2706 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2707 * Description 2708 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2709 * **bpf_sk_release**\ () is unnecessary and not allowed. 2710 * Return 2711 * A **struct bpf_sock** pointer on success, or **NULL** in 2712 * case of failure. 2713 * 2714 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2715 * Description 2716 * Look for TCP socket matching *tuple*, optionally in a child 2717 * network namespace *netns*. The return value must be checked, 2718 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2719 * 2720 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2721 * that it also returns timewait or request sockets. Use 2722 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2723 * full structure. 2724 * 2725 * This helper is available only if the kernel was compiled with 2726 * **CONFIG_NET** configuration option. 2727 * Return 2728 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2729 * For sockets with reuseport option, the **struct bpf_sock** 2730 * result is from *reuse*\ **->socks**\ [] using the hash of the 2731 * tuple. 2732 * 2733 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2734 * Description 2735 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2736 * the listening socket in *sk*. 2737 * 2738 * *iph* points to the start of the IPv4 or IPv6 header, while 2739 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2740 * **sizeof**\ (**struct ip6hdr**). 2741 * 2742 * *th* points to the start of the TCP header, while *th_len* 2743 * contains **sizeof**\ (**struct tcphdr**). 2744 * Return 2745 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2746 * error otherwise. 2747 * 2748 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2749 * Description 2750 * Get name of sysctl in /proc/sys/ and copy it into provided by 2751 * program buffer *buf* of size *buf_len*. 2752 * 2753 * The buffer is always NUL terminated, unless it's zero-sized. 2754 * 2755 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2756 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2757 * only (e.g. "tcp_mem"). 2758 * Return 2759 * Number of character copied (not including the trailing NUL). 2760 * 2761 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2762 * truncated name in this case). 2763 * 2764 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2765 * Description 2766 * Get current value of sysctl as it is presented in /proc/sys 2767 * (incl. newline, etc), and copy it as a string into provided 2768 * by program buffer *buf* of size *buf_len*. 2769 * 2770 * The whole value is copied, no matter what file position user 2771 * space issued e.g. sys_read at. 2772 * 2773 * The buffer is always NUL terminated, unless it's zero-sized. 2774 * Return 2775 * Number of character copied (not including the trailing NUL). 2776 * 2777 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2778 * truncated name in this case). 2779 * 2780 * **-EINVAL** if current value was unavailable, e.g. because 2781 * sysctl is uninitialized and read returns -EIO for it. 2782 * 2783 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2784 * Description 2785 * Get new value being written by user space to sysctl (before 2786 * the actual write happens) and copy it as a string into 2787 * provided by program buffer *buf* of size *buf_len*. 2788 * 2789 * User space may write new value at file position > 0. 2790 * 2791 * The buffer is always NUL terminated, unless it's zero-sized. 2792 * Return 2793 * Number of character copied (not including the trailing NUL). 2794 * 2795 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2796 * truncated name in this case). 2797 * 2798 * **-EINVAL** if sysctl is being read. 2799 * 2800 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2801 * Description 2802 * Override new value being written by user space to sysctl with 2803 * value provided by program in buffer *buf* of size *buf_len*. 2804 * 2805 * *buf* should contain a string in same form as provided by user 2806 * space on sysctl write. 2807 * 2808 * User space may write new value at file position > 0. To override 2809 * the whole sysctl value file position should be set to zero. 2810 * Return 2811 * 0 on success. 2812 * 2813 * **-E2BIG** if the *buf_len* is too big. 2814 * 2815 * **-EINVAL** if sysctl is being read. 2816 * 2817 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2818 * Description 2819 * Convert the initial part of the string from buffer *buf* of 2820 * size *buf_len* to a long integer according to the given base 2821 * and save the result in *res*. 2822 * 2823 * The string may begin with an arbitrary amount of white space 2824 * (as determined by **isspace**\ (3)) followed by a single 2825 * optional '**-**' sign. 2826 * 2827 * Five least significant bits of *flags* encode base, other bits 2828 * are currently unused. 2829 * 2830 * Base must be either 8, 10, 16 or 0 to detect it automatically 2831 * similar to user space **strtol**\ (3). 2832 * Return 2833 * Number of characters consumed on success. Must be positive but 2834 * no more than *buf_len*. 2835 * 2836 * **-EINVAL** if no valid digits were found or unsupported base 2837 * was provided. 2838 * 2839 * **-ERANGE** if resulting value was out of range. 2840 * 2841 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2842 * Description 2843 * Convert the initial part of the string from buffer *buf* of 2844 * size *buf_len* to an unsigned long integer according to the 2845 * given base and save the result in *res*. 2846 * 2847 * The string may begin with an arbitrary amount of white space 2848 * (as determined by **isspace**\ (3)). 2849 * 2850 * Five least significant bits of *flags* encode base, other bits 2851 * are currently unused. 2852 * 2853 * Base must be either 8, 10, 16 or 0 to detect it automatically 2854 * similar to user space **strtoul**\ (3). 2855 * Return 2856 * Number of characters consumed on success. Must be positive but 2857 * no more than *buf_len*. 2858 * 2859 * **-EINVAL** if no valid digits were found or unsupported base 2860 * was provided. 2861 * 2862 * **-ERANGE** if resulting value was out of range. 2863 * 2864 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 2865 * Description 2866 * Get a bpf-local-storage from a *sk*. 2867 * 2868 * Logically, it could be thought of getting the value from 2869 * a *map* with *sk* as the **key**. From this 2870 * perspective, the usage is not much different from 2871 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2872 * helper enforces the key must be a full socket and the map must 2873 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2874 * 2875 * Underneath, the value is stored locally at *sk* instead of 2876 * the *map*. The *map* is used as the bpf-local-storage 2877 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2878 * searched against all bpf-local-storages residing at *sk*. 2879 * 2880 * *sk* is a kernel **struct sock** pointer for LSM program. 2881 * *sk* is a **struct bpf_sock** pointer for other program types. 2882 * 2883 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2884 * used such that a new bpf-local-storage will be 2885 * created if one does not exist. *value* can be used 2886 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2887 * the initial value of a bpf-local-storage. If *value* is 2888 * **NULL**, the new bpf-local-storage will be zero initialized. 2889 * Return 2890 * A bpf-local-storage pointer is returned on success. 2891 * 2892 * **NULL** if not found or there was an error in adding 2893 * a new bpf-local-storage. 2894 * 2895 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 2896 * Description 2897 * Delete a bpf-local-storage from a *sk*. 2898 * Return 2899 * 0 on success. 2900 * 2901 * **-ENOENT** if the bpf-local-storage cannot be found. 2902 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 2903 * 2904 * long bpf_send_signal(u32 sig) 2905 * Description 2906 * Send signal *sig* to the process of the current task. 2907 * The signal may be delivered to any of this process's threads. 2908 * Return 2909 * 0 on success or successfully queued. 2910 * 2911 * **-EBUSY** if work queue under nmi is full. 2912 * 2913 * **-EINVAL** if *sig* is invalid. 2914 * 2915 * **-EPERM** if no permission to send the *sig*. 2916 * 2917 * **-EAGAIN** if bpf program can try again. 2918 * 2919 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2920 * Description 2921 * Try to issue a SYN cookie for the packet with corresponding 2922 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2923 * 2924 * *iph* points to the start of the IPv4 or IPv6 header, while 2925 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2926 * **sizeof**\ (**struct ip6hdr**). 2927 * 2928 * *th* points to the start of the TCP header, while *th_len* 2929 * contains the length of the TCP header. 2930 * Return 2931 * On success, lower 32 bits hold the generated SYN cookie in 2932 * followed by 16 bits which hold the MSS value for that cookie, 2933 * and the top 16 bits are unused. 2934 * 2935 * On failure, the returned value is one of the following: 2936 * 2937 * **-EINVAL** SYN cookie cannot be issued due to error 2938 * 2939 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2940 * 2941 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2942 * 2943 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2944 * 2945 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2946 * Description 2947 * Write raw *data* blob into a special BPF perf event held by 2948 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2949 * event must have the following attributes: **PERF_SAMPLE_RAW** 2950 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2951 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2952 * 2953 * The *flags* are used to indicate the index in *map* for which 2954 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2955 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2956 * to indicate that the index of the current CPU core should be 2957 * used. 2958 * 2959 * The value to write, of *size*, is passed through eBPF stack and 2960 * pointed by *data*. 2961 * 2962 * *ctx* is a pointer to in-kernel struct sk_buff. 2963 * 2964 * This helper is similar to **bpf_perf_event_output**\ () but 2965 * restricted to raw_tracepoint bpf programs. 2966 * Return 2967 * 0 on success, or a negative error in case of failure. 2968 * 2969 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2970 * Description 2971 * Safely attempt to read *size* bytes from user space address 2972 * *unsafe_ptr* and store the data in *dst*. 2973 * Return 2974 * 0 on success, or a negative error in case of failure. 2975 * 2976 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2977 * Description 2978 * Safely attempt to read *size* bytes from kernel space address 2979 * *unsafe_ptr* and store the data in *dst*. 2980 * Return 2981 * 0 on success, or a negative error in case of failure. 2982 * 2983 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2984 * Description 2985 * Copy a NUL terminated string from an unsafe user address 2986 * *unsafe_ptr* to *dst*. The *size* should include the 2987 * terminating NUL byte. In case the string length is smaller than 2988 * *size*, the target is not padded with further NUL bytes. If the 2989 * string length is larger than *size*, just *size*-1 bytes are 2990 * copied and the last byte is set to NUL. 2991 * 2992 * On success, the length of the copied string is returned. This 2993 * makes this helper useful in tracing programs for reading 2994 * strings, and more importantly to get its length at runtime. See 2995 * the following snippet: 2996 * 2997 * :: 2998 * 2999 * SEC("kprobe/sys_open") 3000 * void bpf_sys_open(struct pt_regs *ctx) 3001 * { 3002 * char buf[PATHLEN]; // PATHLEN is defined to 256 3003 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3004 * ctx->di); 3005 * 3006 * // Consume buf, for example push it to 3007 * // userspace via bpf_perf_event_output(); we 3008 * // can use res (the string length) as event 3009 * // size, after checking its boundaries. 3010 * } 3011 * 3012 * In comparison, using **bpf_probe_read_user**\ () helper here 3013 * instead to read the string would require to estimate the length 3014 * at compile time, and would often result in copying more memory 3015 * than necessary. 3016 * 3017 * Another useful use case is when parsing individual process 3018 * arguments or individual environment variables navigating 3019 * *current*\ **->mm->arg_start** and *current*\ 3020 * **->mm->env_start**: using this helper and the return value, 3021 * one can quickly iterate at the right offset of the memory area. 3022 * Return 3023 * On success, the strictly positive length of the string, 3024 * including the trailing NUL character. On error, a negative 3025 * value. 3026 * 3027 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3028 * Description 3029 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3030 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3031 * Return 3032 * On success, the strictly positive length of the string, including 3033 * the trailing NUL character. On error, a negative value. 3034 * 3035 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3036 * Description 3037 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3038 * *rcv_nxt* is the ack_seq to be sent out. 3039 * Return 3040 * 0 on success, or a negative error in case of failure. 3041 * 3042 * long bpf_send_signal_thread(u32 sig) 3043 * Description 3044 * Send signal *sig* to the thread corresponding to the current task. 3045 * Return 3046 * 0 on success or successfully queued. 3047 * 3048 * **-EBUSY** if work queue under nmi is full. 3049 * 3050 * **-EINVAL** if *sig* is invalid. 3051 * 3052 * **-EPERM** if no permission to send the *sig*. 3053 * 3054 * **-EAGAIN** if bpf program can try again. 3055 * 3056 * u64 bpf_jiffies64(void) 3057 * Description 3058 * Obtain the 64bit jiffies 3059 * Return 3060 * The 64 bit jiffies 3061 * 3062 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3063 * Description 3064 * For an eBPF program attached to a perf event, retrieve the 3065 * branch records (**struct perf_branch_entry**) associated to *ctx* 3066 * and store it in the buffer pointed by *buf* up to size 3067 * *size* bytes. 3068 * Return 3069 * On success, number of bytes written to *buf*. On error, a 3070 * negative value. 3071 * 3072 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3073 * instead return the number of bytes required to store all the 3074 * branch entries. If this flag is set, *buf* may be NULL. 3075 * 3076 * **-EINVAL** if arguments invalid or **size** not a multiple 3077 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3078 * 3079 * **-ENOENT** if architecture does not support branch records. 3080 * 3081 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3082 * Description 3083 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3084 * *namespace* will be returned in *nsdata*. 3085 * Return 3086 * 0 on success, or one of the following in case of failure: 3087 * 3088 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3089 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3090 * 3091 * **-ENOENT** if pidns does not exists for the current task. 3092 * 3093 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3094 * Description 3095 * Write raw *data* blob into a special BPF perf event held by 3096 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3097 * event must have the following attributes: **PERF_SAMPLE_RAW** 3098 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3099 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3100 * 3101 * The *flags* are used to indicate the index in *map* for which 3102 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3103 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3104 * to indicate that the index of the current CPU core should be 3105 * used. 3106 * 3107 * The value to write, of *size*, is passed through eBPF stack and 3108 * pointed by *data*. 3109 * 3110 * *ctx* is a pointer to in-kernel struct xdp_buff. 3111 * 3112 * This helper is similar to **bpf_perf_eventoutput**\ () but 3113 * restricted to raw_tracepoint bpf programs. 3114 * Return 3115 * 0 on success, or a negative error in case of failure. 3116 * 3117 * u64 bpf_get_netns_cookie(void *ctx) 3118 * Description 3119 * Retrieve the cookie (generated by the kernel) of the network 3120 * namespace the input *ctx* is associated with. The network 3121 * namespace cookie remains stable for its lifetime and provides 3122 * a global identifier that can be assumed unique. If *ctx* is 3123 * NULL, then the helper returns the cookie for the initial 3124 * network namespace. The cookie itself is very similar to that 3125 * of **bpf_get_socket_cookie**\ () helper, but for network 3126 * namespaces instead of sockets. 3127 * Return 3128 * A 8-byte long opaque number. 3129 * 3130 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3131 * Description 3132 * Return id of cgroup v2 that is ancestor of the cgroup associated 3133 * with the current task at the *ancestor_level*. The root cgroup 3134 * is at *ancestor_level* zero and each step down the hierarchy 3135 * increments the level. If *ancestor_level* == level of cgroup 3136 * associated with the current task, then return value will be the 3137 * same as that of **bpf_get_current_cgroup_id**\ (). 3138 * 3139 * The helper is useful to implement policies based on cgroups 3140 * that are upper in hierarchy than immediate cgroup associated 3141 * with the current task. 3142 * 3143 * The format of returned id and helper limitations are same as in 3144 * **bpf_get_current_cgroup_id**\ (). 3145 * Return 3146 * The id is returned or 0 in case the id could not be retrieved. 3147 * 3148 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3149 * Description 3150 * Helper is overloaded depending on BPF program type. This 3151 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3152 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3153 * 3154 * Assign the *sk* to the *skb*. When combined with appropriate 3155 * routing configuration to receive the packet towards the socket, 3156 * will cause *skb* to be delivered to the specified socket. 3157 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3158 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3159 * interfere with successful delivery to the socket. 3160 * 3161 * This operation is only valid from TC ingress path. 3162 * 3163 * The *flags* argument must be zero. 3164 * Return 3165 * 0 on success, or a negative error in case of failure: 3166 * 3167 * **-EINVAL** if specified *flags* are not supported. 3168 * 3169 * **-ENOENT** if the socket is unavailable for assignment. 3170 * 3171 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3172 * 3173 * **-EOPNOTSUPP** if the operation is not supported, for example 3174 * a call from outside of TC ingress. 3175 * 3176 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3177 * (reuseport). 3178 * 3179 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3180 * Description 3181 * Helper is overloaded depending on BPF program type. This 3182 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3183 * 3184 * Select the *sk* as a result of a socket lookup. 3185 * 3186 * For the operation to succeed passed socket must be compatible 3187 * with the packet description provided by the *ctx* object. 3188 * 3189 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3190 * be an exact match. While IP family (**AF_INET** or 3191 * **AF_INET6**) must be compatible, that is IPv6 sockets 3192 * that are not v6-only can be selected for IPv4 packets. 3193 * 3194 * Only TCP listeners and UDP unconnected sockets can be 3195 * selected. *sk* can also be NULL to reset any previous 3196 * selection. 3197 * 3198 * *flags* argument can combination of following values: 3199 * 3200 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3201 * socket selection, potentially done by a BPF program 3202 * that ran before us. 3203 * 3204 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3205 * load-balancing within reuseport group for the socket 3206 * being selected. 3207 * 3208 * On success *ctx->sk* will point to the selected socket. 3209 * 3210 * Return 3211 * 0 on success, or a negative errno in case of failure. 3212 * 3213 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3214 * not compatible with packet family (*ctx->family*). 3215 * 3216 * * **-EEXIST** if socket has been already selected, 3217 * potentially by another program, and 3218 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3219 * 3220 * * **-EINVAL** if unsupported flags were specified. 3221 * 3222 * * **-EPROTOTYPE** if socket L4 protocol 3223 * (*sk->protocol*) doesn't match packet protocol 3224 * (*ctx->protocol*). 3225 * 3226 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3227 * state (TCP listening or UDP unconnected). 3228 * 3229 * u64 bpf_ktime_get_boot_ns(void) 3230 * Description 3231 * Return the time elapsed since system boot, in nanoseconds. 3232 * Does include the time the system was suspended. 3233 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3234 * Return 3235 * Current *ktime*. 3236 * 3237 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3238 * Description 3239 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3240 * out the format string. 3241 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3242 * the format string itself. The *data* and *data_len* are format string 3243 * arguments. The *data* are a **u64** array and corresponding format string 3244 * values are stored in the array. For strings and pointers where pointees 3245 * are accessed, only the pointer values are stored in the *data* array. 3246 * The *data_len* is the size of *data* in bytes. 3247 * 3248 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3249 * Reading kernel memory may fail due to either invalid address or 3250 * valid address but requiring a major memory fault. If reading kernel memory 3251 * fails, the string for **%s** will be an empty string, and the ip 3252 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3253 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3254 * Return 3255 * 0 on success, or a negative error in case of failure: 3256 * 3257 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3258 * by returning 1 from bpf program. 3259 * 3260 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3261 * 3262 * **-E2BIG** if *fmt* contains too many format specifiers. 3263 * 3264 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3265 * 3266 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 3267 * Description 3268 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3269 * The *m* represents the seq_file. The *data* and *len* represent the 3270 * data to write in bytes. 3271 * Return 3272 * 0 on success, or a negative error in case of failure: 3273 * 3274 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3275 * 3276 * u64 bpf_sk_cgroup_id(void *sk) 3277 * Description 3278 * Return the cgroup v2 id of the socket *sk*. 3279 * 3280 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 3281 * returned from **bpf_sk_lookup_xxx**\ (), 3282 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3283 * same as in **bpf_skb_cgroup_id**\ (). 3284 * 3285 * This helper is available only if the kernel was compiled with 3286 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3287 * Return 3288 * The id is returned or 0 in case the id could not be retrieved. 3289 * 3290 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 3291 * Description 3292 * Return id of cgroup v2 that is ancestor of cgroup associated 3293 * with the *sk* at the *ancestor_level*. The root cgroup is at 3294 * *ancestor_level* zero and each step down the hierarchy 3295 * increments the level. If *ancestor_level* == level of cgroup 3296 * associated with *sk*, then return value will be same as that 3297 * of **bpf_sk_cgroup_id**\ (). 3298 * 3299 * The helper is useful to implement policies based on cgroups 3300 * that are upper in hierarchy than immediate cgroup associated 3301 * with *sk*. 3302 * 3303 * The format of returned id and helper limitations are same as in 3304 * **bpf_sk_cgroup_id**\ (). 3305 * Return 3306 * The id is returned or 0 in case the id could not be retrieved. 3307 * 3308 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 3309 * Description 3310 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3311 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3312 * of new data availability is sent. 3313 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3314 * of new data availability is sent unconditionally. 3315 * Return 3316 * 0 on success, or a negative error in case of failure. 3317 * 3318 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 3319 * Description 3320 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3321 * Return 3322 * Valid pointer with *size* bytes of memory available; NULL, 3323 * otherwise. 3324 * 3325 * void bpf_ringbuf_submit(void *data, u64 flags) 3326 * Description 3327 * Submit reserved ring buffer sample, pointed to by *data*. 3328 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3329 * of new data availability is sent. 3330 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3331 * of new data availability is sent unconditionally. 3332 * Return 3333 * Nothing. Always succeeds. 3334 * 3335 * void bpf_ringbuf_discard(void *data, u64 flags) 3336 * Description 3337 * Discard 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 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 3346 * Description 3347 * Query various characteristics of provided ring buffer. What 3348 * exactly is queries is determined by *flags*: 3349 * 3350 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3351 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3352 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3353 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3354 * 3355 * Data returned is just a momentary snapshot of actual values 3356 * and could be inaccurate, so this facility should be used to 3357 * power heuristics and for reporting, not to make 100% correct 3358 * calculation. 3359 * Return 3360 * Requested value, or 0, if *flags* are not recognized. 3361 * 3362 * long bpf_csum_level(struct sk_buff *skb, u64 level) 3363 * Description 3364 * Change the skbs checksum level by one layer up or down, or 3365 * reset it entirely to none in order to have the stack perform 3366 * checksum validation. The level is applicable to the following 3367 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3368 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3369 * through **bpf_skb_adjust_room**\ () helper with passing in 3370 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3371 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3372 * the UDP header is removed. Similarly, an encap of the latter 3373 * into the former could be accompanied by a helper call to 3374 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3375 * skb is still intended to be processed in higher layers of the 3376 * stack instead of just egressing at tc. 3377 * 3378 * There are three supported level settings at this time: 3379 * 3380 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3381 * with CHECKSUM_UNNECESSARY. 3382 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3383 * with CHECKSUM_UNNECESSARY. 3384 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3385 * sets CHECKSUM_NONE to force checksum validation by the stack. 3386 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3387 * skb->csum_level. 3388 * Return 3389 * 0 on success, or a negative error in case of failure. In the 3390 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3391 * is returned or the error code -EACCES in case the skb is not 3392 * subject to CHECKSUM_UNNECESSARY. 3393 * 3394 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 3395 * Description 3396 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3397 * Return 3398 * *sk* if casting is valid, or **NULL** otherwise. 3399 * 3400 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 3401 * Description 3402 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3403 * Return 3404 * *sk* if casting is valid, or **NULL** otherwise. 3405 * 3406 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 3407 * Description 3408 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3409 * Return 3410 * *sk* if casting is valid, or **NULL** otherwise. 3411 * 3412 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 3413 * Description 3414 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3415 * Return 3416 * *sk* if casting is valid, or **NULL** otherwise. 3417 * 3418 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 3419 * Description 3420 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3421 * Return 3422 * *sk* if casting is valid, or **NULL** otherwise. 3423 * 3424 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 3425 * Description 3426 * Return a user or a kernel stack in bpf program provided buffer. 3427 * Note: the user stack will only be populated if the *task* is 3428 * the current task; all other tasks will return -EOPNOTSUPP. 3429 * To achieve this, the helper needs *task*, which is a valid 3430 * pointer to **struct task_struct**. To store the stacktrace, the 3431 * bpf program provides *buf* with a nonnegative *size*. 3432 * 3433 * The last argument, *flags*, holds the number of stack frames to 3434 * skip (from 0 to 255), masked with 3435 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3436 * the following flags: 3437 * 3438 * **BPF_F_USER_STACK** 3439 * Collect a user space stack instead of a kernel stack. 3440 * The *task* must be the current task. 3441 * **BPF_F_USER_BUILD_ID** 3442 * Collect buildid+offset instead of ips for user stack, 3443 * only valid if **BPF_F_USER_STACK** is also specified. 3444 * 3445 * **bpf_get_task_stack**\ () can collect up to 3446 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3447 * to sufficient large buffer size. Note that 3448 * this limit can be controlled with the **sysctl** program, and 3449 * that it should be manually increased in order to profile long 3450 * user stacks (such as stacks for Java programs). To do so, use: 3451 * 3452 * :: 3453 * 3454 * # sysctl kernel.perf_event_max_stack=<new value> 3455 * Return 3456 * The non-negative copied *buf* length equal to or less than 3457 * *size* on success, or a negative error in case of failure. 3458 * 3459 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 3460 * Description 3461 * Load header option. Support reading a particular TCP header 3462 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 3463 * 3464 * If *flags* is 0, it will search the option from the 3465 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 3466 * has details on what skb_data contains under different 3467 * *skops*\ **->op**. 3468 * 3469 * The first byte of the *searchby_res* specifies the 3470 * kind that it wants to search. 3471 * 3472 * If the searching kind is an experimental kind 3473 * (i.e. 253 or 254 according to RFC6994). It also 3474 * needs to specify the "magic" which is either 3475 * 2 bytes or 4 bytes. It then also needs to 3476 * specify the size of the magic by using 3477 * the 2nd byte which is "kind-length" of a TCP 3478 * header option and the "kind-length" also 3479 * includes the first 2 bytes "kind" and "kind-length" 3480 * itself as a normal TCP header option also does. 3481 * 3482 * For example, to search experimental kind 254 with 3483 * 2 byte magic 0xeB9F, the searchby_res should be 3484 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 3485 * 3486 * To search for the standard window scale option (3), 3487 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 3488 * Note, kind-length must be 0 for regular option. 3489 * 3490 * Searching for No-Op (0) and End-of-Option-List (1) are 3491 * not supported. 3492 * 3493 * *len* must be at least 2 bytes which is the minimal size 3494 * of a header option. 3495 * 3496 * Supported flags: 3497 * 3498 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 3499 * saved_syn packet or the just-received syn packet. 3500 * 3501 * Return 3502 * > 0 when found, the header option is copied to *searchby_res*. 3503 * The return value is the total length copied. On failure, a 3504 * negative error code is returned: 3505 * 3506 * **-EINVAL** if a parameter is invalid. 3507 * 3508 * **-ENOMSG** if the option is not found. 3509 * 3510 * **-ENOENT** if no syn packet is available when 3511 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 3512 * 3513 * **-ENOSPC** if there is not enough space. Only *len* number of 3514 * bytes are copied. 3515 * 3516 * **-EFAULT** on failure to parse the header options in the 3517 * packet. 3518 * 3519 * **-EPERM** if the helper cannot be used under the current 3520 * *skops*\ **->op**. 3521 * 3522 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 3523 * Description 3524 * Store header option. The data will be copied 3525 * from buffer *from* with length *len* to the TCP header. 3526 * 3527 * The buffer *from* should have the whole option that 3528 * includes the kind, kind-length, and the actual 3529 * option data. The *len* must be at least kind-length 3530 * long. The kind-length does not have to be 4 byte 3531 * aligned. The kernel will take care of the padding 3532 * and setting the 4 bytes aligned value to th->doff. 3533 * 3534 * This helper will check for duplicated option 3535 * by searching the same option in the outgoing skb. 3536 * 3537 * This helper can only be called during 3538 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3539 * 3540 * Return 3541 * 0 on success, or negative error in case of failure: 3542 * 3543 * **-EINVAL** If param is invalid. 3544 * 3545 * **-ENOSPC** if there is not enough space in the header. 3546 * Nothing has been written 3547 * 3548 * **-EEXIST** if the option already exists. 3549 * 3550 * **-EFAULT** on failrue to parse the existing header options. 3551 * 3552 * **-EPERM** if the helper cannot be used under the current 3553 * *skops*\ **->op**. 3554 * 3555 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 3556 * Description 3557 * Reserve *len* bytes for the bpf header option. The 3558 * space will be used by **bpf_store_hdr_opt**\ () later in 3559 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3560 * 3561 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 3562 * the total number of bytes will be reserved. 3563 * 3564 * This helper can only be called during 3565 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 3566 * 3567 * Return 3568 * 0 on success, or negative error in case of failure: 3569 * 3570 * **-EINVAL** if a parameter is invalid. 3571 * 3572 * **-ENOSPC** if there is not enough space in the header. 3573 * 3574 * **-EPERM** if the helper cannot be used under the current 3575 * *skops*\ **->op**. 3576 * 3577 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 3578 * Description 3579 * Get a bpf_local_storage from an *inode*. 3580 * 3581 * Logically, it could be thought of as getting the value from 3582 * a *map* with *inode* as the **key**. From this 3583 * perspective, the usage is not much different from 3584 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 3585 * helper enforces the key must be an inode and the map must also 3586 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 3587 * 3588 * Underneath, the value is stored locally at *inode* instead of 3589 * the *map*. The *map* is used as the bpf-local-storage 3590 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3591 * searched against all bpf_local_storage residing at *inode*. 3592 * 3593 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3594 * used such that a new bpf_local_storage will be 3595 * created if one does not exist. *value* can be used 3596 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3597 * the initial value of a bpf_local_storage. If *value* is 3598 * **NULL**, the new bpf_local_storage will be zero initialized. 3599 * Return 3600 * A bpf_local_storage pointer is returned on success. 3601 * 3602 * **NULL** if not found or there was an error in adding 3603 * a new bpf_local_storage. 3604 * 3605 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 3606 * Description 3607 * Delete a bpf_local_storage from an *inode*. 3608 * Return 3609 * 0 on success. 3610 * 3611 * **-ENOENT** if the bpf_local_storage cannot be found. 3612 * 3613 * long bpf_d_path(struct path *path, char *buf, u32 sz) 3614 * Description 3615 * Return full path for given **struct path** object, which 3616 * needs to be the kernel BTF *path* object. The path is 3617 * returned in the provided buffer *buf* of size *sz* and 3618 * is zero terminated. 3619 * 3620 * Return 3621 * On success, the strictly positive length of the string, 3622 * including the trailing NUL character. On error, a negative 3623 * value. 3624 * 3625 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 3626 * Description 3627 * Read *size* bytes from user space address *user_ptr* and store 3628 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 3629 * Return 3630 * 0 on success, or a negative error in case of failure. 3631 * 3632 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 3633 * Description 3634 * Use BTF to store a string representation of *ptr*->ptr in *str*, 3635 * using *ptr*->type_id. This value should specify the type 3636 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 3637 * can be used to look up vmlinux BTF type ids. Traversing the 3638 * data structure using BTF, the type information and values are 3639 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 3640 * the pointer data is carried out to avoid kernel crashes during 3641 * operation. Smaller types can use string space on the stack; 3642 * larger programs can use map data to store the string 3643 * representation. 3644 * 3645 * The string can be subsequently shared with userspace via 3646 * bpf_perf_event_output() or ring buffer interfaces. 3647 * bpf_trace_printk() is to be avoided as it places too small 3648 * a limit on string size to be useful. 3649 * 3650 * *flags* is a combination of 3651 * 3652 * **BTF_F_COMPACT** 3653 * no formatting around type information 3654 * **BTF_F_NONAME** 3655 * no struct/union member names/types 3656 * **BTF_F_PTR_RAW** 3657 * show raw (unobfuscated) pointer values; 3658 * equivalent to printk specifier %px. 3659 * **BTF_F_ZERO** 3660 * show zero-valued struct/union members; they 3661 * are not displayed by default 3662 * 3663 * Return 3664 * The number of bytes that were written (or would have been 3665 * written if output had to be truncated due to string size), 3666 * or a negative error in cases of failure. 3667 * 3668 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 3669 * Description 3670 * Use BTF to write to seq_write a string representation of 3671 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 3672 * *flags* are identical to those used for bpf_snprintf_btf. 3673 * Return 3674 * 0 on success or a negative error in case of failure. 3675 * 3676 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 3677 * Description 3678 * See **bpf_get_cgroup_classid**\ () for the main description. 3679 * This helper differs from **bpf_get_cgroup_classid**\ () in that 3680 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 3681 * associated socket instead of the current process. 3682 * Return 3683 * The id is returned or 0 in case the id could not be retrieved. 3684 * 3685 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 3686 * Description 3687 * Redirect the packet to another net device of index *ifindex* 3688 * and fill in L2 addresses from neighboring subsystem. This helper 3689 * is somewhat similar to **bpf_redirect**\ (), except that it 3690 * populates L2 addresses as well, meaning, internally, the helper 3691 * relies on the neighbor lookup for the L2 address of the nexthop. 3692 * 3693 * The helper will perform a FIB lookup based on the skb's 3694 * networking header to get the address of the next hop, unless 3695 * this is supplied by the caller in the *params* argument. The 3696 * *plen* argument indicates the len of *params* and should be set 3697 * to 0 if *params* is NULL. 3698 * 3699 * The *flags* argument is reserved and must be 0. The helper is 3700 * currently only supported for tc BPF program types, and enabled 3701 * for IPv4 and IPv6 protocols. 3702 * Return 3703 * The helper returns **TC_ACT_REDIRECT** on success or 3704 * **TC_ACT_SHOT** on error. 3705 * 3706 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 3707 * Description 3708 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3709 * pointer to the percpu kernel variable on *cpu*. A ksym is an 3710 * extern variable decorated with '__ksym'. For ksym, there is a 3711 * global var (either static or global) defined of the same name 3712 * in the kernel. The ksym is percpu if the global var is percpu. 3713 * The returned pointer points to the global percpu var on *cpu*. 3714 * 3715 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 3716 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 3717 * happens if *cpu* is larger than nr_cpu_ids. The caller of 3718 * bpf_per_cpu_ptr() must check the returned value. 3719 * Return 3720 * A pointer pointing to the kernel percpu variable on *cpu*, or 3721 * NULL, if *cpu* is invalid. 3722 * 3723 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 3724 * Description 3725 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3726 * pointer to the percpu kernel variable on this cpu. See the 3727 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 3728 * 3729 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 3730 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 3731 * never return NULL. 3732 * Return 3733 * A pointer pointing to the kernel percpu variable on this cpu. 3734 * 3735 * long bpf_redirect_peer(u32 ifindex, u64 flags) 3736 * Description 3737 * Redirect the packet to another net device of index *ifindex*. 3738 * This helper is somewhat similar to **bpf_redirect**\ (), except 3739 * that the redirection happens to the *ifindex*' peer device and 3740 * the netns switch takes place from ingress to ingress without 3741 * going through the CPU's backlog queue. 3742 * 3743 * The *flags* argument is reserved and must be 0. The helper is 3744 * currently only supported for tc BPF program types at the ingress 3745 * hook and for veth device types. The peer device must reside in a 3746 * different network namespace. 3747 * Return 3748 * The helper returns **TC_ACT_REDIRECT** on success or 3749 * **TC_ACT_SHOT** on error. 3750 */ 3751 #define __BPF_FUNC_MAPPER(FN) \ 3752 FN(unspec), \ 3753 FN(map_lookup_elem), \ 3754 FN(map_update_elem), \ 3755 FN(map_delete_elem), \ 3756 FN(probe_read), \ 3757 FN(ktime_get_ns), \ 3758 FN(trace_printk), \ 3759 FN(get_prandom_u32), \ 3760 FN(get_smp_processor_id), \ 3761 FN(skb_store_bytes), \ 3762 FN(l3_csum_replace), \ 3763 FN(l4_csum_replace), \ 3764 FN(tail_call), \ 3765 FN(clone_redirect), \ 3766 FN(get_current_pid_tgid), \ 3767 FN(get_current_uid_gid), \ 3768 FN(get_current_comm), \ 3769 FN(get_cgroup_classid), \ 3770 FN(skb_vlan_push), \ 3771 FN(skb_vlan_pop), \ 3772 FN(skb_get_tunnel_key), \ 3773 FN(skb_set_tunnel_key), \ 3774 FN(perf_event_read), \ 3775 FN(redirect), \ 3776 FN(get_route_realm), \ 3777 FN(perf_event_output), \ 3778 FN(skb_load_bytes), \ 3779 FN(get_stackid), \ 3780 FN(csum_diff), \ 3781 FN(skb_get_tunnel_opt), \ 3782 FN(skb_set_tunnel_opt), \ 3783 FN(skb_change_proto), \ 3784 FN(skb_change_type), \ 3785 FN(skb_under_cgroup), \ 3786 FN(get_hash_recalc), \ 3787 FN(get_current_task), \ 3788 FN(probe_write_user), \ 3789 FN(current_task_under_cgroup), \ 3790 FN(skb_change_tail), \ 3791 FN(skb_pull_data), \ 3792 FN(csum_update), \ 3793 FN(set_hash_invalid), \ 3794 FN(get_numa_node_id), \ 3795 FN(skb_change_head), \ 3796 FN(xdp_adjust_head), \ 3797 FN(probe_read_str), \ 3798 FN(get_socket_cookie), \ 3799 FN(get_socket_uid), \ 3800 FN(set_hash), \ 3801 FN(setsockopt), \ 3802 FN(skb_adjust_room), \ 3803 FN(redirect_map), \ 3804 FN(sk_redirect_map), \ 3805 FN(sock_map_update), \ 3806 FN(xdp_adjust_meta), \ 3807 FN(perf_event_read_value), \ 3808 FN(perf_prog_read_value), \ 3809 FN(getsockopt), \ 3810 FN(override_return), \ 3811 FN(sock_ops_cb_flags_set), \ 3812 FN(msg_redirect_map), \ 3813 FN(msg_apply_bytes), \ 3814 FN(msg_cork_bytes), \ 3815 FN(msg_pull_data), \ 3816 FN(bind), \ 3817 FN(xdp_adjust_tail), \ 3818 FN(skb_get_xfrm_state), \ 3819 FN(get_stack), \ 3820 FN(skb_load_bytes_relative), \ 3821 FN(fib_lookup), \ 3822 FN(sock_hash_update), \ 3823 FN(msg_redirect_hash), \ 3824 FN(sk_redirect_hash), \ 3825 FN(lwt_push_encap), \ 3826 FN(lwt_seg6_store_bytes), \ 3827 FN(lwt_seg6_adjust_srh), \ 3828 FN(lwt_seg6_action), \ 3829 FN(rc_repeat), \ 3830 FN(rc_keydown), \ 3831 FN(skb_cgroup_id), \ 3832 FN(get_current_cgroup_id), \ 3833 FN(get_local_storage), \ 3834 FN(sk_select_reuseport), \ 3835 FN(skb_ancestor_cgroup_id), \ 3836 FN(sk_lookup_tcp), \ 3837 FN(sk_lookup_udp), \ 3838 FN(sk_release), \ 3839 FN(map_push_elem), \ 3840 FN(map_pop_elem), \ 3841 FN(map_peek_elem), \ 3842 FN(msg_push_data), \ 3843 FN(msg_pop_data), \ 3844 FN(rc_pointer_rel), \ 3845 FN(spin_lock), \ 3846 FN(spin_unlock), \ 3847 FN(sk_fullsock), \ 3848 FN(tcp_sock), \ 3849 FN(skb_ecn_set_ce), \ 3850 FN(get_listener_sock), \ 3851 FN(skc_lookup_tcp), \ 3852 FN(tcp_check_syncookie), \ 3853 FN(sysctl_get_name), \ 3854 FN(sysctl_get_current_value), \ 3855 FN(sysctl_get_new_value), \ 3856 FN(sysctl_set_new_value), \ 3857 FN(strtol), \ 3858 FN(strtoul), \ 3859 FN(sk_storage_get), \ 3860 FN(sk_storage_delete), \ 3861 FN(send_signal), \ 3862 FN(tcp_gen_syncookie), \ 3863 FN(skb_output), \ 3864 FN(probe_read_user), \ 3865 FN(probe_read_kernel), \ 3866 FN(probe_read_user_str), \ 3867 FN(probe_read_kernel_str), \ 3868 FN(tcp_send_ack), \ 3869 FN(send_signal_thread), \ 3870 FN(jiffies64), \ 3871 FN(read_branch_records), \ 3872 FN(get_ns_current_pid_tgid), \ 3873 FN(xdp_output), \ 3874 FN(get_netns_cookie), \ 3875 FN(get_current_ancestor_cgroup_id), \ 3876 FN(sk_assign), \ 3877 FN(ktime_get_boot_ns), \ 3878 FN(seq_printf), \ 3879 FN(seq_write), \ 3880 FN(sk_cgroup_id), \ 3881 FN(sk_ancestor_cgroup_id), \ 3882 FN(ringbuf_output), \ 3883 FN(ringbuf_reserve), \ 3884 FN(ringbuf_submit), \ 3885 FN(ringbuf_discard), \ 3886 FN(ringbuf_query), \ 3887 FN(csum_level), \ 3888 FN(skc_to_tcp6_sock), \ 3889 FN(skc_to_tcp_sock), \ 3890 FN(skc_to_tcp_timewait_sock), \ 3891 FN(skc_to_tcp_request_sock), \ 3892 FN(skc_to_udp6_sock), \ 3893 FN(get_task_stack), \ 3894 FN(load_hdr_opt), \ 3895 FN(store_hdr_opt), \ 3896 FN(reserve_hdr_opt), \ 3897 FN(inode_storage_get), \ 3898 FN(inode_storage_delete), \ 3899 FN(d_path), \ 3900 FN(copy_from_user), \ 3901 FN(snprintf_btf), \ 3902 FN(seq_printf_btf), \ 3903 FN(skb_cgroup_classid), \ 3904 FN(redirect_neigh), \ 3905 FN(per_cpu_ptr), \ 3906 FN(this_cpu_ptr), \ 3907 FN(redirect_peer), \ 3908 /* */ 3909 3910 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3911 * function eBPF program intends to call 3912 */ 3913 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3914 enum bpf_func_id { 3915 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3916 __BPF_FUNC_MAX_ID, 3917 }; 3918 #undef __BPF_ENUM_FN 3919 3920 /* All flags used by eBPF helper functions, placed here. */ 3921 3922 /* BPF_FUNC_skb_store_bytes flags. */ 3923 enum { 3924 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 3925 BPF_F_INVALIDATE_HASH = (1ULL << 1), 3926 }; 3927 3928 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3929 * First 4 bits are for passing the header field size. 3930 */ 3931 enum { 3932 BPF_F_HDR_FIELD_MASK = 0xfULL, 3933 }; 3934 3935 /* BPF_FUNC_l4_csum_replace flags. */ 3936 enum { 3937 BPF_F_PSEUDO_HDR = (1ULL << 4), 3938 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 3939 BPF_F_MARK_ENFORCE = (1ULL << 6), 3940 }; 3941 3942 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3943 enum { 3944 BPF_F_INGRESS = (1ULL << 0), 3945 }; 3946 3947 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3948 enum { 3949 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 3950 }; 3951 3952 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3953 enum { 3954 BPF_F_SKIP_FIELD_MASK = 0xffULL, 3955 BPF_F_USER_STACK = (1ULL << 8), 3956 /* flags used by BPF_FUNC_get_stackid only. */ 3957 BPF_F_FAST_STACK_CMP = (1ULL << 9), 3958 BPF_F_REUSE_STACKID = (1ULL << 10), 3959 /* flags used by BPF_FUNC_get_stack only. */ 3960 BPF_F_USER_BUILD_ID = (1ULL << 11), 3961 }; 3962 3963 /* BPF_FUNC_skb_set_tunnel_key flags. */ 3964 enum { 3965 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 3966 BPF_F_DONT_FRAGMENT = (1ULL << 2), 3967 BPF_F_SEQ_NUMBER = (1ULL << 3), 3968 }; 3969 3970 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3971 * BPF_FUNC_perf_event_read_value flags. 3972 */ 3973 enum { 3974 BPF_F_INDEX_MASK = 0xffffffffULL, 3975 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 3976 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 3977 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 3978 }; 3979 3980 /* Current network namespace */ 3981 enum { 3982 BPF_F_CURRENT_NETNS = (-1L), 3983 }; 3984 3985 /* BPF_FUNC_csum_level level values. */ 3986 enum { 3987 BPF_CSUM_LEVEL_QUERY, 3988 BPF_CSUM_LEVEL_INC, 3989 BPF_CSUM_LEVEL_DEC, 3990 BPF_CSUM_LEVEL_RESET, 3991 }; 3992 3993 /* BPF_FUNC_skb_adjust_room flags. */ 3994 enum { 3995 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 3996 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 3997 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 3998 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 3999 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 4000 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 4001 }; 4002 4003 enum { 4004 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 4005 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 4006 }; 4007 4008 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 4009 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 4010 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 4011 4012 /* BPF_FUNC_sysctl_get_name flags. */ 4013 enum { 4014 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 4015 }; 4016 4017 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 4018 enum { 4019 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 4020 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 4021 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 4022 */ 4023 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 4024 }; 4025 4026 /* BPF_FUNC_read_branch_records flags. */ 4027 enum { 4028 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 4029 }; 4030 4031 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 4032 * BPF_FUNC_bpf_ringbuf_output flags. 4033 */ 4034 enum { 4035 BPF_RB_NO_WAKEUP = (1ULL << 0), 4036 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 4037 }; 4038 4039 /* BPF_FUNC_bpf_ringbuf_query flags */ 4040 enum { 4041 BPF_RB_AVAIL_DATA = 0, 4042 BPF_RB_RING_SIZE = 1, 4043 BPF_RB_CONS_POS = 2, 4044 BPF_RB_PROD_POS = 3, 4045 }; 4046 4047 /* BPF ring buffer constants */ 4048 enum { 4049 BPF_RINGBUF_BUSY_BIT = (1U << 31), 4050 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 4051 BPF_RINGBUF_HDR_SZ = 8, 4052 }; 4053 4054 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 4055 enum { 4056 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 4057 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 4058 }; 4059 4060 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 4061 enum bpf_adj_room_mode { 4062 BPF_ADJ_ROOM_NET, 4063 BPF_ADJ_ROOM_MAC, 4064 }; 4065 4066 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 4067 enum bpf_hdr_start_off { 4068 BPF_HDR_START_MAC, 4069 BPF_HDR_START_NET, 4070 }; 4071 4072 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 4073 enum bpf_lwt_encap_mode { 4074 BPF_LWT_ENCAP_SEG6, 4075 BPF_LWT_ENCAP_SEG6_INLINE, 4076 BPF_LWT_ENCAP_IP, 4077 }; 4078 4079 #define __bpf_md_ptr(type, name) \ 4080 union { \ 4081 type name; \ 4082 __u64 :64; \ 4083 } __attribute__((aligned(8))) 4084 4085 /* user accessible mirror of in-kernel sk_buff. 4086 * new fields can only be added to the end of this structure 4087 */ 4088 struct __sk_buff { 4089 __u32 len; 4090 __u32 pkt_type; 4091 __u32 mark; 4092 __u32 queue_mapping; 4093 __u32 protocol; 4094 __u32 vlan_present; 4095 __u32 vlan_tci; 4096 __u32 vlan_proto; 4097 __u32 priority; 4098 __u32 ingress_ifindex; 4099 __u32 ifindex; 4100 __u32 tc_index; 4101 __u32 cb[5]; 4102 __u32 hash; 4103 __u32 tc_classid; 4104 __u32 data; 4105 __u32 data_end; 4106 __u32 napi_id; 4107 4108 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 4109 __u32 family; 4110 __u32 remote_ip4; /* Stored in network byte order */ 4111 __u32 local_ip4; /* Stored in network byte order */ 4112 __u32 remote_ip6[4]; /* Stored in network byte order */ 4113 __u32 local_ip6[4]; /* Stored in network byte order */ 4114 __u32 remote_port; /* Stored in network byte order */ 4115 __u32 local_port; /* stored in host byte order */ 4116 /* ... here. */ 4117 4118 __u32 data_meta; 4119 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 4120 __u64 tstamp; 4121 __u32 wire_len; 4122 __u32 gso_segs; 4123 __bpf_md_ptr(struct bpf_sock *, sk); 4124 __u32 gso_size; 4125 }; 4126 4127 struct bpf_tunnel_key { 4128 __u32 tunnel_id; 4129 union { 4130 __u32 remote_ipv4; 4131 __u32 remote_ipv6[4]; 4132 }; 4133 __u8 tunnel_tos; 4134 __u8 tunnel_ttl; 4135 __u16 tunnel_ext; /* Padding, future use. */ 4136 __u32 tunnel_label; 4137 }; 4138 4139 /* user accessible mirror of in-kernel xfrm_state. 4140 * new fields can only be added to the end of this structure 4141 */ 4142 struct bpf_xfrm_state { 4143 __u32 reqid; 4144 __u32 spi; /* Stored in network byte order */ 4145 __u16 family; 4146 __u16 ext; /* Padding, future use. */ 4147 union { 4148 __u32 remote_ipv4; /* Stored in network byte order */ 4149 __u32 remote_ipv6[4]; /* Stored in network byte order */ 4150 }; 4151 }; 4152 4153 /* Generic BPF return codes which all BPF program types may support. 4154 * The values are binary compatible with their TC_ACT_* counter-part to 4155 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 4156 * programs. 4157 * 4158 * XDP is handled seprately, see XDP_*. 4159 */ 4160 enum bpf_ret_code { 4161 BPF_OK = 0, 4162 /* 1 reserved */ 4163 BPF_DROP = 2, 4164 /* 3-6 reserved */ 4165 BPF_REDIRECT = 7, 4166 /* >127 are reserved for prog type specific return codes. 4167 * 4168 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 4169 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 4170 * changed and should be routed based on its new L3 header. 4171 * (This is an L3 redirect, as opposed to L2 redirect 4172 * represented by BPF_REDIRECT above). 4173 */ 4174 BPF_LWT_REROUTE = 128, 4175 }; 4176 4177 struct bpf_sock { 4178 __u32 bound_dev_if; 4179 __u32 family; 4180 __u32 type; 4181 __u32 protocol; 4182 __u32 mark; 4183 __u32 priority; 4184 /* IP address also allows 1 and 2 bytes access */ 4185 __u32 src_ip4; 4186 __u32 src_ip6[4]; 4187 __u32 src_port; /* host byte order */ 4188 __be16 dst_port; /* network byte order */ 4189 __u16 :16; /* zero padding */ 4190 __u32 dst_ip4; 4191 __u32 dst_ip6[4]; 4192 __u32 state; 4193 __s32 rx_queue_mapping; 4194 }; 4195 4196 struct bpf_tcp_sock { 4197 __u32 snd_cwnd; /* Sending congestion window */ 4198 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 4199 __u32 rtt_min; 4200 __u32 snd_ssthresh; /* Slow start size threshold */ 4201 __u32 rcv_nxt; /* What we want to receive next */ 4202 __u32 snd_nxt; /* Next sequence we send */ 4203 __u32 snd_una; /* First byte we want an ack for */ 4204 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 4205 __u32 ecn_flags; /* ECN status bits. */ 4206 __u32 rate_delivered; /* saved rate sample: packets delivered */ 4207 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 4208 __u32 packets_out; /* Packets which are "in flight" */ 4209 __u32 retrans_out; /* Retransmitted packets out */ 4210 __u32 total_retrans; /* Total retransmits for entire connection */ 4211 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 4212 * total number of segments in. 4213 */ 4214 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 4215 * total number of data segments in. 4216 */ 4217 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 4218 * The total number of segments sent. 4219 */ 4220 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 4221 * total number of data segments sent. 4222 */ 4223 __u32 lost_out; /* Lost packets */ 4224 __u32 sacked_out; /* SACK'd packets */ 4225 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 4226 * sum(delta(rcv_nxt)), or how many bytes 4227 * were acked. 4228 */ 4229 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 4230 * sum(delta(snd_una)), or how many bytes 4231 * were acked. 4232 */ 4233 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 4234 * total number of DSACK blocks received 4235 */ 4236 __u32 delivered; /* Total data packets delivered incl. rexmits */ 4237 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 4238 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 4239 }; 4240 4241 struct bpf_sock_tuple { 4242 union { 4243 struct { 4244 __be32 saddr; 4245 __be32 daddr; 4246 __be16 sport; 4247 __be16 dport; 4248 } ipv4; 4249 struct { 4250 __be32 saddr[4]; 4251 __be32 daddr[4]; 4252 __be16 sport; 4253 __be16 dport; 4254 } ipv6; 4255 }; 4256 }; 4257 4258 struct bpf_xdp_sock { 4259 __u32 queue_id; 4260 }; 4261 4262 #define XDP_PACKET_HEADROOM 256 4263 4264 /* User return codes for XDP prog type. 4265 * A valid XDP program must return one of these defined values. All other 4266 * return codes are reserved for future use. Unknown return codes will 4267 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 4268 */ 4269 enum xdp_action { 4270 XDP_ABORTED = 0, 4271 XDP_DROP, 4272 XDP_PASS, 4273 XDP_TX, 4274 XDP_REDIRECT, 4275 }; 4276 4277 /* user accessible metadata for XDP packet hook 4278 * new fields must be added to the end of this structure 4279 */ 4280 struct xdp_md { 4281 __u32 data; 4282 __u32 data_end; 4283 __u32 data_meta; 4284 /* Below access go through struct xdp_rxq_info */ 4285 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 4286 __u32 rx_queue_index; /* rxq->queue_index */ 4287 4288 __u32 egress_ifindex; /* txq->dev->ifindex */ 4289 }; 4290 4291 /* DEVMAP map-value layout 4292 * 4293 * The struct data-layout of map-value is a configuration interface. 4294 * New members can only be added to the end of this structure. 4295 */ 4296 struct bpf_devmap_val { 4297 __u32 ifindex; /* device index */ 4298 union { 4299 int fd; /* prog fd on map write */ 4300 __u32 id; /* prog id on map read */ 4301 } bpf_prog; 4302 }; 4303 4304 /* CPUMAP map-value layout 4305 * 4306 * The struct data-layout of map-value is a configuration interface. 4307 * New members can only be added to the end of this structure. 4308 */ 4309 struct bpf_cpumap_val { 4310 __u32 qsize; /* queue size to remote target CPU */ 4311 union { 4312 int fd; /* prog fd on map write */ 4313 __u32 id; /* prog id on map read */ 4314 } bpf_prog; 4315 }; 4316 4317 enum sk_action { 4318 SK_DROP = 0, 4319 SK_PASS, 4320 }; 4321 4322 /* user accessible metadata for SK_MSG packet hook, new fields must 4323 * be added to the end of this structure 4324 */ 4325 struct sk_msg_md { 4326 __bpf_md_ptr(void *, data); 4327 __bpf_md_ptr(void *, data_end); 4328 4329 __u32 family; 4330 __u32 remote_ip4; /* Stored in network byte order */ 4331 __u32 local_ip4; /* Stored in network byte order */ 4332 __u32 remote_ip6[4]; /* Stored in network byte order */ 4333 __u32 local_ip6[4]; /* Stored in network byte order */ 4334 __u32 remote_port; /* Stored in network byte order */ 4335 __u32 local_port; /* stored in host byte order */ 4336 __u32 size; /* Total size of sk_msg */ 4337 4338 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 4339 }; 4340 4341 struct sk_reuseport_md { 4342 /* 4343 * Start of directly accessible data. It begins from 4344 * the tcp/udp header. 4345 */ 4346 __bpf_md_ptr(void *, data); 4347 /* End of directly accessible data */ 4348 __bpf_md_ptr(void *, data_end); 4349 /* 4350 * Total length of packet (starting from the tcp/udp header). 4351 * Note that the directly accessible bytes (data_end - data) 4352 * could be less than this "len". Those bytes could be 4353 * indirectly read by a helper "bpf_skb_load_bytes()". 4354 */ 4355 __u32 len; 4356 /* 4357 * Eth protocol in the mac header (network byte order). e.g. 4358 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 4359 */ 4360 __u32 eth_protocol; 4361 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 4362 __u32 bind_inany; /* Is sock bound to an INANY address? */ 4363 __u32 hash; /* A hash of the packet 4 tuples */ 4364 }; 4365 4366 #define BPF_TAG_SIZE 8 4367 4368 struct bpf_prog_info { 4369 __u32 type; 4370 __u32 id; 4371 __u8 tag[BPF_TAG_SIZE]; 4372 __u32 jited_prog_len; 4373 __u32 xlated_prog_len; 4374 __aligned_u64 jited_prog_insns; 4375 __aligned_u64 xlated_prog_insns; 4376 __u64 load_time; /* ns since boottime */ 4377 __u32 created_by_uid; 4378 __u32 nr_map_ids; 4379 __aligned_u64 map_ids; 4380 char name[BPF_OBJ_NAME_LEN]; 4381 __u32 ifindex; 4382 __u32 gpl_compatible:1; 4383 __u32 :31; /* alignment pad */ 4384 __u64 netns_dev; 4385 __u64 netns_ino; 4386 __u32 nr_jited_ksyms; 4387 __u32 nr_jited_func_lens; 4388 __aligned_u64 jited_ksyms; 4389 __aligned_u64 jited_func_lens; 4390 __u32 btf_id; 4391 __u32 func_info_rec_size; 4392 __aligned_u64 func_info; 4393 __u32 nr_func_info; 4394 __u32 nr_line_info; 4395 __aligned_u64 line_info; 4396 __aligned_u64 jited_line_info; 4397 __u32 nr_jited_line_info; 4398 __u32 line_info_rec_size; 4399 __u32 jited_line_info_rec_size; 4400 __u32 nr_prog_tags; 4401 __aligned_u64 prog_tags; 4402 __u64 run_time_ns; 4403 __u64 run_cnt; 4404 } __attribute__((aligned(8))); 4405 4406 struct bpf_map_info { 4407 __u32 type; 4408 __u32 id; 4409 __u32 key_size; 4410 __u32 value_size; 4411 __u32 max_entries; 4412 __u32 map_flags; 4413 char name[BPF_OBJ_NAME_LEN]; 4414 __u32 ifindex; 4415 __u32 btf_vmlinux_value_type_id; 4416 __u64 netns_dev; 4417 __u64 netns_ino; 4418 __u32 btf_id; 4419 __u32 btf_key_type_id; 4420 __u32 btf_value_type_id; 4421 } __attribute__((aligned(8))); 4422 4423 struct bpf_btf_info { 4424 __aligned_u64 btf; 4425 __u32 btf_size; 4426 __u32 id; 4427 } __attribute__((aligned(8))); 4428 4429 struct bpf_link_info { 4430 __u32 type; 4431 __u32 id; 4432 __u32 prog_id; 4433 union { 4434 struct { 4435 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 4436 __u32 tp_name_len; /* in/out: tp_name buffer len */ 4437 } raw_tracepoint; 4438 struct { 4439 __u32 attach_type; 4440 } tracing; 4441 struct { 4442 __u64 cgroup_id; 4443 __u32 attach_type; 4444 } cgroup; 4445 struct { 4446 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 4447 __u32 target_name_len; /* in/out: target_name buffer len */ 4448 union { 4449 struct { 4450 __u32 map_id; 4451 } map; 4452 }; 4453 } iter; 4454 struct { 4455 __u32 netns_ino; 4456 __u32 attach_type; 4457 } netns; 4458 struct { 4459 __u32 ifindex; 4460 } xdp; 4461 }; 4462 } __attribute__((aligned(8))); 4463 4464 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 4465 * by user and intended to be used by socket (e.g. to bind to, depends on 4466 * attach type). 4467 */ 4468 struct bpf_sock_addr { 4469 __u32 user_family; /* Allows 4-byte read, but no write. */ 4470 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4471 * Stored in network byte order. 4472 */ 4473 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4474 * Stored in network byte order. 4475 */ 4476 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 4477 * Stored in network byte order 4478 */ 4479 __u32 family; /* Allows 4-byte read, but no write */ 4480 __u32 type; /* Allows 4-byte read, but no write */ 4481 __u32 protocol; /* Allows 4-byte read, but no write */ 4482 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4483 * Stored in network byte order. 4484 */ 4485 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4486 * Stored in network byte order. 4487 */ 4488 __bpf_md_ptr(struct bpf_sock *, sk); 4489 }; 4490 4491 /* User bpf_sock_ops struct to access socket values and specify request ops 4492 * and their replies. 4493 * Some of this fields are in network (bigendian) byte order and may need 4494 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 4495 * New fields can only be added at the end of this structure 4496 */ 4497 struct bpf_sock_ops { 4498 __u32 op; 4499 union { 4500 __u32 args[4]; /* Optionally passed to bpf program */ 4501 __u32 reply; /* Returned by bpf program */ 4502 __u32 replylong[4]; /* Optionally returned by bpf prog */ 4503 }; 4504 __u32 family; 4505 __u32 remote_ip4; /* Stored in network byte order */ 4506 __u32 local_ip4; /* Stored in network byte order */ 4507 __u32 remote_ip6[4]; /* Stored in network byte order */ 4508 __u32 local_ip6[4]; /* Stored in network byte order */ 4509 __u32 remote_port; /* Stored in network byte order */ 4510 __u32 local_port; /* stored in host byte order */ 4511 __u32 is_fullsock; /* Some TCP fields are only valid if 4512 * there is a full socket. If not, the 4513 * fields read as zero. 4514 */ 4515 __u32 snd_cwnd; 4516 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 4517 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 4518 __u32 state; 4519 __u32 rtt_min; 4520 __u32 snd_ssthresh; 4521 __u32 rcv_nxt; 4522 __u32 snd_nxt; 4523 __u32 snd_una; 4524 __u32 mss_cache; 4525 __u32 ecn_flags; 4526 __u32 rate_delivered; 4527 __u32 rate_interval_us; 4528 __u32 packets_out; 4529 __u32 retrans_out; 4530 __u32 total_retrans; 4531 __u32 segs_in; 4532 __u32 data_segs_in; 4533 __u32 segs_out; 4534 __u32 data_segs_out; 4535 __u32 lost_out; 4536 __u32 sacked_out; 4537 __u32 sk_txhash; 4538 __u64 bytes_received; 4539 __u64 bytes_acked; 4540 __bpf_md_ptr(struct bpf_sock *, sk); 4541 /* [skb_data, skb_data_end) covers the whole TCP header. 4542 * 4543 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 4544 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 4545 * header has not been written. 4546 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 4547 * been written so far. 4548 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 4549 * the 3WHS. 4550 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 4551 * the 3WHS. 4552 * 4553 * bpf_load_hdr_opt() can also be used to read a particular option. 4554 */ 4555 __bpf_md_ptr(void *, skb_data); 4556 __bpf_md_ptr(void *, skb_data_end); 4557 __u32 skb_len; /* The total length of a packet. 4558 * It includes the header, options, 4559 * and payload. 4560 */ 4561 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 4562 * an easy way to check for tcp_flags 4563 * without parsing skb_data. 4564 * 4565 * In particular, the skb_tcp_flags 4566 * will still be available in 4567 * BPF_SOCK_OPS_HDR_OPT_LEN even though 4568 * the outgoing header has not 4569 * been written yet. 4570 */ 4571 }; 4572 4573 /* Definitions for bpf_sock_ops_cb_flags */ 4574 enum { 4575 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 4576 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 4577 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 4578 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 4579 /* Call bpf for all received TCP headers. The bpf prog will be 4580 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4581 * 4582 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4583 * for the header option related helpers that will be useful 4584 * to the bpf programs. 4585 * 4586 * It could be used at the client/active side (i.e. connect() side) 4587 * when the server told it that the server was in syncookie 4588 * mode and required the active side to resend the bpf-written 4589 * options. The active side can keep writing the bpf-options until 4590 * it received a valid packet from the server side to confirm 4591 * the earlier packet (and options) has been received. The later 4592 * example patch is using it like this at the active side when the 4593 * server is in syncookie mode. 4594 * 4595 * The bpf prog will usually turn this off in the common cases. 4596 */ 4597 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 4598 /* Call bpf when kernel has received a header option that 4599 * the kernel cannot handle. The bpf prog will be called under 4600 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 4601 * 4602 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4603 * for the header option related helpers that will be useful 4604 * to the bpf programs. 4605 */ 4606 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 4607 /* Call bpf when the kernel is writing header options for the 4608 * outgoing packet. The bpf prog will first be called 4609 * to reserve space in a skb under 4610 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 4611 * the bpf prog will be called to write the header option(s) 4612 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4613 * 4614 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 4615 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 4616 * related helpers that will be useful to the bpf programs. 4617 * 4618 * The kernel gets its chance to reserve space and write 4619 * options first before the BPF program does. 4620 */ 4621 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 4622 /* Mask of all currently supported cb flags */ 4623 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 4624 }; 4625 4626 /* List of known BPF sock_ops operators. 4627 * New entries can only be added at the end 4628 */ 4629 enum { 4630 BPF_SOCK_OPS_VOID, 4631 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 4632 * -1 if default value should be used 4633 */ 4634 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 4635 * window (in packets) or -1 if default 4636 * value should be used 4637 */ 4638 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 4639 * active connection is initialized 4640 */ 4641 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 4642 * active connection is 4643 * established 4644 */ 4645 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 4646 * passive connection is 4647 * established 4648 */ 4649 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 4650 * needs ECN 4651 */ 4652 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 4653 * based on the path and may be 4654 * dependent on the congestion control 4655 * algorithm. In general it indicates 4656 * a congestion threshold. RTTs above 4657 * this indicate congestion 4658 */ 4659 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 4660 * Arg1: value of icsk_retransmits 4661 * Arg2: value of icsk_rto 4662 * Arg3: whether RTO has expired 4663 */ 4664 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 4665 * Arg1: sequence number of 1st byte 4666 * Arg2: # segments 4667 * Arg3: return value of 4668 * tcp_transmit_skb (0 => success) 4669 */ 4670 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 4671 * Arg1: old_state 4672 * Arg2: new_state 4673 */ 4674 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 4675 * socket transition to LISTEN state. 4676 */ 4677 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 4678 */ 4679 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 4680 * It will be called to handle 4681 * the packets received at 4682 * an already established 4683 * connection. 4684 * 4685 * sock_ops->skb_data: 4686 * Referring to the received skb. 4687 * It covers the TCP header only. 4688 * 4689 * bpf_load_hdr_opt() can also 4690 * be used to search for a 4691 * particular option. 4692 */ 4693 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 4694 * header option later in 4695 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4696 * Arg1: bool want_cookie. (in 4697 * writing SYNACK only) 4698 * 4699 * sock_ops->skb_data: 4700 * Not available because no header has 4701 * been written yet. 4702 * 4703 * sock_ops->skb_tcp_flags: 4704 * The tcp_flags of the 4705 * outgoing skb. (e.g. SYN, ACK, FIN). 4706 * 4707 * bpf_reserve_hdr_opt() should 4708 * be used to reserve space. 4709 */ 4710 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 4711 * Arg1: bool want_cookie. (in 4712 * writing SYNACK only) 4713 * 4714 * sock_ops->skb_data: 4715 * Referring to the outgoing skb. 4716 * It covers the TCP header 4717 * that has already been written 4718 * by the kernel and the 4719 * earlier bpf-progs. 4720 * 4721 * sock_ops->skb_tcp_flags: 4722 * The tcp_flags of the outgoing 4723 * skb. (e.g. SYN, ACK, FIN). 4724 * 4725 * bpf_store_hdr_opt() should 4726 * be used to write the 4727 * option. 4728 * 4729 * bpf_load_hdr_opt() can also 4730 * be used to search for a 4731 * particular option that 4732 * has already been written 4733 * by the kernel or the 4734 * earlier bpf-progs. 4735 */ 4736 }; 4737 4738 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 4739 * changes between the TCP and BPF versions. Ideally this should never happen. 4740 * If it does, we need to add code to convert them before calling 4741 * the BPF sock_ops function. 4742 */ 4743 enum { 4744 BPF_TCP_ESTABLISHED = 1, 4745 BPF_TCP_SYN_SENT, 4746 BPF_TCP_SYN_RECV, 4747 BPF_TCP_FIN_WAIT1, 4748 BPF_TCP_FIN_WAIT2, 4749 BPF_TCP_TIME_WAIT, 4750 BPF_TCP_CLOSE, 4751 BPF_TCP_CLOSE_WAIT, 4752 BPF_TCP_LAST_ACK, 4753 BPF_TCP_LISTEN, 4754 BPF_TCP_CLOSING, /* Now a valid state */ 4755 BPF_TCP_NEW_SYN_RECV, 4756 4757 BPF_TCP_MAX_STATES /* Leave at the end! */ 4758 }; 4759 4760 enum { 4761 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 4762 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 4763 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 4764 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 4765 /* Copy the SYN pkt to optval 4766 * 4767 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 4768 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 4769 * to only getting from the saved_syn. It can either get the 4770 * syn packet from: 4771 * 4772 * 1. the just-received SYN packet (only available when writing the 4773 * SYNACK). It will be useful when it is not necessary to 4774 * save the SYN packet for latter use. It is also the only way 4775 * to get the SYN during syncookie mode because the syn 4776 * packet cannot be saved during syncookie. 4777 * 4778 * OR 4779 * 4780 * 2. the earlier saved syn which was done by 4781 * bpf_setsockopt(TCP_SAVE_SYN). 4782 * 4783 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 4784 * SYN packet is obtained. 4785 * 4786 * If the bpf-prog does not need the IP[46] header, the 4787 * bpf-prog can avoid parsing the IP header by using 4788 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 4789 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 4790 * 4791 * >0: Total number of bytes copied 4792 * -ENOSPC: Not enough space in optval. Only optlen number of 4793 * bytes is copied. 4794 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 4795 * is not saved by setsockopt(TCP_SAVE_SYN). 4796 */ 4797 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 4798 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 4799 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 4800 }; 4801 4802 enum { 4803 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 4804 }; 4805 4806 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 4807 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4808 */ 4809 enum { 4810 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 4811 * total option spaces 4812 * required for an established 4813 * sk in order to calculate the 4814 * MSS. No skb is actually 4815 * sent. 4816 */ 4817 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 4818 * when sending a SYN. 4819 */ 4820 }; 4821 4822 struct bpf_perf_event_value { 4823 __u64 counter; 4824 __u64 enabled; 4825 __u64 running; 4826 }; 4827 4828 enum { 4829 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 4830 BPF_DEVCG_ACC_READ = (1ULL << 1), 4831 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 4832 }; 4833 4834 enum { 4835 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 4836 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 4837 }; 4838 4839 struct bpf_cgroup_dev_ctx { 4840 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 4841 __u32 access_type; 4842 __u32 major; 4843 __u32 minor; 4844 }; 4845 4846 struct bpf_raw_tracepoint_args { 4847 __u64 args[0]; 4848 }; 4849 4850 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 4851 * OUTPUT: Do lookup from egress perspective; default is ingress 4852 */ 4853 enum { 4854 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 4855 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 4856 }; 4857 4858 enum { 4859 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 4860 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 4861 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 4862 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 4863 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 4864 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 4865 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 4866 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 4867 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 4868 }; 4869 4870 struct bpf_fib_lookup { 4871 /* input: network family for lookup (AF_INET, AF_INET6) 4872 * output: network family of egress nexthop 4873 */ 4874 __u8 family; 4875 4876 /* set if lookup is to consider L4 data - e.g., FIB rules */ 4877 __u8 l4_protocol; 4878 __be16 sport; 4879 __be16 dport; 4880 4881 /* total length of packet from network header - used for MTU check */ 4882 __u16 tot_len; 4883 4884 /* input: L3 device index for lookup 4885 * output: device index from FIB lookup 4886 */ 4887 __u32 ifindex; 4888 4889 union { 4890 /* inputs to lookup */ 4891 __u8 tos; /* AF_INET */ 4892 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 4893 4894 /* output: metric of fib result (IPv4/IPv6 only) */ 4895 __u32 rt_metric; 4896 }; 4897 4898 union { 4899 __be32 ipv4_src; 4900 __u32 ipv6_src[4]; /* in6_addr; network order */ 4901 }; 4902 4903 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 4904 * network header. output: bpf_fib_lookup sets to gateway address 4905 * if FIB lookup returns gateway route 4906 */ 4907 union { 4908 __be32 ipv4_dst; 4909 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4910 }; 4911 4912 /* output */ 4913 __be16 h_vlan_proto; 4914 __be16 h_vlan_TCI; 4915 __u8 smac[6]; /* ETH_ALEN */ 4916 __u8 dmac[6]; /* ETH_ALEN */ 4917 }; 4918 4919 struct bpf_redir_neigh { 4920 /* network family for lookup (AF_INET, AF_INET6) */ 4921 __u32 nh_family; 4922 /* network address of nexthop; skips fib lookup to find gateway */ 4923 union { 4924 __be32 ipv4_nh; 4925 __u32 ipv6_nh[4]; /* in6_addr; network order */ 4926 }; 4927 }; 4928 4929 enum bpf_task_fd_type { 4930 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 4931 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 4932 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 4933 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 4934 BPF_FD_TYPE_UPROBE, /* filename + offset */ 4935 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 4936 }; 4937 4938 enum { 4939 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 4940 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 4941 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 4942 }; 4943 4944 struct bpf_flow_keys { 4945 __u16 nhoff; 4946 __u16 thoff; 4947 __u16 addr_proto; /* ETH_P_* of valid addrs */ 4948 __u8 is_frag; 4949 __u8 is_first_frag; 4950 __u8 is_encap; 4951 __u8 ip_proto; 4952 __be16 n_proto; 4953 __be16 sport; 4954 __be16 dport; 4955 union { 4956 struct { 4957 __be32 ipv4_src; 4958 __be32 ipv4_dst; 4959 }; 4960 struct { 4961 __u32 ipv6_src[4]; /* in6_addr; network order */ 4962 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4963 }; 4964 }; 4965 __u32 flags; 4966 __be32 flow_label; 4967 }; 4968 4969 struct bpf_func_info { 4970 __u32 insn_off; 4971 __u32 type_id; 4972 }; 4973 4974 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 4975 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 4976 4977 struct bpf_line_info { 4978 __u32 insn_off; 4979 __u32 file_name_off; 4980 __u32 line_off; 4981 __u32 line_col; 4982 }; 4983 4984 struct bpf_spin_lock { 4985 __u32 val; 4986 }; 4987 4988 struct bpf_sysctl { 4989 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 4990 * Allows 1,2,4-byte read, but no write. 4991 */ 4992 __u32 file_pos; /* Sysctl file position to read from, write to. 4993 * Allows 1,2,4-byte read an 4-byte write. 4994 */ 4995 }; 4996 4997 struct bpf_sockopt { 4998 __bpf_md_ptr(struct bpf_sock *, sk); 4999 __bpf_md_ptr(void *, optval); 5000 __bpf_md_ptr(void *, optval_end); 5001 5002 __s32 level; 5003 __s32 optname; 5004 __s32 optlen; 5005 __s32 retval; 5006 }; 5007 5008 struct bpf_pidns_info { 5009 __u32 pid; 5010 __u32 tgid; 5011 }; 5012 5013 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 5014 struct bpf_sk_lookup { 5015 union { 5016 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 5017 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 5018 }; 5019 5020 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 5021 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 5022 __u32 remote_ip4; /* Network byte order */ 5023 __u32 remote_ip6[4]; /* Network byte order */ 5024 __u32 remote_port; /* Network byte order */ 5025 __u32 local_ip4; /* Network byte order */ 5026 __u32 local_ip6[4]; /* Network byte order */ 5027 __u32 local_port; /* Host byte order */ 5028 }; 5029 5030 /* 5031 * struct btf_ptr is used for typed pointer representation; the 5032 * type id is used to render the pointer data as the appropriate type 5033 * via the bpf_snprintf_btf() helper described above. A flags field - 5034 * potentially to specify additional details about the BTF pointer 5035 * (rather than its mode of display) - is included for future use. 5036 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 5037 */ 5038 struct btf_ptr { 5039 void *ptr; 5040 __u32 type_id; 5041 __u32 flags; /* BTF ptr flags; unused at present. */ 5042 }; 5043 5044 /* 5045 * Flags to control bpf_snprintf_btf() behaviour. 5046 * - BTF_F_COMPACT: no formatting around type information 5047 * - BTF_F_NONAME: no struct/union member names/types 5048 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 5049 * equivalent to %px. 5050 * - BTF_F_ZERO: show zero-valued struct/union members; they 5051 * are not displayed by default 5052 */ 5053 enum { 5054 BTF_F_COMPACT = (1ULL << 0), 5055 BTF_F_NONAME = (1ULL << 1), 5056 BTF_F_PTR_RAW = (1ULL << 2), 5057 BTF_F_ZERO = (1ULL << 3), 5058 }; 5059 5060 #endif /* _UAPI__LINUX_BPF_H__ */ 5061