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