1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 union bpf_iter_link_info { 85 struct { 86 __u32 map_fd; 87 } map; 88 }; 89 90 /* BPF syscall commands, see bpf(2) man-page for details. */ 91 enum bpf_cmd { 92 BPF_MAP_CREATE, 93 BPF_MAP_LOOKUP_ELEM, 94 BPF_MAP_UPDATE_ELEM, 95 BPF_MAP_DELETE_ELEM, 96 BPF_MAP_GET_NEXT_KEY, 97 BPF_PROG_LOAD, 98 BPF_OBJ_PIN, 99 BPF_OBJ_GET, 100 BPF_PROG_ATTACH, 101 BPF_PROG_DETACH, 102 BPF_PROG_TEST_RUN, 103 BPF_PROG_GET_NEXT_ID, 104 BPF_MAP_GET_NEXT_ID, 105 BPF_PROG_GET_FD_BY_ID, 106 BPF_MAP_GET_FD_BY_ID, 107 BPF_OBJ_GET_INFO_BY_FD, 108 BPF_PROG_QUERY, 109 BPF_RAW_TRACEPOINT_OPEN, 110 BPF_BTF_LOAD, 111 BPF_BTF_GET_FD_BY_ID, 112 BPF_TASK_FD_QUERY, 113 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 114 BPF_MAP_FREEZE, 115 BPF_BTF_GET_NEXT_ID, 116 BPF_MAP_LOOKUP_BATCH, 117 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 118 BPF_MAP_UPDATE_BATCH, 119 BPF_MAP_DELETE_BATCH, 120 BPF_LINK_CREATE, 121 BPF_LINK_UPDATE, 122 BPF_LINK_GET_FD_BY_ID, 123 BPF_LINK_GET_NEXT_ID, 124 BPF_ENABLE_STATS, 125 BPF_ITER_CREATE, 126 BPF_LINK_DETACH, 127 BPF_PROG_BIND_MAP, 128 }; 129 130 enum bpf_map_type { 131 BPF_MAP_TYPE_UNSPEC, 132 BPF_MAP_TYPE_HASH, 133 BPF_MAP_TYPE_ARRAY, 134 BPF_MAP_TYPE_PROG_ARRAY, 135 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 136 BPF_MAP_TYPE_PERCPU_HASH, 137 BPF_MAP_TYPE_PERCPU_ARRAY, 138 BPF_MAP_TYPE_STACK_TRACE, 139 BPF_MAP_TYPE_CGROUP_ARRAY, 140 BPF_MAP_TYPE_LRU_HASH, 141 BPF_MAP_TYPE_LRU_PERCPU_HASH, 142 BPF_MAP_TYPE_LPM_TRIE, 143 BPF_MAP_TYPE_ARRAY_OF_MAPS, 144 BPF_MAP_TYPE_HASH_OF_MAPS, 145 BPF_MAP_TYPE_DEVMAP, 146 BPF_MAP_TYPE_SOCKMAP, 147 BPF_MAP_TYPE_CPUMAP, 148 BPF_MAP_TYPE_XSKMAP, 149 BPF_MAP_TYPE_SOCKHASH, 150 BPF_MAP_TYPE_CGROUP_STORAGE, 151 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 152 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 153 BPF_MAP_TYPE_QUEUE, 154 BPF_MAP_TYPE_STACK, 155 BPF_MAP_TYPE_SK_STORAGE, 156 BPF_MAP_TYPE_DEVMAP_HASH, 157 BPF_MAP_TYPE_STRUCT_OPS, 158 BPF_MAP_TYPE_RINGBUF, 159 BPF_MAP_TYPE_INODE_STORAGE, 160 }; 161 162 /* Note that tracing related programs such as 163 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 164 * are not subject to a stable API since kernel internal data 165 * structures can change from release to release and may 166 * therefore break existing tracing BPF programs. Tracing BPF 167 * programs correspond to /a/ specific kernel which is to be 168 * analyzed, and not /a/ specific kernel /and/ all future ones. 169 */ 170 enum bpf_prog_type { 171 BPF_PROG_TYPE_UNSPEC, 172 BPF_PROG_TYPE_SOCKET_FILTER, 173 BPF_PROG_TYPE_KPROBE, 174 BPF_PROG_TYPE_SCHED_CLS, 175 BPF_PROG_TYPE_SCHED_ACT, 176 BPF_PROG_TYPE_TRACEPOINT, 177 BPF_PROG_TYPE_XDP, 178 BPF_PROG_TYPE_PERF_EVENT, 179 BPF_PROG_TYPE_CGROUP_SKB, 180 BPF_PROG_TYPE_CGROUP_SOCK, 181 BPF_PROG_TYPE_LWT_IN, 182 BPF_PROG_TYPE_LWT_OUT, 183 BPF_PROG_TYPE_LWT_XMIT, 184 BPF_PROG_TYPE_SOCK_OPS, 185 BPF_PROG_TYPE_SK_SKB, 186 BPF_PROG_TYPE_CGROUP_DEVICE, 187 BPF_PROG_TYPE_SK_MSG, 188 BPF_PROG_TYPE_RAW_TRACEPOINT, 189 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 190 BPF_PROG_TYPE_LWT_SEG6LOCAL, 191 BPF_PROG_TYPE_LIRC_MODE2, 192 BPF_PROG_TYPE_SK_REUSEPORT, 193 BPF_PROG_TYPE_FLOW_DISSECTOR, 194 BPF_PROG_TYPE_CGROUP_SYSCTL, 195 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 196 BPF_PROG_TYPE_CGROUP_SOCKOPT, 197 BPF_PROG_TYPE_TRACING, 198 BPF_PROG_TYPE_STRUCT_OPS, 199 BPF_PROG_TYPE_EXT, 200 BPF_PROG_TYPE_LSM, 201 BPF_PROG_TYPE_SK_LOOKUP, 202 }; 203 204 enum bpf_attach_type { 205 BPF_CGROUP_INET_INGRESS, 206 BPF_CGROUP_INET_EGRESS, 207 BPF_CGROUP_INET_SOCK_CREATE, 208 BPF_CGROUP_SOCK_OPS, 209 BPF_SK_SKB_STREAM_PARSER, 210 BPF_SK_SKB_STREAM_VERDICT, 211 BPF_CGROUP_DEVICE, 212 BPF_SK_MSG_VERDICT, 213 BPF_CGROUP_INET4_BIND, 214 BPF_CGROUP_INET6_BIND, 215 BPF_CGROUP_INET4_CONNECT, 216 BPF_CGROUP_INET6_CONNECT, 217 BPF_CGROUP_INET4_POST_BIND, 218 BPF_CGROUP_INET6_POST_BIND, 219 BPF_CGROUP_UDP4_SENDMSG, 220 BPF_CGROUP_UDP6_SENDMSG, 221 BPF_LIRC_MODE2, 222 BPF_FLOW_DISSECTOR, 223 BPF_CGROUP_SYSCTL, 224 BPF_CGROUP_UDP4_RECVMSG, 225 BPF_CGROUP_UDP6_RECVMSG, 226 BPF_CGROUP_GETSOCKOPT, 227 BPF_CGROUP_SETSOCKOPT, 228 BPF_TRACE_RAW_TP, 229 BPF_TRACE_FENTRY, 230 BPF_TRACE_FEXIT, 231 BPF_MODIFY_RETURN, 232 BPF_LSM_MAC, 233 BPF_TRACE_ITER, 234 BPF_CGROUP_INET4_GETPEERNAME, 235 BPF_CGROUP_INET6_GETPEERNAME, 236 BPF_CGROUP_INET4_GETSOCKNAME, 237 BPF_CGROUP_INET6_GETSOCKNAME, 238 BPF_XDP_DEVMAP, 239 BPF_CGROUP_INET_SOCK_RELEASE, 240 BPF_XDP_CPUMAP, 241 BPF_SK_LOOKUP, 242 BPF_XDP, 243 __MAX_BPF_ATTACH_TYPE 244 }; 245 246 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 247 248 enum bpf_link_type { 249 BPF_LINK_TYPE_UNSPEC = 0, 250 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 251 BPF_LINK_TYPE_TRACING = 2, 252 BPF_LINK_TYPE_CGROUP = 3, 253 BPF_LINK_TYPE_ITER = 4, 254 BPF_LINK_TYPE_NETNS = 5, 255 BPF_LINK_TYPE_XDP = 6, 256 257 MAX_BPF_LINK_TYPE, 258 }; 259 260 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 261 * 262 * NONE(default): No further bpf programs allowed in the subtree. 263 * 264 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 265 * the program in this cgroup yields to sub-cgroup program. 266 * 267 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 268 * that cgroup program gets run in addition to the program in this cgroup. 269 * 270 * Only one program is allowed to be attached to a cgroup with 271 * NONE or BPF_F_ALLOW_OVERRIDE flag. 272 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 273 * release old program and attach the new one. Attach flags has to match. 274 * 275 * Multiple programs are allowed to be attached to a cgroup with 276 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 277 * (those that were attached first, run first) 278 * The programs of sub-cgroup are executed first, then programs of 279 * this cgroup and then programs of parent cgroup. 280 * When children program makes decision (like picking TCP CA or sock bind) 281 * parent program has a chance to override it. 282 * 283 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 284 * programs for a cgroup. Though it's possible to replace an old program at 285 * any position by also specifying BPF_F_REPLACE flag and position itself in 286 * replace_bpf_fd attribute. Old program at this position will be released. 287 * 288 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 289 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 290 * Ex1: 291 * cgrp1 (MULTI progs A, B) -> 292 * cgrp2 (OVERRIDE prog C) -> 293 * cgrp3 (MULTI prog D) -> 294 * cgrp4 (OVERRIDE prog E) -> 295 * cgrp5 (NONE prog F) 296 * the event in cgrp5 triggers execution of F,D,A,B in that order. 297 * if prog F is detached, the execution is E,D,A,B 298 * if prog F and D are detached, the execution is E,A,B 299 * if prog F, E and D are detached, the execution is C,A,B 300 * 301 * All eligible programs are executed regardless of return code from 302 * earlier programs. 303 */ 304 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 305 #define BPF_F_ALLOW_MULTI (1U << 1) 306 #define BPF_F_REPLACE (1U << 2) 307 308 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 309 * verifier will perform strict alignment checking as if the kernel 310 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 311 * and NET_IP_ALIGN defined to 2. 312 */ 313 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 314 315 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 316 * verifier will allow any alignment whatsoever. On platforms 317 * with strict alignment requirements for loads ands stores (such 318 * as sparc and mips) the verifier validates that all loads and 319 * stores provably follow this requirement. This flag turns that 320 * checking and enforcement off. 321 * 322 * It is mostly used for testing when we want to validate the 323 * context and memory access aspects of the verifier, but because 324 * of an unaligned access the alignment check would trigger before 325 * the one we are interested in. 326 */ 327 #define BPF_F_ANY_ALIGNMENT (1U << 1) 328 329 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 330 * Verifier does sub-register def/use analysis and identifies instructions whose 331 * def only matters for low 32-bit, high 32-bit is never referenced later 332 * through implicit zero extension. Therefore verifier notifies JIT back-ends 333 * that it is safe to ignore clearing high 32-bit for these instructions. This 334 * saves some back-ends a lot of code-gen. However such optimization is not 335 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 336 * hence hasn't used verifier's analysis result. But, we really want to have a 337 * way to be able to verify the correctness of the described optimization on 338 * x86_64 on which testsuites are frequently exercised. 339 * 340 * So, this flag is introduced. Once it is set, verifier will randomize high 341 * 32-bit for those instructions who has been identified as safe to ignore them. 342 * Then, if verifier is not doing correct analysis, such randomization will 343 * regress tests to expose bugs. 344 */ 345 #define BPF_F_TEST_RND_HI32 (1U << 2) 346 347 /* The verifier internal test flag. Behavior is undefined */ 348 #define BPF_F_TEST_STATE_FREQ (1U << 3) 349 350 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 351 * restrict map and helper usage for such programs. Sleepable BPF programs can 352 * only be attached to hooks where kernel execution context allows sleeping. 353 * Such programs are allowed to use helpers that may sleep like 354 * bpf_copy_from_user(). 355 */ 356 #define BPF_F_SLEEPABLE (1U << 4) 357 358 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 359 * two extensions: 360 * 361 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 362 * insn[0].imm: map fd map fd 363 * insn[1].imm: 0 offset into value 364 * insn[0].off: 0 0 365 * insn[1].off: 0 0 366 * ldimm64 rewrite: address of map address of map[0]+offset 367 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 368 */ 369 #define BPF_PSEUDO_MAP_FD 1 370 #define BPF_PSEUDO_MAP_VALUE 2 371 372 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 373 * offset to another bpf function 374 */ 375 #define BPF_PSEUDO_CALL 1 376 377 /* flags for BPF_MAP_UPDATE_ELEM command */ 378 enum { 379 BPF_ANY = 0, /* create new element or update existing */ 380 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 381 BPF_EXIST = 2, /* update existing element */ 382 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 383 }; 384 385 /* flags for BPF_MAP_CREATE command */ 386 enum { 387 BPF_F_NO_PREALLOC = (1U << 0), 388 /* Instead of having one common LRU list in the 389 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 390 * which can scale and perform better. 391 * Note, the LRU nodes (including free nodes) cannot be moved 392 * across different LRU lists. 393 */ 394 BPF_F_NO_COMMON_LRU = (1U << 1), 395 /* Specify numa node during map creation */ 396 BPF_F_NUMA_NODE = (1U << 2), 397 398 /* Flags for accessing BPF object from syscall side. */ 399 BPF_F_RDONLY = (1U << 3), 400 BPF_F_WRONLY = (1U << 4), 401 402 /* Flag for stack_map, store build_id+offset instead of pointer */ 403 BPF_F_STACK_BUILD_ID = (1U << 5), 404 405 /* Zero-initialize hash function seed. This should only be used for testing. */ 406 BPF_F_ZERO_SEED = (1U << 6), 407 408 /* Flags for accessing BPF object from program side. */ 409 BPF_F_RDONLY_PROG = (1U << 7), 410 BPF_F_WRONLY_PROG = (1U << 8), 411 412 /* Clone map from listener for newly accepted socket */ 413 BPF_F_CLONE = (1U << 9), 414 415 /* Enable memory-mapping BPF map */ 416 BPF_F_MMAPABLE = (1U << 10), 417 418 /* Share perf_event among processes */ 419 BPF_F_PRESERVE_ELEMS = (1U << 11), 420 }; 421 422 /* Flags for BPF_PROG_QUERY. */ 423 424 /* Query effective (directly attached + inherited from ancestor cgroups) 425 * programs that will be executed for events within a cgroup. 426 * attach_flags with this flag are returned only for directly attached programs. 427 */ 428 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 429 430 /* Flags for BPF_PROG_TEST_RUN */ 431 432 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 433 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 434 435 /* type for BPF_ENABLE_STATS */ 436 enum bpf_stats_type { 437 /* enabled run_time_ns and run_cnt */ 438 BPF_STATS_RUN_TIME = 0, 439 }; 440 441 enum bpf_stack_build_id_status { 442 /* user space need an empty entry to identify end of a trace */ 443 BPF_STACK_BUILD_ID_EMPTY = 0, 444 /* with valid build_id and offset */ 445 BPF_STACK_BUILD_ID_VALID = 1, 446 /* couldn't get build_id, fallback to ip */ 447 BPF_STACK_BUILD_ID_IP = 2, 448 }; 449 450 #define BPF_BUILD_ID_SIZE 20 451 struct bpf_stack_build_id { 452 __s32 status; 453 unsigned char build_id[BPF_BUILD_ID_SIZE]; 454 union { 455 __u64 offset; 456 __u64 ip; 457 }; 458 }; 459 460 #define BPF_OBJ_NAME_LEN 16U 461 462 union bpf_attr { 463 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 464 __u32 map_type; /* one of enum bpf_map_type */ 465 __u32 key_size; /* size of key in bytes */ 466 __u32 value_size; /* size of value in bytes */ 467 __u32 max_entries; /* max number of entries in a map */ 468 __u32 map_flags; /* BPF_MAP_CREATE related 469 * flags defined above. 470 */ 471 __u32 inner_map_fd; /* fd pointing to the inner map */ 472 __u32 numa_node; /* numa node (effective only if 473 * BPF_F_NUMA_NODE is set). 474 */ 475 char map_name[BPF_OBJ_NAME_LEN]; 476 __u32 map_ifindex; /* ifindex of netdev to create on */ 477 __u32 btf_fd; /* fd pointing to a BTF type data */ 478 __u32 btf_key_type_id; /* BTF type_id of the key */ 479 __u32 btf_value_type_id; /* BTF type_id of the value */ 480 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 481 * struct stored as the 482 * map value 483 */ 484 }; 485 486 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 487 __u32 map_fd; 488 __aligned_u64 key; 489 union { 490 __aligned_u64 value; 491 __aligned_u64 next_key; 492 }; 493 __u64 flags; 494 }; 495 496 struct { /* struct used by BPF_MAP_*_BATCH commands */ 497 __aligned_u64 in_batch; /* start batch, 498 * NULL to start from beginning 499 */ 500 __aligned_u64 out_batch; /* output: next start batch */ 501 __aligned_u64 keys; 502 __aligned_u64 values; 503 __u32 count; /* input/output: 504 * input: # of key/value 505 * elements 506 * output: # of filled elements 507 */ 508 __u32 map_fd; 509 __u64 elem_flags; 510 __u64 flags; 511 } batch; 512 513 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 514 __u32 prog_type; /* one of enum bpf_prog_type */ 515 __u32 insn_cnt; 516 __aligned_u64 insns; 517 __aligned_u64 license; 518 __u32 log_level; /* verbosity level of verifier */ 519 __u32 log_size; /* size of user buffer */ 520 __aligned_u64 log_buf; /* user supplied buffer */ 521 __u32 kern_version; /* not used */ 522 __u32 prog_flags; 523 char prog_name[BPF_OBJ_NAME_LEN]; 524 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 525 /* For some prog types expected attach type must be known at 526 * load time to verify attach type specific parts of prog 527 * (context accesses, allowed helpers, etc). 528 */ 529 __u32 expected_attach_type; 530 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 531 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 532 __aligned_u64 func_info; /* func info */ 533 __u32 func_info_cnt; /* number of bpf_func_info records */ 534 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 535 __aligned_u64 line_info; /* line info */ 536 __u32 line_info_cnt; /* number of bpf_line_info records */ 537 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 538 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 539 }; 540 541 struct { /* anonymous struct used by BPF_OBJ_* commands */ 542 __aligned_u64 pathname; 543 __u32 bpf_fd; 544 __u32 file_flags; 545 }; 546 547 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 548 __u32 target_fd; /* container object to attach to */ 549 __u32 attach_bpf_fd; /* eBPF program to attach */ 550 __u32 attach_type; 551 __u32 attach_flags; 552 __u32 replace_bpf_fd; /* previously attached eBPF 553 * program to replace if 554 * BPF_F_REPLACE is used 555 */ 556 }; 557 558 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 559 __u32 prog_fd; 560 __u32 retval; 561 __u32 data_size_in; /* input: len of data_in */ 562 __u32 data_size_out; /* input/output: len of data_out 563 * returns ENOSPC if data_out 564 * is too small. 565 */ 566 __aligned_u64 data_in; 567 __aligned_u64 data_out; 568 __u32 repeat; 569 __u32 duration; 570 __u32 ctx_size_in; /* input: len of ctx_in */ 571 __u32 ctx_size_out; /* input/output: len of ctx_out 572 * returns ENOSPC if ctx_out 573 * is too small. 574 */ 575 __aligned_u64 ctx_in; 576 __aligned_u64 ctx_out; 577 __u32 flags; 578 __u32 cpu; 579 } test; 580 581 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 582 union { 583 __u32 start_id; 584 __u32 prog_id; 585 __u32 map_id; 586 __u32 btf_id; 587 __u32 link_id; 588 }; 589 __u32 next_id; 590 __u32 open_flags; 591 }; 592 593 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 594 __u32 bpf_fd; 595 __u32 info_len; 596 __aligned_u64 info; 597 } info; 598 599 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 600 __u32 target_fd; /* container object to query */ 601 __u32 attach_type; 602 __u32 query_flags; 603 __u32 attach_flags; 604 __aligned_u64 prog_ids; 605 __u32 prog_cnt; 606 } query; 607 608 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 609 __u64 name; 610 __u32 prog_fd; 611 } raw_tracepoint; 612 613 struct { /* anonymous struct for BPF_BTF_LOAD */ 614 __aligned_u64 btf; 615 __aligned_u64 btf_log_buf; 616 __u32 btf_size; 617 __u32 btf_log_size; 618 __u32 btf_log_level; 619 }; 620 621 struct { 622 __u32 pid; /* input: pid */ 623 __u32 fd; /* input: fd */ 624 __u32 flags; /* input: flags */ 625 __u32 buf_len; /* input/output: buf len */ 626 __aligned_u64 buf; /* input/output: 627 * tp_name for tracepoint 628 * symbol for kprobe 629 * filename for uprobe 630 */ 631 __u32 prog_id; /* output: prod_id */ 632 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 633 __u64 probe_offset; /* output: probe_offset */ 634 __u64 probe_addr; /* output: probe_addr */ 635 } task_fd_query; 636 637 struct { /* struct used by BPF_LINK_CREATE command */ 638 __u32 prog_fd; /* eBPF program to attach */ 639 union { 640 __u32 target_fd; /* object to attach to */ 641 __u32 target_ifindex; /* target ifindex */ 642 }; 643 __u32 attach_type; /* attach type */ 644 __u32 flags; /* extra flags */ 645 union { 646 __u32 target_btf_id; /* btf_id of target to attach to */ 647 struct { 648 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 649 __u32 iter_info_len; /* iter_info length */ 650 }; 651 }; 652 } link_create; 653 654 struct { /* struct used by BPF_LINK_UPDATE command */ 655 __u32 link_fd; /* link fd */ 656 /* new program fd to update link with */ 657 __u32 new_prog_fd; 658 __u32 flags; /* extra flags */ 659 /* expected link's program fd; is specified only if 660 * BPF_F_REPLACE flag is set in flags */ 661 __u32 old_prog_fd; 662 } link_update; 663 664 struct { 665 __u32 link_fd; 666 } link_detach; 667 668 struct { /* struct used by BPF_ENABLE_STATS command */ 669 __u32 type; 670 } enable_stats; 671 672 struct { /* struct used by BPF_ITER_CREATE command */ 673 __u32 link_fd; 674 __u32 flags; 675 } iter_create; 676 677 struct { /* struct used by BPF_PROG_BIND_MAP command */ 678 __u32 prog_fd; 679 __u32 map_fd; 680 __u32 flags; /* extra flags */ 681 } prog_bind_map; 682 683 } __attribute__((aligned(8))); 684 685 /* The description below is an attempt at providing documentation to eBPF 686 * developers about the multiple available eBPF helper functions. It can be 687 * parsed and used to produce a manual page. The workflow is the following, 688 * and requires the rst2man utility: 689 * 690 * $ ./scripts/bpf_helpers_doc.py \ 691 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 692 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 693 * $ man /tmp/bpf-helpers.7 694 * 695 * Note that in order to produce this external documentation, some RST 696 * formatting is used in the descriptions to get "bold" and "italics" in 697 * manual pages. Also note that the few trailing white spaces are 698 * intentional, removing them would break paragraphs for rst2man. 699 * 700 * Start of BPF helper function descriptions: 701 * 702 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 703 * Description 704 * Perform a lookup in *map* for an entry associated to *key*. 705 * Return 706 * Map value associated to *key*, or **NULL** if no entry was 707 * found. 708 * 709 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 710 * Description 711 * Add or update the value of the entry associated to *key* in 712 * *map* with *value*. *flags* is one of: 713 * 714 * **BPF_NOEXIST** 715 * The entry for *key* must not exist in the map. 716 * **BPF_EXIST** 717 * The entry for *key* must already exist in the map. 718 * **BPF_ANY** 719 * No condition on the existence of the entry for *key*. 720 * 721 * Flag value **BPF_NOEXIST** cannot be used for maps of types 722 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 723 * elements always exist), the helper would return an error. 724 * Return 725 * 0 on success, or a negative error in case of failure. 726 * 727 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 728 * Description 729 * Delete entry with *key* from *map*. 730 * Return 731 * 0 on success, or a negative error in case of failure. 732 * 733 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 734 * Description 735 * For tracing programs, safely attempt to read *size* bytes from 736 * kernel space address *unsafe_ptr* and store the data in *dst*. 737 * 738 * Generally, use **bpf_probe_read_user**\ () or 739 * **bpf_probe_read_kernel**\ () instead. 740 * Return 741 * 0 on success, or a negative error in case of failure. 742 * 743 * u64 bpf_ktime_get_ns(void) 744 * Description 745 * Return the time elapsed since system boot, in nanoseconds. 746 * Does not include time the system was suspended. 747 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 748 * Return 749 * Current *ktime*. 750 * 751 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 752 * Description 753 * This helper is a "printk()-like" facility for debugging. It 754 * prints a message defined by format *fmt* (of size *fmt_size*) 755 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 756 * available. It can take up to three additional **u64** 757 * arguments (as an eBPF helpers, the total number of arguments is 758 * limited to five). 759 * 760 * Each time the helper is called, it appends a line to the trace. 761 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 762 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 763 * The format of the trace is customizable, and the exact output 764 * one will get depends on the options set in 765 * *\/sys/kernel/debug/tracing/trace_options* (see also the 766 * *README* file under the same directory). However, it usually 767 * defaults to something like: 768 * 769 * :: 770 * 771 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 772 * 773 * In the above: 774 * 775 * * ``telnet`` is the name of the current task. 776 * * ``470`` is the PID of the current task. 777 * * ``001`` is the CPU number on which the task is 778 * running. 779 * * In ``.N..``, each character refers to a set of 780 * options (whether irqs are enabled, scheduling 781 * options, whether hard/softirqs are running, level of 782 * preempt_disabled respectively). **N** means that 783 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 784 * are set. 785 * * ``419421.045894`` is a timestamp. 786 * * ``0x00000001`` is a fake value used by BPF for the 787 * instruction pointer register. 788 * * ``<formatted msg>`` is the message formatted with 789 * *fmt*. 790 * 791 * The conversion specifiers supported by *fmt* are similar, but 792 * more limited than for printk(). They are **%d**, **%i**, 793 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 794 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 795 * of field, padding with zeroes, etc.) is available, and the 796 * helper will return **-EINVAL** (but print nothing) if it 797 * encounters an unknown specifier. 798 * 799 * Also, note that **bpf_trace_printk**\ () is slow, and should 800 * only be used for debugging purposes. For this reason, a notice 801 * block (spanning several lines) is printed to kernel logs and 802 * states that the helper should not be used "for production use" 803 * the first time this helper is used (or more precisely, when 804 * **trace_printk**\ () buffers are allocated). For passing values 805 * to user space, perf events should be preferred. 806 * Return 807 * The number of bytes written to the buffer, or a negative error 808 * in case of failure. 809 * 810 * u32 bpf_get_prandom_u32(void) 811 * Description 812 * Get a pseudo-random number. 813 * 814 * From a security point of view, this helper uses its own 815 * pseudo-random internal state, and cannot be used to infer the 816 * seed of other random functions in the kernel. However, it is 817 * essential to note that the generator used by the helper is not 818 * cryptographically secure. 819 * Return 820 * A random 32-bit unsigned value. 821 * 822 * u32 bpf_get_smp_processor_id(void) 823 * Description 824 * Get the SMP (symmetric multiprocessing) processor id. Note that 825 * all programs run with preemption disabled, which means that the 826 * SMP processor id is stable during all the execution of the 827 * program. 828 * Return 829 * The SMP id of the processor running the program. 830 * 831 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 832 * Description 833 * Store *len* bytes from address *from* into the packet 834 * associated to *skb*, at *offset*. *flags* are a combination of 835 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 836 * checksum for the packet after storing the bytes) and 837 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 838 * **->swhash** and *skb*\ **->l4hash** to 0). 839 * 840 * A call to this helper is susceptible to change the underlying 841 * packet buffer. Therefore, at load time, all checks on pointers 842 * previously done by the verifier are invalidated and must be 843 * performed again, if the helper is used in combination with 844 * direct packet access. 845 * Return 846 * 0 on success, or a negative error in case of failure. 847 * 848 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 849 * Description 850 * Recompute the layer 3 (e.g. IP) checksum for the packet 851 * associated to *skb*. Computation is incremental, so the helper 852 * must know the former value of the header field that was 853 * modified (*from*), the new value of this field (*to*), and the 854 * number of bytes (2 or 4) for this field, stored in *size*. 855 * Alternatively, it is possible to store the difference between 856 * the previous and the new values of the header field in *to*, by 857 * setting *from* and *size* to 0. For both methods, *offset* 858 * indicates the location of the IP checksum within the packet. 859 * 860 * This helper works in combination with **bpf_csum_diff**\ (), 861 * which does not update the checksum in-place, but offers more 862 * flexibility and can handle sizes larger than 2 or 4 for the 863 * checksum to update. 864 * 865 * A call to this helper is susceptible to change the underlying 866 * packet buffer. Therefore, at load time, all checks on pointers 867 * previously done by the verifier are invalidated and must be 868 * performed again, if the helper is used in combination with 869 * direct packet access. 870 * Return 871 * 0 on success, or a negative error in case of failure. 872 * 873 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 874 * Description 875 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 876 * packet associated to *skb*. Computation is incremental, so the 877 * helper must know the former value of the header field that was 878 * modified (*from*), the new value of this field (*to*), and the 879 * number of bytes (2 or 4) for this field, stored on the lowest 880 * four bits of *flags*. Alternatively, it is possible to store 881 * the difference between the previous and the new values of the 882 * header field in *to*, by setting *from* and the four lowest 883 * bits of *flags* to 0. For both methods, *offset* indicates the 884 * location of the IP checksum within the packet. In addition to 885 * the size of the field, *flags* can be added (bitwise OR) actual 886 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 887 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 888 * for updates resulting in a null checksum the value is set to 889 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 890 * the checksum is to be computed against a pseudo-header. 891 * 892 * This helper works in combination with **bpf_csum_diff**\ (), 893 * which does not update the checksum in-place, but offers more 894 * flexibility and can handle sizes larger than 2 or 4 for the 895 * checksum to update. 896 * 897 * A call to this helper is susceptible to change the underlying 898 * packet buffer. Therefore, at load time, all checks on pointers 899 * previously done by the verifier are invalidated and must be 900 * performed again, if the helper is used in combination with 901 * direct packet access. 902 * Return 903 * 0 on success, or a negative error in case of failure. 904 * 905 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 906 * Description 907 * This special helper is used to trigger a "tail call", or in 908 * other words, to jump into another eBPF program. The same stack 909 * frame is used (but values on stack and in registers for the 910 * caller are not accessible to the callee). This mechanism allows 911 * for program chaining, either for raising the maximum number of 912 * available eBPF instructions, or to execute given programs in 913 * conditional blocks. For security reasons, there is an upper 914 * limit to the number of successive tail calls that can be 915 * performed. 916 * 917 * Upon call of this helper, the program attempts to jump into a 918 * program referenced at index *index* in *prog_array_map*, a 919 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 920 * *ctx*, a pointer to the context. 921 * 922 * If the call succeeds, the kernel immediately runs the first 923 * instruction of the new program. This is not a function call, 924 * and it never returns to the previous program. If the call 925 * fails, then the helper has no effect, and the caller continues 926 * to run its subsequent instructions. A call can fail if the 927 * destination program for the jump does not exist (i.e. *index* 928 * is superior to the number of entries in *prog_array_map*), or 929 * if the maximum number of tail calls has been reached for this 930 * chain of programs. This limit is defined in the kernel by the 931 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 932 * which is currently set to 32. 933 * Return 934 * 0 on success, or a negative error in case of failure. 935 * 936 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 937 * Description 938 * Clone and redirect the packet associated to *skb* to another 939 * net device of index *ifindex*. Both ingress and egress 940 * interfaces can be used for redirection. The **BPF_F_INGRESS** 941 * value in *flags* is used to make the distinction (ingress path 942 * is selected if the flag is present, egress path otherwise). 943 * This is the only flag supported for now. 944 * 945 * In comparison with **bpf_redirect**\ () helper, 946 * **bpf_clone_redirect**\ () has the associated cost of 947 * duplicating the packet buffer, but this can be executed out of 948 * the eBPF program. Conversely, **bpf_redirect**\ () is more 949 * efficient, but it is handled through an action code where the 950 * redirection happens only after the eBPF program has returned. 951 * 952 * A call to this helper is susceptible to change the underlying 953 * packet buffer. Therefore, at load time, all checks on pointers 954 * previously done by the verifier are invalidated and must be 955 * performed again, if the helper is used in combination with 956 * direct packet access. 957 * Return 958 * 0 on success, or a negative error in case of failure. 959 * 960 * u64 bpf_get_current_pid_tgid(void) 961 * Return 962 * A 64-bit integer containing the current tgid and pid, and 963 * created as such: 964 * *current_task*\ **->tgid << 32 \|** 965 * *current_task*\ **->pid**. 966 * 967 * u64 bpf_get_current_uid_gid(void) 968 * Return 969 * A 64-bit integer containing the current GID and UID, and 970 * created as such: *current_gid* **<< 32 \|** *current_uid*. 971 * 972 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 973 * Description 974 * Copy the **comm** attribute of the current task into *buf* of 975 * *size_of_buf*. The **comm** attribute contains the name of 976 * the executable (excluding the path) for the current task. The 977 * *size_of_buf* must be strictly positive. On success, the 978 * helper makes sure that the *buf* is NUL-terminated. On failure, 979 * it is filled with zeroes. 980 * Return 981 * 0 on success, or a negative error in case of failure. 982 * 983 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 984 * Description 985 * Retrieve the classid for the current task, i.e. for the net_cls 986 * cgroup to which *skb* belongs. 987 * 988 * This helper can be used on TC egress path, but not on ingress. 989 * 990 * The net_cls cgroup provides an interface to tag network packets 991 * based on a user-provided identifier for all traffic coming from 992 * the tasks belonging to the related cgroup. See also the related 993 * kernel documentation, available from the Linux sources in file 994 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 995 * 996 * The Linux kernel has two versions for cgroups: there are 997 * cgroups v1 and cgroups v2. Both are available to users, who can 998 * use a mixture of them, but note that the net_cls cgroup is for 999 * cgroup v1 only. This makes it incompatible with BPF programs 1000 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1001 * only hold data for one version of cgroups at a time). 1002 * 1003 * This helper is only available is the kernel was compiled with 1004 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1005 * "**y**" or to "**m**". 1006 * Return 1007 * The classid, or 0 for the default unconfigured classid. 1008 * 1009 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1010 * Description 1011 * Push a *vlan_tci* (VLAN tag control information) of protocol 1012 * *vlan_proto* to the packet associated to *skb*, then update 1013 * the checksum. Note that if *vlan_proto* is different from 1014 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1015 * be **ETH_P_8021Q**. 1016 * 1017 * A call to this helper is susceptible to change the underlying 1018 * packet buffer. Therefore, at load time, all checks on pointers 1019 * previously done by the verifier are invalidated and must be 1020 * performed again, if the helper is used in combination with 1021 * direct packet access. 1022 * Return 1023 * 0 on success, or a negative error in case of failure. 1024 * 1025 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1026 * Description 1027 * Pop a VLAN header from the packet associated to *skb*. 1028 * 1029 * A call to this helper is susceptible to change the underlying 1030 * packet buffer. Therefore, at load time, all checks on pointers 1031 * previously done by the verifier are invalidated and must be 1032 * performed again, if the helper is used in combination with 1033 * direct packet access. 1034 * Return 1035 * 0 on success, or a negative error in case of failure. 1036 * 1037 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1038 * Description 1039 * Get tunnel metadata. This helper takes a pointer *key* to an 1040 * empty **struct bpf_tunnel_key** of **size**, that will be 1041 * filled with tunnel metadata for the packet associated to *skb*. 1042 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1043 * indicates that the tunnel is based on IPv6 protocol instead of 1044 * IPv4. 1045 * 1046 * The **struct bpf_tunnel_key** is an object that generalizes the 1047 * principal parameters used by various tunneling protocols into a 1048 * single struct. This way, it can be used to easily make a 1049 * decision based on the contents of the encapsulation header, 1050 * "summarized" in this struct. In particular, it holds the IP 1051 * address of the remote end (IPv4 or IPv6, depending on the case) 1052 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1053 * this struct exposes the *key*\ **->tunnel_id**, which is 1054 * generally mapped to a VNI (Virtual Network Identifier), making 1055 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1056 * () helper. 1057 * 1058 * Let's imagine that the following code is part of a program 1059 * attached to the TC ingress interface, on one end of a GRE 1060 * tunnel, and is supposed to filter out all messages coming from 1061 * remote ends with IPv4 address other than 10.0.0.1: 1062 * 1063 * :: 1064 * 1065 * int ret; 1066 * struct bpf_tunnel_key key = {}; 1067 * 1068 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1069 * if (ret < 0) 1070 * return TC_ACT_SHOT; // drop packet 1071 * 1072 * if (key.remote_ipv4 != 0x0a000001) 1073 * return TC_ACT_SHOT; // drop packet 1074 * 1075 * return TC_ACT_OK; // accept packet 1076 * 1077 * This interface can also be used with all encapsulation devices 1078 * that can operate in "collect metadata" mode: instead of having 1079 * one network device per specific configuration, the "collect 1080 * metadata" mode only requires a single device where the 1081 * configuration can be extracted from this helper. 1082 * 1083 * This can be used together with various tunnels such as VXLan, 1084 * Geneve, GRE or IP in IP (IPIP). 1085 * Return 1086 * 0 on success, or a negative error in case of failure. 1087 * 1088 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1089 * Description 1090 * Populate tunnel metadata for packet associated to *skb.* The 1091 * tunnel metadata is set to the contents of *key*, of *size*. The 1092 * *flags* can be set to a combination of the following values: 1093 * 1094 * **BPF_F_TUNINFO_IPV6** 1095 * Indicate that the tunnel is based on IPv6 protocol 1096 * instead of IPv4. 1097 * **BPF_F_ZERO_CSUM_TX** 1098 * For IPv4 packets, add a flag to tunnel metadata 1099 * indicating that checksum computation should be skipped 1100 * and checksum set to zeroes. 1101 * **BPF_F_DONT_FRAGMENT** 1102 * Add a flag to tunnel metadata indicating that the 1103 * packet should not be fragmented. 1104 * **BPF_F_SEQ_NUMBER** 1105 * Add a flag to tunnel metadata indicating that a 1106 * sequence number should be added to tunnel header before 1107 * sending the packet. This flag was added for GRE 1108 * encapsulation, but might be used with other protocols 1109 * as well in the future. 1110 * 1111 * Here is a typical usage on the transmit path: 1112 * 1113 * :: 1114 * 1115 * struct bpf_tunnel_key key; 1116 * populate key ... 1117 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1118 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1119 * 1120 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1121 * helper for additional information. 1122 * Return 1123 * 0 on success, or a negative error in case of failure. 1124 * 1125 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1126 * Description 1127 * Read the value of a perf event counter. This helper relies on a 1128 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1129 * the perf event counter is selected when *map* is updated with 1130 * perf event file descriptors. The *map* is an array whose size 1131 * is the number of available CPUs, and each cell contains a value 1132 * relative to one CPU. The value to retrieve is indicated by 1133 * *flags*, that contains the index of the CPU to look up, masked 1134 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1135 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1136 * current CPU should be retrieved. 1137 * 1138 * Note that before Linux 4.13, only hardware perf event can be 1139 * retrieved. 1140 * 1141 * Also, be aware that the newer helper 1142 * **bpf_perf_event_read_value**\ () is recommended over 1143 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1144 * quirks where error and counter value are used as a return code 1145 * (which is wrong to do since ranges may overlap). This issue is 1146 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1147 * time provides more features over the **bpf_perf_event_read**\ 1148 * () interface. Please refer to the description of 1149 * **bpf_perf_event_read_value**\ () for details. 1150 * Return 1151 * The value of the perf event counter read from the map, or a 1152 * negative error code in case of failure. 1153 * 1154 * long bpf_redirect(u32 ifindex, u64 flags) 1155 * Description 1156 * Redirect the packet to another net device of index *ifindex*. 1157 * This helper is somewhat similar to **bpf_clone_redirect**\ 1158 * (), except that the packet is not cloned, which provides 1159 * increased performance. 1160 * 1161 * Except for XDP, both ingress and egress interfaces can be used 1162 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1163 * to make the distinction (ingress path is selected if the flag 1164 * is present, egress path otherwise). Currently, XDP only 1165 * supports redirection to the egress interface, and accepts no 1166 * flag at all. 1167 * 1168 * The same effect can also be attained with the more generic 1169 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1170 * redirect target instead of providing it directly to the helper. 1171 * Return 1172 * For XDP, the helper returns **XDP_REDIRECT** on success or 1173 * **XDP_ABORTED** on error. For other program types, the values 1174 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1175 * error. 1176 * 1177 * u32 bpf_get_route_realm(struct sk_buff *skb) 1178 * Description 1179 * Retrieve the realm or the route, that is to say the 1180 * **tclassid** field of the destination for the *skb*. The 1181 * identifier retrieved is a user-provided tag, similar to the 1182 * one used with the net_cls cgroup (see description for 1183 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1184 * held by a route (a destination entry), not by a task. 1185 * 1186 * Retrieving this identifier works with the clsact TC egress hook 1187 * (see also **tc-bpf(8)**), or alternatively on conventional 1188 * classful egress qdiscs, but not on TC ingress path. In case of 1189 * clsact TC egress hook, this has the advantage that, internally, 1190 * the destination entry has not been dropped yet in the transmit 1191 * path. Therefore, the destination entry does not need to be 1192 * artificially held via **netif_keep_dst**\ () for a classful 1193 * qdisc until the *skb* is freed. 1194 * 1195 * This helper is available only if the kernel was compiled with 1196 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1197 * Return 1198 * The realm of the route for the packet associated to *skb*, or 0 1199 * if none was found. 1200 * 1201 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1202 * Description 1203 * Write raw *data* blob into a special BPF perf event held by 1204 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1205 * event must have the following attributes: **PERF_SAMPLE_RAW** 1206 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1207 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1208 * 1209 * The *flags* are used to indicate the index in *map* for which 1210 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1211 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1212 * to indicate that the index of the current CPU core should be 1213 * used. 1214 * 1215 * The value to write, of *size*, is passed through eBPF stack and 1216 * pointed by *data*. 1217 * 1218 * The context of the program *ctx* needs also be passed to the 1219 * helper. 1220 * 1221 * On user space, a program willing to read the values needs to 1222 * call **perf_event_open**\ () on the perf event (either for 1223 * one or for all CPUs) and to store the file descriptor into the 1224 * *map*. This must be done before the eBPF program can send data 1225 * into it. An example is available in file 1226 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1227 * tree (the eBPF program counterpart is in 1228 * *samples/bpf/trace_output_kern.c*). 1229 * 1230 * **bpf_perf_event_output**\ () achieves better performance 1231 * than **bpf_trace_printk**\ () for sharing data with user 1232 * space, and is much better suitable for streaming data from eBPF 1233 * programs. 1234 * 1235 * Note that this helper is not restricted to tracing use cases 1236 * and can be used with programs attached to TC or XDP as well, 1237 * where it allows for passing data to user space listeners. Data 1238 * can be: 1239 * 1240 * * Only custom structs, 1241 * * Only the packet payload, or 1242 * * A combination of both. 1243 * Return 1244 * 0 on success, or a negative error in case of failure. 1245 * 1246 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1247 * Description 1248 * This helper was provided as an easy way to load data from a 1249 * packet. It can be used to load *len* bytes from *offset* from 1250 * the packet associated to *skb*, into the buffer pointed by 1251 * *to*. 1252 * 1253 * Since Linux 4.7, usage of this helper has mostly been replaced 1254 * by "direct packet access", enabling packet data to be 1255 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1256 * pointing respectively to the first byte of packet data and to 1257 * the byte after the last byte of packet data. However, it 1258 * remains useful if one wishes to read large quantities of data 1259 * at once from a packet into the eBPF stack. 1260 * Return 1261 * 0 on success, or a negative error in case of failure. 1262 * 1263 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1264 * Description 1265 * Walk a user or a kernel stack and return its id. To achieve 1266 * this, the helper needs *ctx*, which is a pointer to the context 1267 * on which the tracing program is executed, and a pointer to a 1268 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1269 * 1270 * The last argument, *flags*, holds the number of stack frames to 1271 * skip (from 0 to 255), masked with 1272 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1273 * a combination of the following flags: 1274 * 1275 * **BPF_F_USER_STACK** 1276 * Collect a user space stack instead of a kernel stack. 1277 * **BPF_F_FAST_STACK_CMP** 1278 * Compare stacks by hash only. 1279 * **BPF_F_REUSE_STACKID** 1280 * If two different stacks hash into the same *stackid*, 1281 * discard the old one. 1282 * 1283 * The stack id retrieved is a 32 bit long integer handle which 1284 * can be further combined with other data (including other stack 1285 * ids) and used as a key into maps. This can be useful for 1286 * generating a variety of graphs (such as flame graphs or off-cpu 1287 * graphs). 1288 * 1289 * For walking a stack, this helper is an improvement over 1290 * **bpf_probe_read**\ (), which can be used with unrolled loops 1291 * but is not efficient and consumes a lot of eBPF instructions. 1292 * Instead, **bpf_get_stackid**\ () can collect up to 1293 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1294 * this limit can be controlled with the **sysctl** program, and 1295 * that it should be manually increased in order to profile long 1296 * user stacks (such as stacks for Java programs). To do so, use: 1297 * 1298 * :: 1299 * 1300 * # sysctl kernel.perf_event_max_stack=<new value> 1301 * Return 1302 * The positive or null stack id on success, or a negative error 1303 * in case of failure. 1304 * 1305 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1306 * Description 1307 * Compute a checksum difference, from the raw buffer pointed by 1308 * *from*, of length *from_size* (that must be a multiple of 4), 1309 * towards the raw buffer pointed by *to*, of size *to_size* 1310 * (same remark). An optional *seed* can be added to the value 1311 * (this can be cascaded, the seed may come from a previous call 1312 * to the helper). 1313 * 1314 * This is flexible enough to be used in several ways: 1315 * 1316 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1317 * checksum, it can be used when pushing new data. 1318 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1319 * checksum, it can be used when removing data from a packet. 1320 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1321 * can be used to compute a diff. Note that *from_size* and 1322 * *to_size* do not need to be equal. 1323 * 1324 * This helper can be used in combination with 1325 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1326 * which one can feed in the difference computed with 1327 * **bpf_csum_diff**\ (). 1328 * Return 1329 * The checksum result, or a negative error code in case of 1330 * failure. 1331 * 1332 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1333 * Description 1334 * Retrieve tunnel options metadata for the packet associated to 1335 * *skb*, and store the raw tunnel option data to the buffer *opt* 1336 * of *size*. 1337 * 1338 * This helper can be used with encapsulation devices that can 1339 * operate in "collect metadata" mode (please refer to the related 1340 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1341 * more details). A particular example where this can be used is 1342 * in combination with the Geneve encapsulation protocol, where it 1343 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1344 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1345 * the eBPF program. This allows for full customization of these 1346 * headers. 1347 * Return 1348 * The size of the option data retrieved. 1349 * 1350 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1351 * Description 1352 * Set tunnel options metadata for the packet associated to *skb* 1353 * to the option data contained in the raw buffer *opt* of *size*. 1354 * 1355 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1356 * helper for additional information. 1357 * Return 1358 * 0 on success, or a negative error in case of failure. 1359 * 1360 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1361 * Description 1362 * Change the protocol of the *skb* to *proto*. Currently 1363 * supported are transition from IPv4 to IPv6, and from IPv6 to 1364 * IPv4. The helper takes care of the groundwork for the 1365 * transition, including resizing the socket buffer. The eBPF 1366 * program is expected to fill the new headers, if any, via 1367 * **skb_store_bytes**\ () and to recompute the checksums with 1368 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1369 * (). The main case for this helper is to perform NAT64 1370 * operations out of an eBPF program. 1371 * 1372 * Internally, the GSO type is marked as dodgy so that headers are 1373 * checked and segments are recalculated by the GSO/GRO engine. 1374 * The size for GSO target is adapted as well. 1375 * 1376 * All values for *flags* are reserved for future usage, and must 1377 * be left at zero. 1378 * 1379 * A call to this helper is susceptible to change the underlying 1380 * packet buffer. Therefore, at load time, all checks on pointers 1381 * previously done by the verifier are invalidated and must be 1382 * performed again, if the helper is used in combination with 1383 * direct packet access. 1384 * Return 1385 * 0 on success, or a negative error in case of failure. 1386 * 1387 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 1388 * Description 1389 * Change the packet type for the packet associated to *skb*. This 1390 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1391 * the eBPF program does not have a write access to *skb*\ 1392 * **->pkt_type** beside this helper. Using a helper here allows 1393 * for graceful handling of errors. 1394 * 1395 * The major use case is to change incoming *skb*s to 1396 * **PACKET_HOST** in a programmatic way instead of having to 1397 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1398 * example. 1399 * 1400 * Note that *type* only allows certain values. At this time, they 1401 * are: 1402 * 1403 * **PACKET_HOST** 1404 * Packet is for us. 1405 * **PACKET_BROADCAST** 1406 * Send packet to all. 1407 * **PACKET_MULTICAST** 1408 * Send packet to group. 1409 * **PACKET_OTHERHOST** 1410 * Send packet to someone else. 1411 * Return 1412 * 0 on success, or a negative error in case of failure. 1413 * 1414 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1415 * Description 1416 * Check whether *skb* is a descendant of the cgroup2 held by 1417 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1418 * Return 1419 * The return value depends on the result of the test, and can be: 1420 * 1421 * * 0, if the *skb* failed the cgroup2 descendant test. 1422 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1423 * * A negative error code, if an error occurred. 1424 * 1425 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1426 * Description 1427 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1428 * not set, in particular if the hash was cleared due to mangling, 1429 * recompute this hash. Later accesses to the hash can be done 1430 * directly with *skb*\ **->hash**. 1431 * 1432 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1433 * prototype with **bpf_skb_change_proto**\ (), or calling 1434 * **bpf_skb_store_bytes**\ () with the 1435 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1436 * the hash and to trigger a new computation for the next call to 1437 * **bpf_get_hash_recalc**\ (). 1438 * Return 1439 * The 32-bit hash. 1440 * 1441 * u64 bpf_get_current_task(void) 1442 * Return 1443 * A pointer to the current task struct. 1444 * 1445 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 1446 * Description 1447 * Attempt in a safe way to write *len* bytes from the buffer 1448 * *src* to *dst* in memory. It only works for threads that are in 1449 * user context, and *dst* must be a valid user space address. 1450 * 1451 * This helper should not be used to implement any kind of 1452 * security mechanism because of TOC-TOU attacks, but rather to 1453 * debug, divert, and manipulate execution of semi-cooperative 1454 * processes. 1455 * 1456 * Keep in mind that this feature is meant for experiments, and it 1457 * has a risk of crashing the system and running programs. 1458 * Therefore, when an eBPF program using this helper is attached, 1459 * a warning including PID and process name is printed to kernel 1460 * logs. 1461 * Return 1462 * 0 on success, or a negative error in case of failure. 1463 * 1464 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1465 * Description 1466 * Check whether the probe is being run is the context of a given 1467 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1468 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1469 * Return 1470 * The return value depends on the result of the test, and can be: 1471 * 1472 * * 0, if current task belongs to the cgroup2. 1473 * * 1, if current task does not belong to the cgroup2. 1474 * * A negative error code, if an error occurred. 1475 * 1476 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1477 * Description 1478 * Resize (trim or grow) the packet associated to *skb* to the 1479 * new *len*. The *flags* are reserved for future usage, and must 1480 * be left at zero. 1481 * 1482 * The basic idea is that the helper performs the needed work to 1483 * change the size of the packet, then the eBPF program rewrites 1484 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1485 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1486 * and others. This helper is a slow path utility intended for 1487 * replies with control messages. And because it is targeted for 1488 * slow path, the helper itself can afford to be slow: it 1489 * implicitly linearizes, unclones and drops offloads from the 1490 * *skb*. 1491 * 1492 * A call to this helper is susceptible to change the underlying 1493 * packet buffer. Therefore, at load time, all checks on pointers 1494 * previously done by the verifier are invalidated and must be 1495 * performed again, if the helper is used in combination with 1496 * direct packet access. 1497 * Return 1498 * 0 on success, or a negative error in case of failure. 1499 * 1500 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1501 * Description 1502 * Pull in non-linear data in case the *skb* is non-linear and not 1503 * all of *len* are part of the linear section. Make *len* bytes 1504 * from *skb* readable and writable. If a zero value is passed for 1505 * *len*, then the whole length of the *skb* is pulled. 1506 * 1507 * This helper is only needed for reading and writing with direct 1508 * packet access. 1509 * 1510 * For direct packet access, testing that offsets to access 1511 * are within packet boundaries (test on *skb*\ **->data_end**) is 1512 * susceptible to fail if offsets are invalid, or if the requested 1513 * data is in non-linear parts of the *skb*. On failure the 1514 * program can just bail out, or in the case of a non-linear 1515 * buffer, use a helper to make the data available. The 1516 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1517 * the data. Another one consists in using **bpf_skb_pull_data** 1518 * to pull in once the non-linear parts, then retesting and 1519 * eventually access the data. 1520 * 1521 * At the same time, this also makes sure the *skb* is uncloned, 1522 * which is a necessary condition for direct write. As this needs 1523 * to be an invariant for the write part only, the verifier 1524 * detects writes and adds a prologue that is calling 1525 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1526 * the very beginning in case it is indeed cloned. 1527 * 1528 * A call to this helper is susceptible to change the underlying 1529 * packet buffer. Therefore, at load time, all checks on pointers 1530 * previously done by the verifier are invalidated and must be 1531 * performed again, if the helper is used in combination with 1532 * direct packet access. 1533 * Return 1534 * 0 on success, or a negative error in case of failure. 1535 * 1536 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1537 * Description 1538 * Add the checksum *csum* into *skb*\ **->csum** in case the 1539 * driver has supplied a checksum for the entire packet into that 1540 * field. Return an error otherwise. This helper is intended to be 1541 * used in combination with **bpf_csum_diff**\ (), in particular 1542 * when the checksum needs to be updated after data has been 1543 * written into the packet through direct packet access. 1544 * Return 1545 * The checksum on success, or a negative error code in case of 1546 * failure. 1547 * 1548 * void bpf_set_hash_invalid(struct sk_buff *skb) 1549 * Description 1550 * Invalidate the current *skb*\ **->hash**. It can be used after 1551 * mangling on headers through direct packet access, in order to 1552 * indicate that the hash is outdated and to trigger a 1553 * recalculation the next time the kernel tries to access this 1554 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1555 * 1556 * long bpf_get_numa_node_id(void) 1557 * Description 1558 * Return the id of the current NUMA node. The primary use case 1559 * for this helper is the selection of sockets for the local NUMA 1560 * node, when the program is attached to sockets using the 1561 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1562 * but the helper is also available to other eBPF program types, 1563 * similarly to **bpf_get_smp_processor_id**\ (). 1564 * Return 1565 * The id of current NUMA node. 1566 * 1567 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1568 * Description 1569 * Grows headroom of packet associated to *skb* and adjusts the 1570 * offset of the MAC header accordingly, adding *len* bytes of 1571 * space. It automatically extends and reallocates memory as 1572 * required. 1573 * 1574 * This helper can be used on a layer 3 *skb* to push a MAC header 1575 * for redirection into a layer 2 device. 1576 * 1577 * All values for *flags* are reserved for future usage, and must 1578 * be left at zero. 1579 * 1580 * A call to this helper is susceptible to change the underlying 1581 * packet buffer. Therefore, at load time, all checks on pointers 1582 * previously done by the verifier are invalidated and must be 1583 * performed again, if the helper is used in combination with 1584 * direct packet access. 1585 * Return 1586 * 0 on success, or a negative error in case of failure. 1587 * 1588 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1589 * Description 1590 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1591 * it is possible to use a negative value for *delta*. This helper 1592 * can be used to prepare the packet for pushing or popping 1593 * headers. 1594 * 1595 * A call to this helper is susceptible to change the underlying 1596 * packet buffer. Therefore, at load time, all checks on pointers 1597 * previously done by the verifier are invalidated and must be 1598 * performed again, if the helper is used in combination with 1599 * direct packet access. 1600 * Return 1601 * 0 on success, or a negative error in case of failure. 1602 * 1603 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1604 * Description 1605 * Copy a NUL terminated string from an unsafe kernel address 1606 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1607 * more details. 1608 * 1609 * Generally, use **bpf_probe_read_user_str**\ () or 1610 * **bpf_probe_read_kernel_str**\ () instead. 1611 * Return 1612 * On success, the strictly positive length of the string, 1613 * including the trailing NUL character. On error, a negative 1614 * value. 1615 * 1616 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1617 * Description 1618 * If the **struct sk_buff** pointed by *skb* has a known socket, 1619 * retrieve the cookie (generated by the kernel) of this socket. 1620 * If no cookie has been set yet, generate a new cookie. Once 1621 * generated, the socket cookie remains stable for the life of the 1622 * socket. This helper can be useful for monitoring per socket 1623 * networking traffic statistics as it provides a global socket 1624 * identifier that can be assumed unique. 1625 * Return 1626 * A 8-byte long non-decreasing number on success, or 0 if the 1627 * socket field is missing inside *skb*. 1628 * 1629 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1630 * Description 1631 * Equivalent to bpf_get_socket_cookie() helper that accepts 1632 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1633 * Return 1634 * A 8-byte long non-decreasing number. 1635 * 1636 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1637 * Description 1638 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 1639 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1640 * Return 1641 * A 8-byte long non-decreasing number. 1642 * 1643 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1644 * Return 1645 * The owner UID of the socket associated to *skb*. If the socket 1646 * is **NULL**, or if it is not a full socket (i.e. if it is a 1647 * time-wait or a request socket instead), **overflowuid** value 1648 * is returned (note that **overflowuid** might also be the actual 1649 * UID value for the socket). 1650 * 1651 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 1652 * Description 1653 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1654 * to value *hash*. 1655 * Return 1656 * 0 1657 * 1658 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1659 * Description 1660 * Emulate a call to **setsockopt()** on the socket associated to 1661 * *bpf_socket*, which must be a full socket. The *level* at 1662 * which the option resides and the name *optname* of the option 1663 * must be specified, see **setsockopt(2)** for more information. 1664 * The option value of length *optlen* is pointed by *optval*. 1665 * 1666 * *bpf_socket* should be one of the following: 1667 * 1668 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1669 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1670 * and **BPF_CGROUP_INET6_CONNECT**. 1671 * 1672 * This helper actually implements a subset of **setsockopt()**. 1673 * It supports the following *level*\ s: 1674 * 1675 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1676 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1677 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1678 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 1679 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1680 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1681 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1682 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1683 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**. 1684 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1685 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1686 * Return 1687 * 0 on success, or a negative error in case of failure. 1688 * 1689 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1690 * Description 1691 * Grow or shrink the room for data in the packet associated to 1692 * *skb* by *len_diff*, and according to the selected *mode*. 1693 * 1694 * By default, the helper will reset any offloaded checksum 1695 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1696 * by the following flag: 1697 * 1698 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1699 * checksum data of the skb to CHECKSUM_NONE. 1700 * 1701 * There are two supported modes at this time: 1702 * 1703 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1704 * (room space is added or removed below the layer 2 header). 1705 * 1706 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1707 * (room space is added or removed below the layer 3 header). 1708 * 1709 * The following flags are supported at this time: 1710 * 1711 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1712 * Adjusting mss in this way is not allowed for datagrams. 1713 * 1714 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1715 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1716 * Any new space is reserved to hold a tunnel header. 1717 * Configure skb offsets and other fields accordingly. 1718 * 1719 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1720 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1721 * Use with ENCAP_L3 flags to further specify the tunnel type. 1722 * 1723 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1724 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1725 * type; *len* is the length of the inner MAC header. 1726 * 1727 * A call to this helper is susceptible to change the underlying 1728 * packet buffer. Therefore, at load time, all checks on pointers 1729 * previously done by the verifier are invalidated and must be 1730 * performed again, if the helper is used in combination with 1731 * direct packet access. 1732 * Return 1733 * 0 on success, or a negative error in case of failure. 1734 * 1735 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1736 * Description 1737 * Redirect the packet to the endpoint referenced by *map* at 1738 * index *key*. Depending on its type, this *map* can contain 1739 * references to net devices (for forwarding packets through other 1740 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1741 * but this is only implemented for native XDP (with driver 1742 * support) as of this writing). 1743 * 1744 * The lower two bits of *flags* are used as the return code if 1745 * the map lookup fails. This is so that the return value can be 1746 * one of the XDP program return codes up to **XDP_TX**, as chosen 1747 * by the caller. Any higher bits in the *flags* argument must be 1748 * unset. 1749 * 1750 * See also **bpf_redirect**\ (), which only supports redirecting 1751 * to an ifindex, but doesn't require a map to do so. 1752 * Return 1753 * **XDP_REDIRECT** on success, or the value of the two lower bits 1754 * of the *flags* argument on error. 1755 * 1756 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1757 * Description 1758 * Redirect the packet to the socket referenced by *map* (of type 1759 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1760 * egress interfaces can be used for redirection. The 1761 * **BPF_F_INGRESS** value in *flags* is used to make the 1762 * distinction (ingress path is selected if the flag is present, 1763 * egress path otherwise). This is the only flag supported for now. 1764 * Return 1765 * **SK_PASS** on success, or **SK_DROP** on error. 1766 * 1767 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1768 * Description 1769 * Add an entry to, or update a *map* referencing sockets. The 1770 * *skops* is used as a new value for the entry associated to 1771 * *key*. *flags* is one of: 1772 * 1773 * **BPF_NOEXIST** 1774 * The entry for *key* must not exist in the map. 1775 * **BPF_EXIST** 1776 * The entry for *key* must already exist in the map. 1777 * **BPF_ANY** 1778 * No condition on the existence of the entry for *key*. 1779 * 1780 * If the *map* has eBPF programs (parser and verdict), those will 1781 * be inherited by the socket being added. If the socket is 1782 * already attached to eBPF programs, this results in an error. 1783 * Return 1784 * 0 on success, or a negative error in case of failure. 1785 * 1786 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1787 * Description 1788 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1789 * *delta* (which can be positive or negative). Note that this 1790 * operation modifies the address stored in *xdp_md*\ **->data**, 1791 * so the latter must be loaded only after the helper has been 1792 * called. 1793 * 1794 * The use of *xdp_md*\ **->data_meta** is optional and programs 1795 * are not required to use it. The rationale is that when the 1796 * packet is processed with XDP (e.g. as DoS filter), it is 1797 * possible to push further meta data along with it before passing 1798 * to the stack, and to give the guarantee that an ingress eBPF 1799 * program attached as a TC classifier on the same device can pick 1800 * this up for further post-processing. Since TC works with socket 1801 * buffers, it remains possible to set from XDP the **mark** or 1802 * **priority** pointers, or other pointers for the socket buffer. 1803 * Having this scratch space generic and programmable allows for 1804 * more flexibility as the user is free to store whatever meta 1805 * data they need. 1806 * 1807 * A call to this helper is susceptible to change the underlying 1808 * packet buffer. Therefore, at load time, all checks on pointers 1809 * previously done by the verifier are invalidated and must be 1810 * performed again, if the helper is used in combination with 1811 * direct packet access. 1812 * Return 1813 * 0 on success, or a negative error in case of failure. 1814 * 1815 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1816 * Description 1817 * Read the value of a perf event counter, and store it into *buf* 1818 * of size *buf_size*. This helper relies on a *map* of type 1819 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1820 * counter is selected when *map* is updated with perf event file 1821 * descriptors. The *map* is an array whose size is the number of 1822 * available CPUs, and each cell contains a value relative to one 1823 * CPU. The value to retrieve is indicated by *flags*, that 1824 * contains the index of the CPU to look up, masked with 1825 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1826 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1827 * current CPU should be retrieved. 1828 * 1829 * This helper behaves in a way close to 1830 * **bpf_perf_event_read**\ () helper, save that instead of 1831 * just returning the value observed, it fills the *buf* 1832 * structure. This allows for additional data to be retrieved: in 1833 * particular, the enabled and running times (in *buf*\ 1834 * **->enabled** and *buf*\ **->running**, respectively) are 1835 * copied. In general, **bpf_perf_event_read_value**\ () is 1836 * recommended over **bpf_perf_event_read**\ (), which has some 1837 * ABI issues and provides fewer functionalities. 1838 * 1839 * These values are interesting, because hardware PMU (Performance 1840 * Monitoring Unit) counters are limited resources. When there are 1841 * more PMU based perf events opened than available counters, 1842 * kernel will multiplex these events so each event gets certain 1843 * percentage (but not all) of the PMU time. In case that 1844 * multiplexing happens, the number of samples or counter value 1845 * will not reflect the case compared to when no multiplexing 1846 * occurs. This makes comparison between different runs difficult. 1847 * Typically, the counter value should be normalized before 1848 * comparing to other experiments. The usual normalization is done 1849 * as follows. 1850 * 1851 * :: 1852 * 1853 * normalized_counter = counter * t_enabled / t_running 1854 * 1855 * Where t_enabled is the time enabled for event and t_running is 1856 * the time running for event since last normalization. The 1857 * enabled and running times are accumulated since the perf event 1858 * open. To achieve scaling factor between two invocations of an 1859 * eBPF program, users can use CPU id as the key (which is 1860 * typical for perf array usage model) to remember the previous 1861 * value and do the calculation inside the eBPF program. 1862 * Return 1863 * 0 on success, or a negative error in case of failure. 1864 * 1865 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1866 * Description 1867 * For en eBPF program attached to a perf event, retrieve the 1868 * value of the event counter associated to *ctx* and store it in 1869 * the structure pointed by *buf* and of size *buf_size*. Enabled 1870 * and running times are also stored in the structure (see 1871 * description of helper **bpf_perf_event_read_value**\ () for 1872 * more details). 1873 * Return 1874 * 0 on success, or a negative error in case of failure. 1875 * 1876 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1877 * Description 1878 * Emulate a call to **getsockopt()** on the socket associated to 1879 * *bpf_socket*, which must be a full socket. The *level* at 1880 * which the option resides and the name *optname* of the option 1881 * must be specified, see **getsockopt(2)** for more information. 1882 * The retrieved value is stored in the structure pointed by 1883 * *opval* and of length *optlen*. 1884 * 1885 * *bpf_socket* should be one of the following: 1886 * 1887 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1888 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1889 * and **BPF_CGROUP_INET6_CONNECT**. 1890 * 1891 * This helper actually implements a subset of **getsockopt()**. 1892 * It supports the following *level*\ s: 1893 * 1894 * * **IPPROTO_TCP**, which supports *optname* 1895 * **TCP_CONGESTION**. 1896 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1897 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1898 * Return 1899 * 0 on success, or a negative error in case of failure. 1900 * 1901 * long bpf_override_return(struct pt_regs *regs, u64 rc) 1902 * Description 1903 * Used for error injection, this helper uses kprobes to override 1904 * the return value of the probed function, and to set it to *rc*. 1905 * The first argument is the context *regs* on which the kprobe 1906 * works. 1907 * 1908 * This helper works by setting the PC (program counter) 1909 * to an override function which is run in place of the original 1910 * probed function. This means the probed function is not run at 1911 * all. The replacement function just returns with the required 1912 * value. 1913 * 1914 * This helper has security implications, and thus is subject to 1915 * restrictions. It is only available if the kernel was compiled 1916 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1917 * option, and in this case it only works on functions tagged with 1918 * **ALLOW_ERROR_INJECTION** in the kernel code. 1919 * 1920 * Also, the helper is only available for the architectures having 1921 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1922 * x86 architecture is the only one to support this feature. 1923 * Return 1924 * 0 1925 * 1926 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1927 * Description 1928 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1929 * for the full TCP socket associated to *bpf_sock_ops* to 1930 * *argval*. 1931 * 1932 * The primary use of this field is to determine if there should 1933 * be calls to eBPF programs of type 1934 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1935 * code. A program of the same type can change its value, per 1936 * connection and as necessary, when the connection is 1937 * established. This field is directly accessible for reading, but 1938 * this helper must be used for updates in order to return an 1939 * error if an eBPF program tries to set a callback that is not 1940 * supported in the current kernel. 1941 * 1942 * *argval* is a flag array which can combine these flags: 1943 * 1944 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1945 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1946 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1947 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1948 * 1949 * Therefore, this function can be used to clear a callback flag by 1950 * setting the appropriate bit to zero. e.g. to disable the RTO 1951 * callback: 1952 * 1953 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1954 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1955 * 1956 * Here are some examples of where one could call such eBPF 1957 * program: 1958 * 1959 * * When RTO fires. 1960 * * When a packet is retransmitted. 1961 * * When the connection terminates. 1962 * * When a packet is sent. 1963 * * When a packet is received. 1964 * Return 1965 * Code **-EINVAL** if the socket is not a full TCP socket; 1966 * otherwise, a positive number containing the bits that could not 1967 * be set is returned (which comes down to 0 if all bits were set 1968 * as required). 1969 * 1970 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1971 * Description 1972 * This helper is used in programs implementing policies at the 1973 * socket level. If the message *msg* is allowed to pass (i.e. if 1974 * the verdict eBPF program returns **SK_PASS**), redirect it to 1975 * the socket referenced by *map* (of type 1976 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1977 * egress interfaces can be used for redirection. The 1978 * **BPF_F_INGRESS** value in *flags* is used to make the 1979 * distinction (ingress path is selected if the flag is present, 1980 * egress path otherwise). This is the only flag supported for now. 1981 * Return 1982 * **SK_PASS** on success, or **SK_DROP** on error. 1983 * 1984 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1985 * Description 1986 * For socket policies, apply the verdict of the eBPF program to 1987 * the next *bytes* (number of bytes) of message *msg*. 1988 * 1989 * For example, this helper can be used in the following cases: 1990 * 1991 * * A single **sendmsg**\ () or **sendfile**\ () system call 1992 * contains multiple logical messages that the eBPF program is 1993 * supposed to read and for which it should apply a verdict. 1994 * * An eBPF program only cares to read the first *bytes* of a 1995 * *msg*. If the message has a large payload, then setting up 1996 * and calling the eBPF program repeatedly for all bytes, even 1997 * though the verdict is already known, would create unnecessary 1998 * overhead. 1999 * 2000 * When called from within an eBPF program, the helper sets a 2001 * counter internal to the BPF infrastructure, that is used to 2002 * apply the last verdict to the next *bytes*. If *bytes* is 2003 * smaller than the current data being processed from a 2004 * **sendmsg**\ () or **sendfile**\ () system call, the first 2005 * *bytes* will be sent and the eBPF program will be re-run with 2006 * the pointer for start of data pointing to byte number *bytes* 2007 * **+ 1**. If *bytes* is larger than the current data being 2008 * processed, then the eBPF verdict will be applied to multiple 2009 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2010 * consumed. 2011 * 2012 * Note that if a socket closes with the internal counter holding 2013 * a non-zero value, this is not a problem because data is not 2014 * being buffered for *bytes* and is sent as it is received. 2015 * Return 2016 * 0 2017 * 2018 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2019 * Description 2020 * For socket policies, prevent the execution of the verdict eBPF 2021 * program for message *msg* until *bytes* (byte number) have been 2022 * accumulated. 2023 * 2024 * This can be used when one needs a specific number of bytes 2025 * before a verdict can be assigned, even if the data spans 2026 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2027 * case would be a user calling **sendmsg**\ () repeatedly with 2028 * 1-byte long message segments. Obviously, this is bad for 2029 * performance, but it is still valid. If the eBPF program needs 2030 * *bytes* bytes to validate a header, this helper can be used to 2031 * prevent the eBPF program to be called again until *bytes* have 2032 * been accumulated. 2033 * Return 2034 * 0 2035 * 2036 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2037 * Description 2038 * For socket policies, pull in non-linear data from user space 2039 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2040 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2041 * respectively. 2042 * 2043 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2044 * *msg* it can only parse data that the (**data**, **data_end**) 2045 * pointers have already consumed. For **sendmsg**\ () hooks this 2046 * is likely the first scatterlist element. But for calls relying 2047 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2048 * be the range (**0**, **0**) because the data is shared with 2049 * user space and by default the objective is to avoid allowing 2050 * user space to modify data while (or after) eBPF verdict is 2051 * being decided. This helper can be used to pull in data and to 2052 * set the start and end pointer to given values. Data will be 2053 * copied if necessary (i.e. if data was not linear and if start 2054 * and end pointers do not point to the same chunk). 2055 * 2056 * A call to this helper is susceptible to change the underlying 2057 * packet buffer. Therefore, at load time, all checks on pointers 2058 * previously done by the verifier are invalidated and must be 2059 * performed again, if the helper is used in combination with 2060 * direct packet access. 2061 * 2062 * All values for *flags* are reserved for future usage, and must 2063 * be left at zero. 2064 * Return 2065 * 0 on success, or a negative error in case of failure. 2066 * 2067 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2068 * Description 2069 * Bind the socket associated to *ctx* to the address pointed by 2070 * *addr*, of length *addr_len*. This allows for making outgoing 2071 * connection from the desired IP address, which can be useful for 2072 * example when all processes inside a cgroup should use one 2073 * single IP address on a host that has multiple IP configured. 2074 * 2075 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2076 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2077 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2078 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2079 * behavior and lets the kernel efficiently pick up an unused 2080 * port as long as 4-tuple is unique. Passing non-zero port might 2081 * lead to degraded performance. 2082 * Return 2083 * 0 on success, or a negative error in case of failure. 2084 * 2085 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2086 * Description 2087 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2088 * possible to both shrink and grow the packet tail. 2089 * Shrink done via *delta* being a negative integer. 2090 * 2091 * A call to this helper is susceptible to change the underlying 2092 * packet buffer. Therefore, at load time, all checks on pointers 2093 * previously done by the verifier are invalidated and must be 2094 * performed again, if the helper is used in combination with 2095 * direct packet access. 2096 * Return 2097 * 0 on success, or a negative error in case of failure. 2098 * 2099 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2100 * Description 2101 * Retrieve the XFRM state (IP transform framework, see also 2102 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2103 * 2104 * The retrieved value is stored in the **struct bpf_xfrm_state** 2105 * pointed by *xfrm_state* and of length *size*. 2106 * 2107 * All values for *flags* are reserved for future usage, and must 2108 * be left at zero. 2109 * 2110 * This helper is available only if the kernel was compiled with 2111 * **CONFIG_XFRM** configuration option. 2112 * Return 2113 * 0 on success, or a negative error in case of failure. 2114 * 2115 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2116 * Description 2117 * Return a user or a kernel stack in bpf program provided buffer. 2118 * To achieve this, the helper needs *ctx*, which is a pointer 2119 * to the context on which the tracing program is executed. 2120 * To store the stacktrace, the bpf program provides *buf* with 2121 * a nonnegative *size*. 2122 * 2123 * The last argument, *flags*, holds the number of stack frames to 2124 * skip (from 0 to 255), masked with 2125 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2126 * the following flags: 2127 * 2128 * **BPF_F_USER_STACK** 2129 * Collect a user space stack instead of a kernel stack. 2130 * **BPF_F_USER_BUILD_ID** 2131 * Collect buildid+offset instead of ips for user stack, 2132 * only valid if **BPF_F_USER_STACK** is also specified. 2133 * 2134 * **bpf_get_stack**\ () can collect up to 2135 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2136 * to sufficient large buffer size. Note that 2137 * this limit can be controlled with the **sysctl** program, and 2138 * that it should be manually increased in order to profile long 2139 * user stacks (such as stacks for Java programs). To do so, use: 2140 * 2141 * :: 2142 * 2143 * # sysctl kernel.perf_event_max_stack=<new value> 2144 * Return 2145 * A non-negative value equal to or less than *size* on success, 2146 * or a negative error in case of failure. 2147 * 2148 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2149 * Description 2150 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2151 * it provides an easy way to load *len* bytes from *offset* 2152 * from the packet associated to *skb*, into the buffer pointed 2153 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2154 * a fifth argument *start_header* exists in order to select a 2155 * base offset to start from. *start_header* can be one of: 2156 * 2157 * **BPF_HDR_START_MAC** 2158 * Base offset to load data from is *skb*'s mac header. 2159 * **BPF_HDR_START_NET** 2160 * Base offset to load data from is *skb*'s network header. 2161 * 2162 * In general, "direct packet access" is the preferred method to 2163 * access packet data, however, this helper is in particular useful 2164 * in socket filters where *skb*\ **->data** does not always point 2165 * to the start of the mac header and where "direct packet access" 2166 * is not available. 2167 * Return 2168 * 0 on success, or a negative error in case of failure. 2169 * 2170 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2171 * Description 2172 * Do FIB lookup in kernel tables using parameters in *params*. 2173 * If lookup is successful and result shows packet is to be 2174 * forwarded, the neighbor tables are searched for the nexthop. 2175 * If successful (ie., FIB lookup shows forwarding and nexthop 2176 * is resolved), the nexthop address is returned in ipv4_dst 2177 * or ipv6_dst based on family, smac is set to mac address of 2178 * egress device, dmac is set to nexthop mac address, rt_metric 2179 * is set to metric from route (IPv4/IPv6 only), and ifindex 2180 * is set to the device index of the nexthop from the FIB lookup. 2181 * 2182 * *plen* argument is the size of the passed in struct. 2183 * *flags* argument can be a combination of one or more of the 2184 * following values: 2185 * 2186 * **BPF_FIB_LOOKUP_DIRECT** 2187 * Do a direct table lookup vs full lookup using FIB 2188 * rules. 2189 * **BPF_FIB_LOOKUP_OUTPUT** 2190 * Perform lookup from an egress perspective (default is 2191 * ingress). 2192 * 2193 * *ctx* is either **struct xdp_md** for XDP programs or 2194 * **struct sk_buff** tc cls_act programs. 2195 * Return 2196 * * < 0 if any input argument is invalid 2197 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2198 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2199 * packet is not forwarded or needs assist from full stack 2200 * 2201 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2202 * Description 2203 * Add an entry to, or update a sockhash *map* referencing sockets. 2204 * The *skops* is used as a new value for the entry associated to 2205 * *key*. *flags* is one of: 2206 * 2207 * **BPF_NOEXIST** 2208 * The entry for *key* must not exist in the map. 2209 * **BPF_EXIST** 2210 * The entry for *key* must already exist in the map. 2211 * **BPF_ANY** 2212 * No condition on the existence of the entry for *key*. 2213 * 2214 * If the *map* has eBPF programs (parser and verdict), those will 2215 * be inherited by the socket being added. If the socket is 2216 * already attached to eBPF programs, this results in an error. 2217 * Return 2218 * 0 on success, or a negative error in case of failure. 2219 * 2220 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2221 * Description 2222 * This helper is used in programs implementing policies at the 2223 * socket level. If the message *msg* is allowed to pass (i.e. if 2224 * the verdict eBPF program returns **SK_PASS**), redirect it to 2225 * the socket referenced by *map* (of type 2226 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2227 * egress interfaces can be used for redirection. The 2228 * **BPF_F_INGRESS** value in *flags* is used to make the 2229 * distinction (ingress path is selected if the flag is present, 2230 * egress path otherwise). This is the only flag supported for now. 2231 * Return 2232 * **SK_PASS** on success, or **SK_DROP** on error. 2233 * 2234 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2235 * Description 2236 * This helper is used in programs implementing policies at the 2237 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2238 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2239 * to the socket referenced by *map* (of type 2240 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2241 * egress interfaces can be used for redirection. The 2242 * **BPF_F_INGRESS** value in *flags* is used to make the 2243 * distinction (ingress path is selected if the flag is present, 2244 * egress otherwise). This is the only flag supported for now. 2245 * Return 2246 * **SK_PASS** on success, or **SK_DROP** on error. 2247 * 2248 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2249 * Description 2250 * Encapsulate the packet associated to *skb* within a Layer 3 2251 * protocol header. This header is provided in the buffer at 2252 * address *hdr*, with *len* its size in bytes. *type* indicates 2253 * the protocol of the header and can be one of: 2254 * 2255 * **BPF_LWT_ENCAP_SEG6** 2256 * IPv6 encapsulation with Segment Routing Header 2257 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2258 * the IPv6 header is computed by the kernel. 2259 * **BPF_LWT_ENCAP_SEG6_INLINE** 2260 * Only works if *skb* contains an IPv6 packet. Insert a 2261 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2262 * the IPv6 header. 2263 * **BPF_LWT_ENCAP_IP** 2264 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2265 * must be IPv4 or IPv6, followed by zero or more 2266 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2267 * total bytes in all prepended headers. Please note that 2268 * if **skb_is_gso**\ (*skb*) is true, no more than two 2269 * headers can be prepended, and the inner header, if 2270 * present, should be either GRE or UDP/GUE. 2271 * 2272 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2273 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2274 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2275 * **BPF_PROG_TYPE_LWT_XMIT**. 2276 * 2277 * A call to this helper is susceptible to change the underlying 2278 * packet buffer. Therefore, at load time, all checks on pointers 2279 * previously done by the verifier are invalidated and must be 2280 * performed again, if the helper is used in combination with 2281 * direct packet access. 2282 * Return 2283 * 0 on success, or a negative error in case of failure. 2284 * 2285 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2286 * Description 2287 * Store *len* bytes from address *from* into the packet 2288 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2289 * inside the outermost IPv6 Segment Routing Header can be 2290 * modified through this helper. 2291 * 2292 * A call to this helper is susceptible to change the underlying 2293 * packet buffer. Therefore, at load time, all checks on pointers 2294 * previously done by the verifier are invalidated and must be 2295 * performed again, if the helper is used in combination with 2296 * direct packet access. 2297 * Return 2298 * 0 on success, or a negative error in case of failure. 2299 * 2300 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2301 * Description 2302 * Adjust the size allocated to TLVs in the outermost IPv6 2303 * Segment Routing Header contained in the packet associated to 2304 * *skb*, at position *offset* by *delta* bytes. Only offsets 2305 * after the segments are accepted. *delta* can be as well 2306 * positive (growing) as negative (shrinking). 2307 * 2308 * A call to this helper is susceptible to change the underlying 2309 * packet buffer. Therefore, at load time, all checks on pointers 2310 * previously done by the verifier are invalidated and must be 2311 * performed again, if the helper is used in combination with 2312 * direct packet access. 2313 * Return 2314 * 0 on success, or a negative error in case of failure. 2315 * 2316 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2317 * Description 2318 * Apply an IPv6 Segment Routing action of type *action* to the 2319 * packet associated to *skb*. Each action takes a parameter 2320 * contained at address *param*, and of length *param_len* bytes. 2321 * *action* can be one of: 2322 * 2323 * **SEG6_LOCAL_ACTION_END_X** 2324 * End.X action: Endpoint with Layer-3 cross-connect. 2325 * Type of *param*: **struct in6_addr**. 2326 * **SEG6_LOCAL_ACTION_END_T** 2327 * End.T action: Endpoint with specific IPv6 table lookup. 2328 * Type of *param*: **int**. 2329 * **SEG6_LOCAL_ACTION_END_B6** 2330 * End.B6 action: Endpoint bound to an SRv6 policy. 2331 * Type of *param*: **struct ipv6_sr_hdr**. 2332 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2333 * End.B6.Encap action: Endpoint bound to an SRv6 2334 * encapsulation policy. 2335 * Type of *param*: **struct ipv6_sr_hdr**. 2336 * 2337 * A call to this helper is susceptible to change the underlying 2338 * packet buffer. Therefore, at load time, all checks on pointers 2339 * previously done by the verifier are invalidated and must be 2340 * performed again, if the helper is used in combination with 2341 * direct packet access. 2342 * Return 2343 * 0 on success, or a negative error in case of failure. 2344 * 2345 * long bpf_rc_repeat(void *ctx) 2346 * Description 2347 * This helper is used in programs implementing IR decoding, to 2348 * report a successfully decoded repeat key message. This delays 2349 * the generation of a key up event for previously generated 2350 * key down event. 2351 * 2352 * Some IR protocols like NEC have a special IR message for 2353 * repeating last button, for when a button is held down. 2354 * 2355 * The *ctx* should point to the lirc sample as passed into 2356 * the program. 2357 * 2358 * This helper is only available is the kernel was compiled with 2359 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2360 * "**y**". 2361 * Return 2362 * 0 2363 * 2364 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2365 * Description 2366 * This helper is used in programs implementing IR decoding, to 2367 * report a successfully decoded key press with *scancode*, 2368 * *toggle* value in the given *protocol*. The scancode will be 2369 * translated to a keycode using the rc keymap, and reported as 2370 * an input key down event. After a period a key up event is 2371 * generated. This period can be extended by calling either 2372 * **bpf_rc_keydown**\ () again with the same values, or calling 2373 * **bpf_rc_repeat**\ (). 2374 * 2375 * Some protocols include a toggle bit, in case the button was 2376 * released and pressed again between consecutive scancodes. 2377 * 2378 * The *ctx* should point to the lirc sample as passed into 2379 * the program. 2380 * 2381 * The *protocol* is the decoded protocol number (see 2382 * **enum rc_proto** for some predefined values). 2383 * 2384 * This helper is only available is the kernel was compiled with 2385 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2386 * "**y**". 2387 * Return 2388 * 0 2389 * 2390 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2391 * Description 2392 * Return the cgroup v2 id of the socket associated with the *skb*. 2393 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2394 * helper for cgroup v1 by providing a tag resp. identifier that 2395 * can be matched on or used for map lookups e.g. to implement 2396 * policy. The cgroup v2 id of a given path in the hierarchy is 2397 * exposed in user space through the f_handle API in order to get 2398 * to the same 64-bit id. 2399 * 2400 * This helper can be used on TC egress path, but not on ingress, 2401 * and is available only if the kernel was compiled with the 2402 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2403 * Return 2404 * The id is returned or 0 in case the id could not be retrieved. 2405 * 2406 * u64 bpf_get_current_cgroup_id(void) 2407 * Return 2408 * A 64-bit integer containing the current cgroup id based 2409 * on the cgroup within which the current task is running. 2410 * 2411 * void *bpf_get_local_storage(void *map, u64 flags) 2412 * Description 2413 * Get the pointer to the local storage area. 2414 * The type and the size of the local storage is defined 2415 * by the *map* argument. 2416 * The *flags* meaning is specific for each map type, 2417 * and has to be 0 for cgroup local storage. 2418 * 2419 * Depending on the BPF program type, a local storage area 2420 * can be shared between multiple instances of the BPF program, 2421 * running simultaneously. 2422 * 2423 * A user should care about the synchronization by himself. 2424 * For example, by using the **BPF_STX_XADD** instruction to alter 2425 * the shared data. 2426 * Return 2427 * A pointer to the local storage area. 2428 * 2429 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2430 * Description 2431 * Select a **SO_REUSEPORT** socket from a 2432 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2433 * It checks the selected socket is matching the incoming 2434 * request in the socket buffer. 2435 * Return 2436 * 0 on success, or a negative error in case of failure. 2437 * 2438 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2439 * Description 2440 * Return id of cgroup v2 that is ancestor of cgroup associated 2441 * with the *skb* at the *ancestor_level*. The root cgroup is at 2442 * *ancestor_level* zero and each step down the hierarchy 2443 * increments the level. If *ancestor_level* == level of cgroup 2444 * associated with *skb*, then return value will be same as that 2445 * of **bpf_skb_cgroup_id**\ (). 2446 * 2447 * The helper is useful to implement policies based on cgroups 2448 * that are upper in hierarchy than immediate cgroup associated 2449 * with *skb*. 2450 * 2451 * The format of returned id and helper limitations are same as in 2452 * **bpf_skb_cgroup_id**\ (). 2453 * Return 2454 * The id is returned or 0 in case the id could not be retrieved. 2455 * 2456 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2457 * Description 2458 * Look for TCP socket matching *tuple*, optionally in a child 2459 * network namespace *netns*. The return value must be checked, 2460 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2461 * 2462 * The *ctx* should point to the context of the program, such as 2463 * the skb or socket (depending on the hook in use). This is used 2464 * to determine the base network namespace for the lookup. 2465 * 2466 * *tuple_size* must be one of: 2467 * 2468 * **sizeof**\ (*tuple*\ **->ipv4**) 2469 * Look for an IPv4 socket. 2470 * **sizeof**\ (*tuple*\ **->ipv6**) 2471 * Look for an IPv6 socket. 2472 * 2473 * If the *netns* is a negative signed 32-bit integer, then the 2474 * socket lookup table in the netns associated with the *ctx* 2475 * will be used. For the TC hooks, this is the netns of the device 2476 * in the skb. For socket hooks, this is the netns of the socket. 2477 * If *netns* is any other signed 32-bit value greater than or 2478 * equal to zero then it specifies the ID of the netns relative to 2479 * the netns associated with the *ctx*. *netns* values beyond the 2480 * range of 32-bit integers are reserved for future use. 2481 * 2482 * All values for *flags* are reserved for future usage, and must 2483 * be left at zero. 2484 * 2485 * This helper is available only if the kernel was compiled with 2486 * **CONFIG_NET** configuration option. 2487 * Return 2488 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2489 * For sockets with reuseport option, the **struct bpf_sock** 2490 * result is from *reuse*\ **->socks**\ [] using the hash of the 2491 * tuple. 2492 * 2493 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2494 * Description 2495 * Look for UDP socket matching *tuple*, optionally in a child 2496 * network namespace *netns*. The return value must be checked, 2497 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2498 * 2499 * The *ctx* should point to the context of the program, such as 2500 * the skb or socket (depending on the hook in use). This is used 2501 * to determine the base network namespace for the lookup. 2502 * 2503 * *tuple_size* must be one of: 2504 * 2505 * **sizeof**\ (*tuple*\ **->ipv4**) 2506 * Look for an IPv4 socket. 2507 * **sizeof**\ (*tuple*\ **->ipv6**) 2508 * Look for an IPv6 socket. 2509 * 2510 * If the *netns* is a negative signed 32-bit integer, then the 2511 * socket lookup table in the netns associated with the *ctx* 2512 * will be used. For the TC hooks, this is the netns of the device 2513 * in the skb. For socket hooks, this is the netns of the socket. 2514 * If *netns* is any other signed 32-bit value greater than or 2515 * equal to zero then it specifies the ID of the netns relative to 2516 * the netns associated with the *ctx*. *netns* values beyond the 2517 * range of 32-bit integers are reserved for future use. 2518 * 2519 * All values for *flags* are reserved for future usage, and must 2520 * be left at zero. 2521 * 2522 * This helper is available only if the kernel was compiled with 2523 * **CONFIG_NET** configuration option. 2524 * Return 2525 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2526 * For sockets with reuseport option, the **struct bpf_sock** 2527 * result is from *reuse*\ **->socks**\ [] using the hash of the 2528 * tuple. 2529 * 2530 * long bpf_sk_release(void *sock) 2531 * Description 2532 * Release the reference held by *sock*. *sock* must be a 2533 * non-**NULL** pointer that was returned from 2534 * **bpf_sk_lookup_xxx**\ (). 2535 * Return 2536 * 0 on success, or a negative error in case of failure. 2537 * 2538 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2539 * Description 2540 * Push an element *value* in *map*. *flags* is one of: 2541 * 2542 * **BPF_EXIST** 2543 * If the queue/stack is full, the oldest element is 2544 * removed to make room for this. 2545 * Return 2546 * 0 on success, or a negative error in case of failure. 2547 * 2548 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 2549 * Description 2550 * Pop an element from *map*. 2551 * Return 2552 * 0 on success, or a negative error in case of failure. 2553 * 2554 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 2555 * Description 2556 * Get an element from *map* without removing it. 2557 * Return 2558 * 0 on success, or a negative error in case of failure. 2559 * 2560 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2561 * Description 2562 * For socket policies, insert *len* bytes into *msg* at offset 2563 * *start*. 2564 * 2565 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2566 * *msg* it may want to insert metadata or options into the *msg*. 2567 * This can later be read and used by any of the lower layer BPF 2568 * hooks. 2569 * 2570 * This helper may fail if under memory pressure (a malloc 2571 * fails) in these cases BPF programs will get an appropriate 2572 * error and BPF programs will need to handle them. 2573 * Return 2574 * 0 on success, or a negative error in case of failure. 2575 * 2576 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2577 * Description 2578 * Will remove *len* bytes from a *msg* starting at byte *start*. 2579 * This may result in **ENOMEM** errors under certain situations if 2580 * an allocation and copy are required due to a full ring buffer. 2581 * However, the helper will try to avoid doing the allocation 2582 * if possible. Other errors can occur if input parameters are 2583 * invalid either due to *start* byte not being valid part of *msg* 2584 * payload and/or *pop* value being to large. 2585 * Return 2586 * 0 on success, or a negative error in case of failure. 2587 * 2588 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2589 * Description 2590 * This helper is used in programs implementing IR decoding, to 2591 * report a successfully decoded pointer movement. 2592 * 2593 * The *ctx* should point to the lirc sample as passed into 2594 * the program. 2595 * 2596 * This helper is only available is the kernel was compiled with 2597 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2598 * "**y**". 2599 * Return 2600 * 0 2601 * 2602 * long bpf_spin_lock(struct bpf_spin_lock *lock) 2603 * Description 2604 * Acquire a spinlock represented by the pointer *lock*, which is 2605 * stored as part of a value of a map. Taking the lock allows to 2606 * safely update the rest of the fields in that value. The 2607 * spinlock can (and must) later be released with a call to 2608 * **bpf_spin_unlock**\ (\ *lock*\ ). 2609 * 2610 * Spinlocks in BPF programs come with a number of restrictions 2611 * and constraints: 2612 * 2613 * * **bpf_spin_lock** objects are only allowed inside maps of 2614 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2615 * list could be extended in the future). 2616 * * BTF description of the map is mandatory. 2617 * * The BPF program can take ONE lock at a time, since taking two 2618 * or more could cause dead locks. 2619 * * Only one **struct bpf_spin_lock** is allowed per map element. 2620 * * When the lock is taken, calls (either BPF to BPF or helpers) 2621 * are not allowed. 2622 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2623 * allowed inside a spinlock-ed region. 2624 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2625 * the lock, on all execution paths, before it returns. 2626 * * The BPF program can access **struct bpf_spin_lock** only via 2627 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2628 * helpers. Loading or storing data into the **struct 2629 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2630 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2631 * of the map value must be a struct and have **struct 2632 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2633 * Nested lock inside another struct is not allowed. 2634 * * The **struct bpf_spin_lock** *lock* field in a map value must 2635 * be aligned on a multiple of 4 bytes in that value. 2636 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2637 * the **bpf_spin_lock** field to user space. 2638 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2639 * a BPF program, do not update the **bpf_spin_lock** field. 2640 * * **bpf_spin_lock** cannot be on the stack or inside a 2641 * networking packet (it can only be inside of a map values). 2642 * * **bpf_spin_lock** is available to root only. 2643 * * Tracing programs and socket filter programs cannot use 2644 * **bpf_spin_lock**\ () due to insufficient preemption checks 2645 * (but this may change in the future). 2646 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2647 * Return 2648 * 0 2649 * 2650 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 2651 * Description 2652 * Release the *lock* previously locked by a call to 2653 * **bpf_spin_lock**\ (\ *lock*\ ). 2654 * Return 2655 * 0 2656 * 2657 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2658 * Description 2659 * This helper gets a **struct bpf_sock** pointer such 2660 * that all the fields in this **bpf_sock** can be accessed. 2661 * Return 2662 * A **struct bpf_sock** pointer on success, or **NULL** in 2663 * case of failure. 2664 * 2665 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2666 * Description 2667 * This helper gets a **struct bpf_tcp_sock** pointer from a 2668 * **struct bpf_sock** pointer. 2669 * Return 2670 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2671 * case of failure. 2672 * 2673 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 2674 * Description 2675 * Set ECN (Explicit Congestion Notification) field of IP header 2676 * to **CE** (Congestion Encountered) if current value is **ECT** 2677 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2678 * and IPv4. 2679 * Return 2680 * 1 if the **CE** flag is set (either by the current helper call 2681 * or because it was already present), 0 if it is not set. 2682 * 2683 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2684 * Description 2685 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2686 * **bpf_sk_release**\ () is unnecessary and not allowed. 2687 * Return 2688 * A **struct bpf_sock** pointer on success, or **NULL** in 2689 * case of failure. 2690 * 2691 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2692 * Description 2693 * Look for TCP socket matching *tuple*, optionally in a child 2694 * network namespace *netns*. The return value must be checked, 2695 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2696 * 2697 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2698 * that it also returns timewait or request sockets. Use 2699 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2700 * full structure. 2701 * 2702 * This helper is available only if the kernel was compiled with 2703 * **CONFIG_NET** configuration option. 2704 * Return 2705 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2706 * For sockets with reuseport option, the **struct bpf_sock** 2707 * result is from *reuse*\ **->socks**\ [] using the hash of the 2708 * tuple. 2709 * 2710 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2711 * Description 2712 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2713 * the listening socket in *sk*. 2714 * 2715 * *iph* points to the start of the IPv4 or IPv6 header, while 2716 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2717 * **sizeof**\ (**struct ip6hdr**). 2718 * 2719 * *th* points to the start of the TCP header, while *th_len* 2720 * contains **sizeof**\ (**struct tcphdr**). 2721 * Return 2722 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2723 * error otherwise. 2724 * 2725 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2726 * Description 2727 * Get name of sysctl in /proc/sys/ and copy it into provided by 2728 * program buffer *buf* of size *buf_len*. 2729 * 2730 * The buffer is always NUL terminated, unless it's zero-sized. 2731 * 2732 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2733 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2734 * only (e.g. "tcp_mem"). 2735 * Return 2736 * Number of character copied (not including the trailing NUL). 2737 * 2738 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2739 * truncated name in this case). 2740 * 2741 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2742 * Description 2743 * Get current value of sysctl as it is presented in /proc/sys 2744 * (incl. newline, etc), and copy it as a string into provided 2745 * by program buffer *buf* of size *buf_len*. 2746 * 2747 * The whole value is copied, no matter what file position user 2748 * space issued e.g. sys_read at. 2749 * 2750 * The buffer is always NUL terminated, unless it's zero-sized. 2751 * Return 2752 * Number of character copied (not including the trailing NUL). 2753 * 2754 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2755 * truncated name in this case). 2756 * 2757 * **-EINVAL** if current value was unavailable, e.g. because 2758 * sysctl is uninitialized and read returns -EIO for it. 2759 * 2760 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2761 * Description 2762 * Get new value being written by user space to sysctl (before 2763 * the actual write happens) and copy it as a string into 2764 * provided by program buffer *buf* of size *buf_len*. 2765 * 2766 * User space may write new value at file position > 0. 2767 * 2768 * The buffer is always NUL terminated, unless it's zero-sized. 2769 * Return 2770 * Number of character copied (not including the trailing NUL). 2771 * 2772 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2773 * truncated name in this case). 2774 * 2775 * **-EINVAL** if sysctl is being read. 2776 * 2777 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2778 * Description 2779 * Override new value being written by user space to sysctl with 2780 * value provided by program in buffer *buf* of size *buf_len*. 2781 * 2782 * *buf* should contain a string in same form as provided by user 2783 * space on sysctl write. 2784 * 2785 * User space may write new value at file position > 0. To override 2786 * the whole sysctl value file position should be set to zero. 2787 * Return 2788 * 0 on success. 2789 * 2790 * **-E2BIG** if the *buf_len* is too big. 2791 * 2792 * **-EINVAL** if sysctl is being read. 2793 * 2794 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2795 * Description 2796 * Convert the initial part of the string from buffer *buf* of 2797 * size *buf_len* to a long integer according to the given base 2798 * and save the result in *res*. 2799 * 2800 * The string may begin with an arbitrary amount of white space 2801 * (as determined by **isspace**\ (3)) followed by a single 2802 * optional '**-**' sign. 2803 * 2804 * Five least significant bits of *flags* encode base, other bits 2805 * are currently unused. 2806 * 2807 * Base must be either 8, 10, 16 or 0 to detect it automatically 2808 * similar to user space **strtol**\ (3). 2809 * Return 2810 * Number of characters consumed on success. Must be positive but 2811 * no more than *buf_len*. 2812 * 2813 * **-EINVAL** if no valid digits were found or unsupported base 2814 * was provided. 2815 * 2816 * **-ERANGE** if resulting value was out of range. 2817 * 2818 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2819 * Description 2820 * Convert the initial part of the string from buffer *buf* of 2821 * size *buf_len* to an unsigned long integer according to the 2822 * given base and save the result in *res*. 2823 * 2824 * The string may begin with an arbitrary amount of white space 2825 * (as determined by **isspace**\ (3)). 2826 * 2827 * Five least significant bits of *flags* encode base, other bits 2828 * are currently unused. 2829 * 2830 * Base must be either 8, 10, 16 or 0 to detect it automatically 2831 * similar to user space **strtoul**\ (3). 2832 * Return 2833 * Number of characters consumed on success. Must be positive but 2834 * no more than *buf_len*. 2835 * 2836 * **-EINVAL** if no valid digits were found or unsupported base 2837 * was provided. 2838 * 2839 * **-ERANGE** if resulting value was out of range. 2840 * 2841 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 2842 * Description 2843 * Get a bpf-local-storage from a *sk*. 2844 * 2845 * Logically, it could be thought of getting the value from 2846 * a *map* with *sk* as the **key**. From this 2847 * perspective, the usage is not much different from 2848 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2849 * helper enforces the key must be a full socket and the map must 2850 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2851 * 2852 * Underneath, the value is stored locally at *sk* instead of 2853 * the *map*. The *map* is used as the bpf-local-storage 2854 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2855 * searched against all bpf-local-storages residing at *sk*. 2856 * 2857 * *sk* is a kernel **struct sock** pointer for LSM program. 2858 * *sk* is a **struct bpf_sock** pointer for other program types. 2859 * 2860 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2861 * used such that a new bpf-local-storage will be 2862 * created if one does not exist. *value* can be used 2863 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2864 * the initial value of a bpf-local-storage. If *value* is 2865 * **NULL**, the new bpf-local-storage will be zero initialized. 2866 * Return 2867 * A bpf-local-storage pointer is returned on success. 2868 * 2869 * **NULL** if not found or there was an error in adding 2870 * a new bpf-local-storage. 2871 * 2872 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 2873 * Description 2874 * Delete a bpf-local-storage from a *sk*. 2875 * Return 2876 * 0 on success. 2877 * 2878 * **-ENOENT** if the bpf-local-storage cannot be found. 2879 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 2880 * 2881 * long bpf_send_signal(u32 sig) 2882 * Description 2883 * Send signal *sig* to the process of the current task. 2884 * The signal may be delivered to any of this process's threads. 2885 * Return 2886 * 0 on success or successfully queued. 2887 * 2888 * **-EBUSY** if work queue under nmi is full. 2889 * 2890 * **-EINVAL** if *sig* is invalid. 2891 * 2892 * **-EPERM** if no permission to send the *sig*. 2893 * 2894 * **-EAGAIN** if bpf program can try again. 2895 * 2896 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2897 * Description 2898 * Try to issue a SYN cookie for the packet with corresponding 2899 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2900 * 2901 * *iph* points to the start of the IPv4 or IPv6 header, while 2902 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2903 * **sizeof**\ (**struct ip6hdr**). 2904 * 2905 * *th* points to the start of the TCP header, while *th_len* 2906 * contains the length of the TCP header. 2907 * Return 2908 * On success, lower 32 bits hold the generated SYN cookie in 2909 * followed by 16 bits which hold the MSS value for that cookie, 2910 * and the top 16 bits are unused. 2911 * 2912 * On failure, the returned value is one of the following: 2913 * 2914 * **-EINVAL** SYN cookie cannot be issued due to error 2915 * 2916 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2917 * 2918 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2919 * 2920 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2921 * 2922 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2923 * Description 2924 * Write raw *data* blob into a special BPF perf event held by 2925 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2926 * event must have the following attributes: **PERF_SAMPLE_RAW** 2927 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2928 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2929 * 2930 * The *flags* are used to indicate the index in *map* for which 2931 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2932 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2933 * to indicate that the index of the current CPU core should be 2934 * used. 2935 * 2936 * The value to write, of *size*, is passed through eBPF stack and 2937 * pointed by *data*. 2938 * 2939 * *ctx* is a pointer to in-kernel struct sk_buff. 2940 * 2941 * This helper is similar to **bpf_perf_event_output**\ () but 2942 * restricted to raw_tracepoint bpf programs. 2943 * Return 2944 * 0 on success, or a negative error in case of failure. 2945 * 2946 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2947 * Description 2948 * Safely attempt to read *size* bytes from user space address 2949 * *unsafe_ptr* and store the data in *dst*. 2950 * Return 2951 * 0 on success, or a negative error in case of failure. 2952 * 2953 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2954 * Description 2955 * Safely attempt to read *size* bytes from kernel space address 2956 * *unsafe_ptr* and store the data in *dst*. 2957 * Return 2958 * 0 on success, or a negative error in case of failure. 2959 * 2960 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2961 * Description 2962 * Copy a NUL terminated string from an unsafe user address 2963 * *unsafe_ptr* to *dst*. The *size* should include the 2964 * terminating NUL byte. In case the string length is smaller than 2965 * *size*, the target is not padded with further NUL bytes. If the 2966 * string length is larger than *size*, just *size*-1 bytes are 2967 * copied and the last byte is set to NUL. 2968 * 2969 * On success, the length of the copied string is returned. This 2970 * makes this helper useful in tracing programs for reading 2971 * strings, and more importantly to get its length at runtime. See 2972 * the following snippet: 2973 * 2974 * :: 2975 * 2976 * SEC("kprobe/sys_open") 2977 * void bpf_sys_open(struct pt_regs *ctx) 2978 * { 2979 * char buf[PATHLEN]; // PATHLEN is defined to 256 2980 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2981 * ctx->di); 2982 * 2983 * // Consume buf, for example push it to 2984 * // userspace via bpf_perf_event_output(); we 2985 * // can use res (the string length) as event 2986 * // size, after checking its boundaries. 2987 * } 2988 * 2989 * In comparison, using **bpf_probe_read_user**\ () helper here 2990 * instead to read the string would require to estimate the length 2991 * at compile time, and would often result in copying more memory 2992 * than necessary. 2993 * 2994 * Another useful use case is when parsing individual process 2995 * arguments or individual environment variables navigating 2996 * *current*\ **->mm->arg_start** and *current*\ 2997 * **->mm->env_start**: using this helper and the return value, 2998 * one can quickly iterate at the right offset of the memory area. 2999 * Return 3000 * On success, the strictly positive length of the string, 3001 * including the trailing NUL character. On error, a negative 3002 * value. 3003 * 3004 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3005 * Description 3006 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3007 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3008 * Return 3009 * On success, the strictly positive length of the string, including 3010 * the trailing NUL character. On error, a negative value. 3011 * 3012 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3013 * Description 3014 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3015 * *rcv_nxt* is the ack_seq to be sent out. 3016 * Return 3017 * 0 on success, or a negative error in case of failure. 3018 * 3019 * long bpf_send_signal_thread(u32 sig) 3020 * Description 3021 * Send signal *sig* to the thread corresponding to the current task. 3022 * Return 3023 * 0 on success or successfully queued. 3024 * 3025 * **-EBUSY** if work queue under nmi is full. 3026 * 3027 * **-EINVAL** if *sig* is invalid. 3028 * 3029 * **-EPERM** if no permission to send the *sig*. 3030 * 3031 * **-EAGAIN** if bpf program can try again. 3032 * 3033 * u64 bpf_jiffies64(void) 3034 * Description 3035 * Obtain the 64bit jiffies 3036 * Return 3037 * The 64 bit jiffies 3038 * 3039 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3040 * Description 3041 * For an eBPF program attached to a perf event, retrieve the 3042 * branch records (**struct perf_branch_entry**) associated to *ctx* 3043 * and store it in the buffer pointed by *buf* up to size 3044 * *size* bytes. 3045 * Return 3046 * On success, number of bytes written to *buf*. On error, a 3047 * negative value. 3048 * 3049 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3050 * instead return the number of bytes required to store all the 3051 * branch entries. If this flag is set, *buf* may be NULL. 3052 * 3053 * **-EINVAL** if arguments invalid or **size** not a multiple 3054 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3055 * 3056 * **-ENOENT** if architecture does not support branch records. 3057 * 3058 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3059 * Description 3060 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3061 * *namespace* will be returned in *nsdata*. 3062 * Return 3063 * 0 on success, or one of the following in case of failure: 3064 * 3065 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3066 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3067 * 3068 * **-ENOENT** if pidns does not exists for the current task. 3069 * 3070 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3071 * Description 3072 * Write raw *data* blob into a special BPF perf event held by 3073 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3074 * event must have the following attributes: **PERF_SAMPLE_RAW** 3075 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3076 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3077 * 3078 * The *flags* are used to indicate the index in *map* for which 3079 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3080 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3081 * to indicate that the index of the current CPU core should be 3082 * used. 3083 * 3084 * The value to write, of *size*, is passed through eBPF stack and 3085 * pointed by *data*. 3086 * 3087 * *ctx* is a pointer to in-kernel struct xdp_buff. 3088 * 3089 * This helper is similar to **bpf_perf_eventoutput**\ () but 3090 * restricted to raw_tracepoint bpf programs. 3091 * Return 3092 * 0 on success, or a negative error in case of failure. 3093 * 3094 * u64 bpf_get_netns_cookie(void *ctx) 3095 * Description 3096 * Retrieve the cookie (generated by the kernel) of the network 3097 * namespace the input *ctx* is associated with. The network 3098 * namespace cookie remains stable for its lifetime and provides 3099 * a global identifier that can be assumed unique. If *ctx* is 3100 * NULL, then the helper returns the cookie for the initial 3101 * network namespace. The cookie itself is very similar to that 3102 * of **bpf_get_socket_cookie**\ () helper, but for network 3103 * namespaces instead of sockets. 3104 * Return 3105 * A 8-byte long opaque number. 3106 * 3107 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3108 * Description 3109 * Return id of cgroup v2 that is ancestor of the cgroup associated 3110 * with the current task at the *ancestor_level*. The root cgroup 3111 * is at *ancestor_level* zero and each step down the hierarchy 3112 * increments the level. If *ancestor_level* == level of cgroup 3113 * associated with the current task, then return value will be the 3114 * same as that of **bpf_get_current_cgroup_id**\ (). 3115 * 3116 * The helper is useful to implement policies based on cgroups 3117 * that are upper in hierarchy than immediate cgroup associated 3118 * with the current task. 3119 * 3120 * The format of returned id and helper limitations are same as in 3121 * **bpf_get_current_cgroup_id**\ (). 3122 * Return 3123 * The id is returned or 0 in case the id could not be retrieved. 3124 * 3125 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3126 * Description 3127 * Helper is overloaded depending on BPF program type. This 3128 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3129 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3130 * 3131 * Assign the *sk* to the *skb*. When combined with appropriate 3132 * routing configuration to receive the packet towards the socket, 3133 * will cause *skb* to be delivered to the specified socket. 3134 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3135 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3136 * interfere with successful delivery to the socket. 3137 * 3138 * This operation is only valid from TC ingress path. 3139 * 3140 * The *flags* argument must be zero. 3141 * Return 3142 * 0 on success, or a negative error in case of failure: 3143 * 3144 * **-EINVAL** if specified *flags* are not supported. 3145 * 3146 * **-ENOENT** if the socket is unavailable for assignment. 3147 * 3148 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3149 * 3150 * **-EOPNOTSUPP** if the operation is not supported, for example 3151 * a call from outside of TC ingress. 3152 * 3153 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3154 * (reuseport). 3155 * 3156 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3157 * Description 3158 * Helper is overloaded depending on BPF program type. This 3159 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3160 * 3161 * Select the *sk* as a result of a socket lookup. 3162 * 3163 * For the operation to succeed passed socket must be compatible 3164 * with the packet description provided by the *ctx* object. 3165 * 3166 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3167 * be an exact match. While IP family (**AF_INET** or 3168 * **AF_INET6**) must be compatible, that is IPv6 sockets 3169 * that are not v6-only can be selected for IPv4 packets. 3170 * 3171 * Only TCP listeners and UDP unconnected sockets can be 3172 * selected. *sk* can also be NULL to reset any previous 3173 * selection. 3174 * 3175 * *flags* argument can combination of following values: 3176 * 3177 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3178 * socket selection, potentially done by a BPF program 3179 * that ran before us. 3180 * 3181 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3182 * load-balancing within reuseport group for the socket 3183 * being selected. 3184 * 3185 * On success *ctx->sk* will point to the selected socket. 3186 * 3187 * Return 3188 * 0 on success, or a negative errno in case of failure. 3189 * 3190 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3191 * not compatible with packet family (*ctx->family*). 3192 * 3193 * * **-EEXIST** if socket has been already selected, 3194 * potentially by another program, and 3195 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3196 * 3197 * * **-EINVAL** if unsupported flags were specified. 3198 * 3199 * * **-EPROTOTYPE** if socket L4 protocol 3200 * (*sk->protocol*) doesn't match packet protocol 3201 * (*ctx->protocol*). 3202 * 3203 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3204 * state (TCP listening or UDP unconnected). 3205 * 3206 * u64 bpf_ktime_get_boot_ns(void) 3207 * Description 3208 * Return the time elapsed since system boot, in nanoseconds. 3209 * Does include the time the system was suspended. 3210 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3211 * Return 3212 * Current *ktime*. 3213 * 3214 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3215 * Description 3216 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3217 * out the format string. 3218 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3219 * the format string itself. The *data* and *data_len* are format string 3220 * arguments. The *data* are a **u64** array and corresponding format string 3221 * values are stored in the array. For strings and pointers where pointees 3222 * are accessed, only the pointer values are stored in the *data* array. 3223 * The *data_len* is the size of *data* in bytes. 3224 * 3225 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3226 * Reading kernel memory may fail due to either invalid address or 3227 * valid address but requiring a major memory fault. If reading kernel memory 3228 * fails, the string for **%s** will be an empty string, and the ip 3229 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3230 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3231 * Return 3232 * 0 on success, or a negative error in case of failure: 3233 * 3234 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3235 * by returning 1 from bpf program. 3236 * 3237 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3238 * 3239 * **-E2BIG** if *fmt* contains too many format specifiers. 3240 * 3241 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3242 * 3243 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 3244 * Description 3245 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3246 * The *m* represents the seq_file. The *data* and *len* represent the 3247 * data to write in bytes. 3248 * Return 3249 * 0 on success, or a negative error in case of failure: 3250 * 3251 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3252 * 3253 * u64 bpf_sk_cgroup_id(void *sk) 3254 * Description 3255 * Return the cgroup v2 id of the socket *sk*. 3256 * 3257 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 3258 * returned from **bpf_sk_lookup_xxx**\ (), 3259 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3260 * same as in **bpf_skb_cgroup_id**\ (). 3261 * 3262 * This helper is available only if the kernel was compiled with 3263 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3264 * Return 3265 * The id is returned or 0 in case the id could not be retrieved. 3266 * 3267 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 3268 * Description 3269 * Return id of cgroup v2 that is ancestor of cgroup associated 3270 * with the *sk* at the *ancestor_level*. The root cgroup is at 3271 * *ancestor_level* zero and each step down the hierarchy 3272 * increments the level. If *ancestor_level* == level of cgroup 3273 * associated with *sk*, then return value will be same as that 3274 * of **bpf_sk_cgroup_id**\ (). 3275 * 3276 * The helper is useful to implement policies based on cgroups 3277 * that are upper in hierarchy than immediate cgroup associated 3278 * with *sk*. 3279 * 3280 * The format of returned id and helper limitations are same as in 3281 * **bpf_sk_cgroup_id**\ (). 3282 * Return 3283 * The id is returned or 0 in case the id could not be retrieved. 3284 * 3285 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 3286 * Description 3287 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3288 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3289 * of new data availability is sent. 3290 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3291 * of new data availability is sent unconditionally. 3292 * Return 3293 * 0 on success, or a negative error in case of failure. 3294 * 3295 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 3296 * Description 3297 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3298 * Return 3299 * Valid pointer with *size* bytes of memory available; NULL, 3300 * otherwise. 3301 * 3302 * void bpf_ringbuf_submit(void *data, u64 flags) 3303 * Description 3304 * Submit reserved ring buffer sample, pointed to by *data*. 3305 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3306 * of new data availability is sent. 3307 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3308 * of new data availability is sent unconditionally. 3309 * Return 3310 * Nothing. Always succeeds. 3311 * 3312 * void bpf_ringbuf_discard(void *data, u64 flags) 3313 * Description 3314 * Discard reserved ring buffer sample, pointed to by *data*. 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 * Nothing. Always succeeds. 3321 * 3322 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 3323 * Description 3324 * Query various characteristics of provided ring buffer. What 3325 * exactly is queries is determined by *flags*: 3326 * 3327 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3328 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3329 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3330 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3331 * 3332 * Data returned is just a momentary snapshot of actual values 3333 * and could be inaccurate, so this facility should be used to 3334 * power heuristics and for reporting, not to make 100% correct 3335 * calculation. 3336 * Return 3337 * Requested value, or 0, if *flags* are not recognized. 3338 * 3339 * long bpf_csum_level(struct sk_buff *skb, u64 level) 3340 * Description 3341 * Change the skbs checksum level by one layer up or down, or 3342 * reset it entirely to none in order to have the stack perform 3343 * checksum validation. The level is applicable to the following 3344 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3345 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3346 * through **bpf_skb_adjust_room**\ () helper with passing in 3347 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3348 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3349 * the UDP header is removed. Similarly, an encap of the latter 3350 * into the former could be accompanied by a helper call to 3351 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3352 * skb is still intended to be processed in higher layers of the 3353 * stack instead of just egressing at tc. 3354 * 3355 * There are three supported level settings at this time: 3356 * 3357 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3358 * with CHECKSUM_UNNECESSARY. 3359 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3360 * with CHECKSUM_UNNECESSARY. 3361 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3362 * sets CHECKSUM_NONE to force checksum validation by the stack. 3363 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3364 * skb->csum_level. 3365 * Return 3366 * 0 on success, or a negative error in case of failure. In the 3367 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3368 * is returned or the error code -EACCES in case the skb is not 3369 * subject to CHECKSUM_UNNECESSARY. 3370 * 3371 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 3372 * Description 3373 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3374 * Return 3375 * *sk* if casting is valid, or **NULL** otherwise. 3376 * 3377 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 3378 * Description 3379 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3380 * Return 3381 * *sk* if casting is valid, or **NULL** otherwise. 3382 * 3383 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 3384 * Description 3385 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3386 * Return 3387 * *sk* if casting is valid, or **NULL** otherwise. 3388 * 3389 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 3390 * Description 3391 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3392 * Return 3393 * *sk* if casting is valid, or **NULL** otherwise. 3394 * 3395 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 3396 * Description 3397 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3398 * Return 3399 * *sk* if casting is valid, or **NULL** otherwise. 3400 * 3401 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 3402 * Description 3403 * Return a user or a kernel stack in bpf program provided buffer. 3404 * To achieve this, the helper needs *task*, which is a valid 3405 * pointer to **struct task_struct**. To store the stacktrace, the 3406 * bpf program provides *buf* with a nonnegative *size*. 3407 * 3408 * The last argument, *flags*, holds the number of stack frames to 3409 * skip (from 0 to 255), masked with 3410 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3411 * the following flags: 3412 * 3413 * **BPF_F_USER_STACK** 3414 * Collect a user space stack instead of a kernel stack. 3415 * **BPF_F_USER_BUILD_ID** 3416 * Collect buildid+offset instead of ips for user stack, 3417 * only valid if **BPF_F_USER_STACK** is also specified. 3418 * 3419 * **bpf_get_task_stack**\ () can collect up to 3420 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3421 * to sufficient large buffer size. Note that 3422 * this limit can be controlled with the **sysctl** program, and 3423 * that it should be manually increased in order to profile long 3424 * user stacks (such as stacks for Java programs). To do so, use: 3425 * 3426 * :: 3427 * 3428 * # sysctl kernel.perf_event_max_stack=<new value> 3429 * Return 3430 * A non-negative value equal to or less than *size* on success, 3431 * or a negative error in case of failure. 3432 * 3433 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 3434 * Description 3435 * Load header option. Support reading a particular TCP header 3436 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 3437 * 3438 * If *flags* is 0, it will search the option from the 3439 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 3440 * has details on what skb_data contains under different 3441 * *skops*\ **->op**. 3442 * 3443 * The first byte of the *searchby_res* specifies the 3444 * kind that it wants to search. 3445 * 3446 * If the searching kind is an experimental kind 3447 * (i.e. 253 or 254 according to RFC6994). It also 3448 * needs to specify the "magic" which is either 3449 * 2 bytes or 4 bytes. It then also needs to 3450 * specify the size of the magic by using 3451 * the 2nd byte which is "kind-length" of a TCP 3452 * header option and the "kind-length" also 3453 * includes the first 2 bytes "kind" and "kind-length" 3454 * itself as a normal TCP header option also does. 3455 * 3456 * For example, to search experimental kind 254 with 3457 * 2 byte magic 0xeB9F, the searchby_res should be 3458 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 3459 * 3460 * To search for the standard window scale option (3), 3461 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 3462 * Note, kind-length must be 0 for regular option. 3463 * 3464 * Searching for No-Op (0) and End-of-Option-List (1) are 3465 * not supported. 3466 * 3467 * *len* must be at least 2 bytes which is the minimal size 3468 * of a header option. 3469 * 3470 * Supported flags: 3471 * 3472 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 3473 * saved_syn packet or the just-received syn packet. 3474 * 3475 * Return 3476 * > 0 when found, the header option is copied to *searchby_res*. 3477 * The return value is the total length copied. On failure, a 3478 * negative error code is returned: 3479 * 3480 * **-EINVAL** if a parameter is invalid. 3481 * 3482 * **-ENOMSG** if the option is not found. 3483 * 3484 * **-ENOENT** if no syn packet is available when 3485 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 3486 * 3487 * **-ENOSPC** if there is not enough space. Only *len* number of 3488 * bytes are copied. 3489 * 3490 * **-EFAULT** on failure to parse the header options in the 3491 * packet. 3492 * 3493 * **-EPERM** if the helper cannot be used under the current 3494 * *skops*\ **->op**. 3495 * 3496 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 3497 * Description 3498 * Store header option. The data will be copied 3499 * from buffer *from* with length *len* to the TCP header. 3500 * 3501 * The buffer *from* should have the whole option that 3502 * includes the kind, kind-length, and the actual 3503 * option data. The *len* must be at least kind-length 3504 * long. The kind-length does not have to be 4 byte 3505 * aligned. The kernel will take care of the padding 3506 * and setting the 4 bytes aligned value to th->doff. 3507 * 3508 * This helper will check for duplicated option 3509 * by searching the same option in the outgoing skb. 3510 * 3511 * This helper can only be called during 3512 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3513 * 3514 * Return 3515 * 0 on success, or negative error in case of failure: 3516 * 3517 * **-EINVAL** If param is invalid. 3518 * 3519 * **-ENOSPC** if there is not enough space in the header. 3520 * Nothing has been written 3521 * 3522 * **-EEXIST** if the option already exists. 3523 * 3524 * **-EFAULT** on failrue to parse the existing header options. 3525 * 3526 * **-EPERM** if the helper cannot be used under the current 3527 * *skops*\ **->op**. 3528 * 3529 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 3530 * Description 3531 * Reserve *len* bytes for the bpf header option. The 3532 * space will be used by **bpf_store_hdr_opt**\ () later in 3533 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3534 * 3535 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 3536 * the total number of bytes will be reserved. 3537 * 3538 * This helper can only be called during 3539 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 3540 * 3541 * Return 3542 * 0 on success, or negative error in case of failure: 3543 * 3544 * **-EINVAL** if a parameter is invalid. 3545 * 3546 * **-ENOSPC** if there is not enough space in the header. 3547 * 3548 * **-EPERM** if the helper cannot be used under the current 3549 * *skops*\ **->op**. 3550 * 3551 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 3552 * Description 3553 * Get a bpf_local_storage from an *inode*. 3554 * 3555 * Logically, it could be thought of as getting the value from 3556 * a *map* with *inode* as the **key**. From this 3557 * perspective, the usage is not much different from 3558 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 3559 * helper enforces the key must be an inode and the map must also 3560 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 3561 * 3562 * Underneath, the value is stored locally at *inode* instead of 3563 * the *map*. The *map* is used as the bpf-local-storage 3564 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3565 * searched against all bpf_local_storage residing at *inode*. 3566 * 3567 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3568 * used such that a new bpf_local_storage will be 3569 * created if one does not exist. *value* can be used 3570 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3571 * the initial value of a bpf_local_storage. If *value* is 3572 * **NULL**, the new bpf_local_storage will be zero initialized. 3573 * Return 3574 * A bpf_local_storage pointer is returned on success. 3575 * 3576 * **NULL** if not found or there was an error in adding 3577 * a new bpf_local_storage. 3578 * 3579 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 3580 * Description 3581 * Delete a bpf_local_storage from an *inode*. 3582 * Return 3583 * 0 on success. 3584 * 3585 * **-ENOENT** if the bpf_local_storage cannot be found. 3586 * 3587 * long bpf_d_path(struct path *path, char *buf, u32 sz) 3588 * Description 3589 * Return full path for given **struct path** object, which 3590 * needs to be the kernel BTF *path* object. The path is 3591 * returned in the provided buffer *buf* of size *sz* and 3592 * is zero terminated. 3593 * 3594 * Return 3595 * On success, the strictly positive length of the string, 3596 * including the trailing NUL character. On error, a negative 3597 * value. 3598 * 3599 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 3600 * Description 3601 * Read *size* bytes from user space address *user_ptr* and store 3602 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 3603 * Return 3604 * 0 on success, or a negative error in case of failure. 3605 * 3606 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 3607 * Description 3608 * Use BTF to store a string representation of *ptr*->ptr in *str*, 3609 * using *ptr*->type_id. This value should specify the type 3610 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 3611 * can be used to look up vmlinux BTF type ids. Traversing the 3612 * data structure using BTF, the type information and values are 3613 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 3614 * the pointer data is carried out to avoid kernel crashes during 3615 * operation. Smaller types can use string space on the stack; 3616 * larger programs can use map data to store the string 3617 * representation. 3618 * 3619 * The string can be subsequently shared with userspace via 3620 * bpf_perf_event_output() or ring buffer interfaces. 3621 * bpf_trace_printk() is to be avoided as it places too small 3622 * a limit on string size to be useful. 3623 * 3624 * *flags* is a combination of 3625 * 3626 * **BTF_F_COMPACT** 3627 * no formatting around type information 3628 * **BTF_F_NONAME** 3629 * no struct/union member names/types 3630 * **BTF_F_PTR_RAW** 3631 * show raw (unobfuscated) pointer values; 3632 * equivalent to printk specifier %px. 3633 * **BTF_F_ZERO** 3634 * show zero-valued struct/union members; they 3635 * are not displayed by default 3636 * 3637 * Return 3638 * The number of bytes that were written (or would have been 3639 * written if output had to be truncated due to string size), 3640 * or a negative error in cases of failure. 3641 * 3642 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 3643 * Description 3644 * Use BTF to write to seq_write a string representation of 3645 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 3646 * *flags* are identical to those used for bpf_snprintf_btf. 3647 * Return 3648 * 0 on success or a negative error in case of failure. 3649 * 3650 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 3651 * Description 3652 * See **bpf_get_cgroup_classid**\ () for the main description. 3653 * This helper differs from **bpf_get_cgroup_classid**\ () in that 3654 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 3655 * associated socket instead of the current process. 3656 * Return 3657 * The id is returned or 0 in case the id could not be retrieved. 3658 * 3659 * long bpf_redirect_neigh(u32 ifindex, u64 flags) 3660 * Description 3661 * Redirect the packet to another net device of index *ifindex* 3662 * and fill in L2 addresses from neighboring subsystem. This helper 3663 * is somewhat similar to **bpf_redirect**\ (), except that it 3664 * fills in e.g. MAC addresses based on the L3 information from 3665 * the packet. This helper is supported for IPv4 and IPv6 protocols. 3666 * The *flags* argument is reserved and must be 0. The helper is 3667 * currently only supported for tc BPF program types. 3668 * Return 3669 * The helper returns **TC_ACT_REDIRECT** on success or 3670 * **TC_ACT_SHOT** on error. 3671 */ 3672 #define __BPF_FUNC_MAPPER(FN) \ 3673 FN(unspec), \ 3674 FN(map_lookup_elem), \ 3675 FN(map_update_elem), \ 3676 FN(map_delete_elem), \ 3677 FN(probe_read), \ 3678 FN(ktime_get_ns), \ 3679 FN(trace_printk), \ 3680 FN(get_prandom_u32), \ 3681 FN(get_smp_processor_id), \ 3682 FN(skb_store_bytes), \ 3683 FN(l3_csum_replace), \ 3684 FN(l4_csum_replace), \ 3685 FN(tail_call), \ 3686 FN(clone_redirect), \ 3687 FN(get_current_pid_tgid), \ 3688 FN(get_current_uid_gid), \ 3689 FN(get_current_comm), \ 3690 FN(get_cgroup_classid), \ 3691 FN(skb_vlan_push), \ 3692 FN(skb_vlan_pop), \ 3693 FN(skb_get_tunnel_key), \ 3694 FN(skb_set_tunnel_key), \ 3695 FN(perf_event_read), \ 3696 FN(redirect), \ 3697 FN(get_route_realm), \ 3698 FN(perf_event_output), \ 3699 FN(skb_load_bytes), \ 3700 FN(get_stackid), \ 3701 FN(csum_diff), \ 3702 FN(skb_get_tunnel_opt), \ 3703 FN(skb_set_tunnel_opt), \ 3704 FN(skb_change_proto), \ 3705 FN(skb_change_type), \ 3706 FN(skb_under_cgroup), \ 3707 FN(get_hash_recalc), \ 3708 FN(get_current_task), \ 3709 FN(probe_write_user), \ 3710 FN(current_task_under_cgroup), \ 3711 FN(skb_change_tail), \ 3712 FN(skb_pull_data), \ 3713 FN(csum_update), \ 3714 FN(set_hash_invalid), \ 3715 FN(get_numa_node_id), \ 3716 FN(skb_change_head), \ 3717 FN(xdp_adjust_head), \ 3718 FN(probe_read_str), \ 3719 FN(get_socket_cookie), \ 3720 FN(get_socket_uid), \ 3721 FN(set_hash), \ 3722 FN(setsockopt), \ 3723 FN(skb_adjust_room), \ 3724 FN(redirect_map), \ 3725 FN(sk_redirect_map), \ 3726 FN(sock_map_update), \ 3727 FN(xdp_adjust_meta), \ 3728 FN(perf_event_read_value), \ 3729 FN(perf_prog_read_value), \ 3730 FN(getsockopt), \ 3731 FN(override_return), \ 3732 FN(sock_ops_cb_flags_set), \ 3733 FN(msg_redirect_map), \ 3734 FN(msg_apply_bytes), \ 3735 FN(msg_cork_bytes), \ 3736 FN(msg_pull_data), \ 3737 FN(bind), \ 3738 FN(xdp_adjust_tail), \ 3739 FN(skb_get_xfrm_state), \ 3740 FN(get_stack), \ 3741 FN(skb_load_bytes_relative), \ 3742 FN(fib_lookup), \ 3743 FN(sock_hash_update), \ 3744 FN(msg_redirect_hash), \ 3745 FN(sk_redirect_hash), \ 3746 FN(lwt_push_encap), \ 3747 FN(lwt_seg6_store_bytes), \ 3748 FN(lwt_seg6_adjust_srh), \ 3749 FN(lwt_seg6_action), \ 3750 FN(rc_repeat), \ 3751 FN(rc_keydown), \ 3752 FN(skb_cgroup_id), \ 3753 FN(get_current_cgroup_id), \ 3754 FN(get_local_storage), \ 3755 FN(sk_select_reuseport), \ 3756 FN(skb_ancestor_cgroup_id), \ 3757 FN(sk_lookup_tcp), \ 3758 FN(sk_lookup_udp), \ 3759 FN(sk_release), \ 3760 FN(map_push_elem), \ 3761 FN(map_pop_elem), \ 3762 FN(map_peek_elem), \ 3763 FN(msg_push_data), \ 3764 FN(msg_pop_data), \ 3765 FN(rc_pointer_rel), \ 3766 FN(spin_lock), \ 3767 FN(spin_unlock), \ 3768 FN(sk_fullsock), \ 3769 FN(tcp_sock), \ 3770 FN(skb_ecn_set_ce), \ 3771 FN(get_listener_sock), \ 3772 FN(skc_lookup_tcp), \ 3773 FN(tcp_check_syncookie), \ 3774 FN(sysctl_get_name), \ 3775 FN(sysctl_get_current_value), \ 3776 FN(sysctl_get_new_value), \ 3777 FN(sysctl_set_new_value), \ 3778 FN(strtol), \ 3779 FN(strtoul), \ 3780 FN(sk_storage_get), \ 3781 FN(sk_storage_delete), \ 3782 FN(send_signal), \ 3783 FN(tcp_gen_syncookie), \ 3784 FN(skb_output), \ 3785 FN(probe_read_user), \ 3786 FN(probe_read_kernel), \ 3787 FN(probe_read_user_str), \ 3788 FN(probe_read_kernel_str), \ 3789 FN(tcp_send_ack), \ 3790 FN(send_signal_thread), \ 3791 FN(jiffies64), \ 3792 FN(read_branch_records), \ 3793 FN(get_ns_current_pid_tgid), \ 3794 FN(xdp_output), \ 3795 FN(get_netns_cookie), \ 3796 FN(get_current_ancestor_cgroup_id), \ 3797 FN(sk_assign), \ 3798 FN(ktime_get_boot_ns), \ 3799 FN(seq_printf), \ 3800 FN(seq_write), \ 3801 FN(sk_cgroup_id), \ 3802 FN(sk_ancestor_cgroup_id), \ 3803 FN(ringbuf_output), \ 3804 FN(ringbuf_reserve), \ 3805 FN(ringbuf_submit), \ 3806 FN(ringbuf_discard), \ 3807 FN(ringbuf_query), \ 3808 FN(csum_level), \ 3809 FN(skc_to_tcp6_sock), \ 3810 FN(skc_to_tcp_sock), \ 3811 FN(skc_to_tcp_timewait_sock), \ 3812 FN(skc_to_tcp_request_sock), \ 3813 FN(skc_to_udp6_sock), \ 3814 FN(get_task_stack), \ 3815 FN(load_hdr_opt), \ 3816 FN(store_hdr_opt), \ 3817 FN(reserve_hdr_opt), \ 3818 FN(inode_storage_get), \ 3819 FN(inode_storage_delete), \ 3820 FN(d_path), \ 3821 FN(copy_from_user), \ 3822 FN(snprintf_btf), \ 3823 FN(seq_printf_btf), \ 3824 FN(skb_cgroup_classid), \ 3825 FN(redirect_neigh), \ 3826 /* */ 3827 3828 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3829 * function eBPF program intends to call 3830 */ 3831 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3832 enum bpf_func_id { 3833 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3834 __BPF_FUNC_MAX_ID, 3835 }; 3836 #undef __BPF_ENUM_FN 3837 3838 /* All flags used by eBPF helper functions, placed here. */ 3839 3840 /* BPF_FUNC_skb_store_bytes flags. */ 3841 enum { 3842 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 3843 BPF_F_INVALIDATE_HASH = (1ULL << 1), 3844 }; 3845 3846 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3847 * First 4 bits are for passing the header field size. 3848 */ 3849 enum { 3850 BPF_F_HDR_FIELD_MASK = 0xfULL, 3851 }; 3852 3853 /* BPF_FUNC_l4_csum_replace flags. */ 3854 enum { 3855 BPF_F_PSEUDO_HDR = (1ULL << 4), 3856 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 3857 BPF_F_MARK_ENFORCE = (1ULL << 6), 3858 }; 3859 3860 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3861 enum { 3862 BPF_F_INGRESS = (1ULL << 0), 3863 }; 3864 3865 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3866 enum { 3867 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 3868 }; 3869 3870 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3871 enum { 3872 BPF_F_SKIP_FIELD_MASK = 0xffULL, 3873 BPF_F_USER_STACK = (1ULL << 8), 3874 /* flags used by BPF_FUNC_get_stackid only. */ 3875 BPF_F_FAST_STACK_CMP = (1ULL << 9), 3876 BPF_F_REUSE_STACKID = (1ULL << 10), 3877 /* flags used by BPF_FUNC_get_stack only. */ 3878 BPF_F_USER_BUILD_ID = (1ULL << 11), 3879 }; 3880 3881 /* BPF_FUNC_skb_set_tunnel_key flags. */ 3882 enum { 3883 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 3884 BPF_F_DONT_FRAGMENT = (1ULL << 2), 3885 BPF_F_SEQ_NUMBER = (1ULL << 3), 3886 }; 3887 3888 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3889 * BPF_FUNC_perf_event_read_value flags. 3890 */ 3891 enum { 3892 BPF_F_INDEX_MASK = 0xffffffffULL, 3893 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 3894 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 3895 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 3896 }; 3897 3898 /* Current network namespace */ 3899 enum { 3900 BPF_F_CURRENT_NETNS = (-1L), 3901 }; 3902 3903 /* BPF_FUNC_csum_level level values. */ 3904 enum { 3905 BPF_CSUM_LEVEL_QUERY, 3906 BPF_CSUM_LEVEL_INC, 3907 BPF_CSUM_LEVEL_DEC, 3908 BPF_CSUM_LEVEL_RESET, 3909 }; 3910 3911 /* BPF_FUNC_skb_adjust_room flags. */ 3912 enum { 3913 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 3914 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 3915 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 3916 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 3917 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 3918 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 3919 }; 3920 3921 enum { 3922 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 3923 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 3924 }; 3925 3926 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3927 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3928 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3929 3930 /* BPF_FUNC_sysctl_get_name flags. */ 3931 enum { 3932 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 3933 }; 3934 3935 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 3936 enum { 3937 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 3938 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 3939 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 3940 */ 3941 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 3942 }; 3943 3944 /* BPF_FUNC_read_branch_records flags. */ 3945 enum { 3946 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 3947 }; 3948 3949 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 3950 * BPF_FUNC_bpf_ringbuf_output flags. 3951 */ 3952 enum { 3953 BPF_RB_NO_WAKEUP = (1ULL << 0), 3954 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 3955 }; 3956 3957 /* BPF_FUNC_bpf_ringbuf_query flags */ 3958 enum { 3959 BPF_RB_AVAIL_DATA = 0, 3960 BPF_RB_RING_SIZE = 1, 3961 BPF_RB_CONS_POS = 2, 3962 BPF_RB_PROD_POS = 3, 3963 }; 3964 3965 /* BPF ring buffer constants */ 3966 enum { 3967 BPF_RINGBUF_BUSY_BIT = (1U << 31), 3968 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 3969 BPF_RINGBUF_HDR_SZ = 8, 3970 }; 3971 3972 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 3973 enum { 3974 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 3975 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 3976 }; 3977 3978 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 3979 enum bpf_adj_room_mode { 3980 BPF_ADJ_ROOM_NET, 3981 BPF_ADJ_ROOM_MAC, 3982 }; 3983 3984 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3985 enum bpf_hdr_start_off { 3986 BPF_HDR_START_MAC, 3987 BPF_HDR_START_NET, 3988 }; 3989 3990 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3991 enum bpf_lwt_encap_mode { 3992 BPF_LWT_ENCAP_SEG6, 3993 BPF_LWT_ENCAP_SEG6_INLINE, 3994 BPF_LWT_ENCAP_IP, 3995 }; 3996 3997 #define __bpf_md_ptr(type, name) \ 3998 union { \ 3999 type name; \ 4000 __u64 :64; \ 4001 } __attribute__((aligned(8))) 4002 4003 /* user accessible mirror of in-kernel sk_buff. 4004 * new fields can only be added to the end of this structure 4005 */ 4006 struct __sk_buff { 4007 __u32 len; 4008 __u32 pkt_type; 4009 __u32 mark; 4010 __u32 queue_mapping; 4011 __u32 protocol; 4012 __u32 vlan_present; 4013 __u32 vlan_tci; 4014 __u32 vlan_proto; 4015 __u32 priority; 4016 __u32 ingress_ifindex; 4017 __u32 ifindex; 4018 __u32 tc_index; 4019 __u32 cb[5]; 4020 __u32 hash; 4021 __u32 tc_classid; 4022 __u32 data; 4023 __u32 data_end; 4024 __u32 napi_id; 4025 4026 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 4027 __u32 family; 4028 __u32 remote_ip4; /* Stored in network byte order */ 4029 __u32 local_ip4; /* Stored in network byte order */ 4030 __u32 remote_ip6[4]; /* Stored in network byte order */ 4031 __u32 local_ip6[4]; /* Stored in network byte order */ 4032 __u32 remote_port; /* Stored in network byte order */ 4033 __u32 local_port; /* stored in host byte order */ 4034 /* ... here. */ 4035 4036 __u32 data_meta; 4037 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 4038 __u64 tstamp; 4039 __u32 wire_len; 4040 __u32 gso_segs; 4041 __bpf_md_ptr(struct bpf_sock *, sk); 4042 __u32 gso_size; 4043 }; 4044 4045 struct bpf_tunnel_key { 4046 __u32 tunnel_id; 4047 union { 4048 __u32 remote_ipv4; 4049 __u32 remote_ipv6[4]; 4050 }; 4051 __u8 tunnel_tos; 4052 __u8 tunnel_ttl; 4053 __u16 tunnel_ext; /* Padding, future use. */ 4054 __u32 tunnel_label; 4055 }; 4056 4057 /* user accessible mirror of in-kernel xfrm_state. 4058 * new fields can only be added to the end of this structure 4059 */ 4060 struct bpf_xfrm_state { 4061 __u32 reqid; 4062 __u32 spi; /* Stored in network byte order */ 4063 __u16 family; 4064 __u16 ext; /* Padding, future use. */ 4065 union { 4066 __u32 remote_ipv4; /* Stored in network byte order */ 4067 __u32 remote_ipv6[4]; /* Stored in network byte order */ 4068 }; 4069 }; 4070 4071 /* Generic BPF return codes which all BPF program types may support. 4072 * The values are binary compatible with their TC_ACT_* counter-part to 4073 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 4074 * programs. 4075 * 4076 * XDP is handled seprately, see XDP_*. 4077 */ 4078 enum bpf_ret_code { 4079 BPF_OK = 0, 4080 /* 1 reserved */ 4081 BPF_DROP = 2, 4082 /* 3-6 reserved */ 4083 BPF_REDIRECT = 7, 4084 /* >127 are reserved for prog type specific return codes. 4085 * 4086 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 4087 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 4088 * changed and should be routed based on its new L3 header. 4089 * (This is an L3 redirect, as opposed to L2 redirect 4090 * represented by BPF_REDIRECT above). 4091 */ 4092 BPF_LWT_REROUTE = 128, 4093 }; 4094 4095 struct bpf_sock { 4096 __u32 bound_dev_if; 4097 __u32 family; 4098 __u32 type; 4099 __u32 protocol; 4100 __u32 mark; 4101 __u32 priority; 4102 /* IP address also allows 1 and 2 bytes access */ 4103 __u32 src_ip4; 4104 __u32 src_ip6[4]; 4105 __u32 src_port; /* host byte order */ 4106 __u32 dst_port; /* network byte order */ 4107 __u32 dst_ip4; 4108 __u32 dst_ip6[4]; 4109 __u32 state; 4110 __s32 rx_queue_mapping; 4111 }; 4112 4113 struct bpf_tcp_sock { 4114 __u32 snd_cwnd; /* Sending congestion window */ 4115 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 4116 __u32 rtt_min; 4117 __u32 snd_ssthresh; /* Slow start size threshold */ 4118 __u32 rcv_nxt; /* What we want to receive next */ 4119 __u32 snd_nxt; /* Next sequence we send */ 4120 __u32 snd_una; /* First byte we want an ack for */ 4121 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 4122 __u32 ecn_flags; /* ECN status bits. */ 4123 __u32 rate_delivered; /* saved rate sample: packets delivered */ 4124 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 4125 __u32 packets_out; /* Packets which are "in flight" */ 4126 __u32 retrans_out; /* Retransmitted packets out */ 4127 __u32 total_retrans; /* Total retransmits for entire connection */ 4128 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 4129 * total number of segments in. 4130 */ 4131 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 4132 * total number of data segments in. 4133 */ 4134 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 4135 * The total number of segments sent. 4136 */ 4137 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 4138 * total number of data segments sent. 4139 */ 4140 __u32 lost_out; /* Lost packets */ 4141 __u32 sacked_out; /* SACK'd packets */ 4142 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 4143 * sum(delta(rcv_nxt)), or how many bytes 4144 * were acked. 4145 */ 4146 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 4147 * sum(delta(snd_una)), or how many bytes 4148 * were acked. 4149 */ 4150 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 4151 * total number of DSACK blocks received 4152 */ 4153 __u32 delivered; /* Total data packets delivered incl. rexmits */ 4154 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 4155 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 4156 }; 4157 4158 struct bpf_sock_tuple { 4159 union { 4160 struct { 4161 __be32 saddr; 4162 __be32 daddr; 4163 __be16 sport; 4164 __be16 dport; 4165 } ipv4; 4166 struct { 4167 __be32 saddr[4]; 4168 __be32 daddr[4]; 4169 __be16 sport; 4170 __be16 dport; 4171 } ipv6; 4172 }; 4173 }; 4174 4175 struct bpf_xdp_sock { 4176 __u32 queue_id; 4177 }; 4178 4179 #define XDP_PACKET_HEADROOM 256 4180 4181 /* User return codes for XDP prog type. 4182 * A valid XDP program must return one of these defined values. All other 4183 * return codes are reserved for future use. Unknown return codes will 4184 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 4185 */ 4186 enum xdp_action { 4187 XDP_ABORTED = 0, 4188 XDP_DROP, 4189 XDP_PASS, 4190 XDP_TX, 4191 XDP_REDIRECT, 4192 }; 4193 4194 /* user accessible metadata for XDP packet hook 4195 * new fields must be added to the end of this structure 4196 */ 4197 struct xdp_md { 4198 __u32 data; 4199 __u32 data_end; 4200 __u32 data_meta; 4201 /* Below access go through struct xdp_rxq_info */ 4202 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 4203 __u32 rx_queue_index; /* rxq->queue_index */ 4204 4205 __u32 egress_ifindex; /* txq->dev->ifindex */ 4206 }; 4207 4208 /* DEVMAP map-value layout 4209 * 4210 * The struct data-layout of map-value is a configuration interface. 4211 * New members can only be added to the end of this structure. 4212 */ 4213 struct bpf_devmap_val { 4214 __u32 ifindex; /* device index */ 4215 union { 4216 int fd; /* prog fd on map write */ 4217 __u32 id; /* prog id on map read */ 4218 } bpf_prog; 4219 }; 4220 4221 /* CPUMAP map-value layout 4222 * 4223 * The struct data-layout of map-value is a configuration interface. 4224 * New members can only be added to the end of this structure. 4225 */ 4226 struct bpf_cpumap_val { 4227 __u32 qsize; /* queue size to remote target CPU */ 4228 union { 4229 int fd; /* prog fd on map write */ 4230 __u32 id; /* prog id on map read */ 4231 } bpf_prog; 4232 }; 4233 4234 enum sk_action { 4235 SK_DROP = 0, 4236 SK_PASS, 4237 }; 4238 4239 /* user accessible metadata for SK_MSG packet hook, new fields must 4240 * be added to the end of this structure 4241 */ 4242 struct sk_msg_md { 4243 __bpf_md_ptr(void *, data); 4244 __bpf_md_ptr(void *, data_end); 4245 4246 __u32 family; 4247 __u32 remote_ip4; /* Stored in network byte order */ 4248 __u32 local_ip4; /* Stored in network byte order */ 4249 __u32 remote_ip6[4]; /* Stored in network byte order */ 4250 __u32 local_ip6[4]; /* Stored in network byte order */ 4251 __u32 remote_port; /* Stored in network byte order */ 4252 __u32 local_port; /* stored in host byte order */ 4253 __u32 size; /* Total size of sk_msg */ 4254 4255 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 4256 }; 4257 4258 struct sk_reuseport_md { 4259 /* 4260 * Start of directly accessible data. It begins from 4261 * the tcp/udp header. 4262 */ 4263 __bpf_md_ptr(void *, data); 4264 /* End of directly accessible data */ 4265 __bpf_md_ptr(void *, data_end); 4266 /* 4267 * Total length of packet (starting from the tcp/udp header). 4268 * Note that the directly accessible bytes (data_end - data) 4269 * could be less than this "len". Those bytes could be 4270 * indirectly read by a helper "bpf_skb_load_bytes()". 4271 */ 4272 __u32 len; 4273 /* 4274 * Eth protocol in the mac header (network byte order). e.g. 4275 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 4276 */ 4277 __u32 eth_protocol; 4278 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 4279 __u32 bind_inany; /* Is sock bound to an INANY address? */ 4280 __u32 hash; /* A hash of the packet 4 tuples */ 4281 }; 4282 4283 #define BPF_TAG_SIZE 8 4284 4285 struct bpf_prog_info { 4286 __u32 type; 4287 __u32 id; 4288 __u8 tag[BPF_TAG_SIZE]; 4289 __u32 jited_prog_len; 4290 __u32 xlated_prog_len; 4291 __aligned_u64 jited_prog_insns; 4292 __aligned_u64 xlated_prog_insns; 4293 __u64 load_time; /* ns since boottime */ 4294 __u32 created_by_uid; 4295 __u32 nr_map_ids; 4296 __aligned_u64 map_ids; 4297 char name[BPF_OBJ_NAME_LEN]; 4298 __u32 ifindex; 4299 __u32 gpl_compatible:1; 4300 __u32 :31; /* alignment pad */ 4301 __u64 netns_dev; 4302 __u64 netns_ino; 4303 __u32 nr_jited_ksyms; 4304 __u32 nr_jited_func_lens; 4305 __aligned_u64 jited_ksyms; 4306 __aligned_u64 jited_func_lens; 4307 __u32 btf_id; 4308 __u32 func_info_rec_size; 4309 __aligned_u64 func_info; 4310 __u32 nr_func_info; 4311 __u32 nr_line_info; 4312 __aligned_u64 line_info; 4313 __aligned_u64 jited_line_info; 4314 __u32 nr_jited_line_info; 4315 __u32 line_info_rec_size; 4316 __u32 jited_line_info_rec_size; 4317 __u32 nr_prog_tags; 4318 __aligned_u64 prog_tags; 4319 __u64 run_time_ns; 4320 __u64 run_cnt; 4321 } __attribute__((aligned(8))); 4322 4323 struct bpf_map_info { 4324 __u32 type; 4325 __u32 id; 4326 __u32 key_size; 4327 __u32 value_size; 4328 __u32 max_entries; 4329 __u32 map_flags; 4330 char name[BPF_OBJ_NAME_LEN]; 4331 __u32 ifindex; 4332 __u32 btf_vmlinux_value_type_id; 4333 __u64 netns_dev; 4334 __u64 netns_ino; 4335 __u32 btf_id; 4336 __u32 btf_key_type_id; 4337 __u32 btf_value_type_id; 4338 } __attribute__((aligned(8))); 4339 4340 struct bpf_btf_info { 4341 __aligned_u64 btf; 4342 __u32 btf_size; 4343 __u32 id; 4344 } __attribute__((aligned(8))); 4345 4346 struct bpf_link_info { 4347 __u32 type; 4348 __u32 id; 4349 __u32 prog_id; 4350 union { 4351 struct { 4352 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 4353 __u32 tp_name_len; /* in/out: tp_name buffer len */ 4354 } raw_tracepoint; 4355 struct { 4356 __u32 attach_type; 4357 } tracing; 4358 struct { 4359 __u64 cgroup_id; 4360 __u32 attach_type; 4361 } cgroup; 4362 struct { 4363 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 4364 __u32 target_name_len; /* in/out: target_name buffer len */ 4365 union { 4366 struct { 4367 __u32 map_id; 4368 } map; 4369 }; 4370 } iter; 4371 struct { 4372 __u32 netns_ino; 4373 __u32 attach_type; 4374 } netns; 4375 struct { 4376 __u32 ifindex; 4377 } xdp; 4378 }; 4379 } __attribute__((aligned(8))); 4380 4381 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 4382 * by user and intended to be used by socket (e.g. to bind to, depends on 4383 * attach type). 4384 */ 4385 struct bpf_sock_addr { 4386 __u32 user_family; /* Allows 4-byte read, but no write. */ 4387 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4388 * Stored in network byte order. 4389 */ 4390 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4391 * Stored in network byte order. 4392 */ 4393 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 4394 * Stored in network byte order 4395 */ 4396 __u32 family; /* Allows 4-byte read, but no write */ 4397 __u32 type; /* Allows 4-byte read, but no write */ 4398 __u32 protocol; /* Allows 4-byte read, but no write */ 4399 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4400 * Stored in network byte order. 4401 */ 4402 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4403 * Stored in network byte order. 4404 */ 4405 __bpf_md_ptr(struct bpf_sock *, sk); 4406 }; 4407 4408 /* User bpf_sock_ops struct to access socket values and specify request ops 4409 * and their replies. 4410 * Some of this fields are in network (bigendian) byte order and may need 4411 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 4412 * New fields can only be added at the end of this structure 4413 */ 4414 struct bpf_sock_ops { 4415 __u32 op; 4416 union { 4417 __u32 args[4]; /* Optionally passed to bpf program */ 4418 __u32 reply; /* Returned by bpf program */ 4419 __u32 replylong[4]; /* Optionally returned by bpf prog */ 4420 }; 4421 __u32 family; 4422 __u32 remote_ip4; /* Stored in network byte order */ 4423 __u32 local_ip4; /* Stored in network byte order */ 4424 __u32 remote_ip6[4]; /* Stored in network byte order */ 4425 __u32 local_ip6[4]; /* Stored in network byte order */ 4426 __u32 remote_port; /* Stored in network byte order */ 4427 __u32 local_port; /* stored in host byte order */ 4428 __u32 is_fullsock; /* Some TCP fields are only valid if 4429 * there is a full socket. If not, the 4430 * fields read as zero. 4431 */ 4432 __u32 snd_cwnd; 4433 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 4434 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 4435 __u32 state; 4436 __u32 rtt_min; 4437 __u32 snd_ssthresh; 4438 __u32 rcv_nxt; 4439 __u32 snd_nxt; 4440 __u32 snd_una; 4441 __u32 mss_cache; 4442 __u32 ecn_flags; 4443 __u32 rate_delivered; 4444 __u32 rate_interval_us; 4445 __u32 packets_out; 4446 __u32 retrans_out; 4447 __u32 total_retrans; 4448 __u32 segs_in; 4449 __u32 data_segs_in; 4450 __u32 segs_out; 4451 __u32 data_segs_out; 4452 __u32 lost_out; 4453 __u32 sacked_out; 4454 __u32 sk_txhash; 4455 __u64 bytes_received; 4456 __u64 bytes_acked; 4457 __bpf_md_ptr(struct bpf_sock *, sk); 4458 /* [skb_data, skb_data_end) covers the whole TCP header. 4459 * 4460 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 4461 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 4462 * header has not been written. 4463 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 4464 * been written so far. 4465 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 4466 * the 3WHS. 4467 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 4468 * the 3WHS. 4469 * 4470 * bpf_load_hdr_opt() can also be used to read a particular option. 4471 */ 4472 __bpf_md_ptr(void *, skb_data); 4473 __bpf_md_ptr(void *, skb_data_end); 4474 __u32 skb_len; /* The total length of a packet. 4475 * It includes the header, options, 4476 * and payload. 4477 */ 4478 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 4479 * an easy way to check for tcp_flags 4480 * without parsing skb_data. 4481 * 4482 * In particular, the skb_tcp_flags 4483 * will still be available in 4484 * BPF_SOCK_OPS_HDR_OPT_LEN even though 4485 * the outgoing header has not 4486 * been written yet. 4487 */ 4488 }; 4489 4490 /* Definitions for bpf_sock_ops_cb_flags */ 4491 enum { 4492 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 4493 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 4494 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 4495 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 4496 /* Call bpf for all received TCP headers. The bpf prog will be 4497 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4498 * 4499 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4500 * for the header option related helpers that will be useful 4501 * to the bpf programs. 4502 * 4503 * It could be used at the client/active side (i.e. connect() side) 4504 * when the server told it that the server was in syncookie 4505 * mode and required the active side to resend the bpf-written 4506 * options. The active side can keep writing the bpf-options until 4507 * it received a valid packet from the server side to confirm 4508 * the earlier packet (and options) has been received. The later 4509 * example patch is using it like this at the active side when the 4510 * server is in syncookie mode. 4511 * 4512 * The bpf prog will usually turn this off in the common cases. 4513 */ 4514 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 4515 /* Call bpf when kernel has received a header option that 4516 * the kernel cannot handle. The bpf prog will be called under 4517 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 4518 * 4519 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 4520 * for the header option related helpers that will be useful 4521 * to the bpf programs. 4522 */ 4523 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 4524 /* Call bpf when the kernel is writing header options for the 4525 * outgoing packet. The bpf prog will first be called 4526 * to reserve space in a skb under 4527 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 4528 * the bpf prog will be called to write the header option(s) 4529 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4530 * 4531 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 4532 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 4533 * related helpers that will be useful to the bpf programs. 4534 * 4535 * The kernel gets its chance to reserve space and write 4536 * options first before the BPF program does. 4537 */ 4538 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 4539 /* Mask of all currently supported cb flags */ 4540 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 4541 }; 4542 4543 /* List of known BPF sock_ops operators. 4544 * New entries can only be added at the end 4545 */ 4546 enum { 4547 BPF_SOCK_OPS_VOID, 4548 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 4549 * -1 if default value should be used 4550 */ 4551 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 4552 * window (in packets) or -1 if default 4553 * value should be used 4554 */ 4555 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 4556 * active connection is initialized 4557 */ 4558 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 4559 * active connection is 4560 * established 4561 */ 4562 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 4563 * passive connection is 4564 * established 4565 */ 4566 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 4567 * needs ECN 4568 */ 4569 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 4570 * based on the path and may be 4571 * dependent on the congestion control 4572 * algorithm. In general it indicates 4573 * a congestion threshold. RTTs above 4574 * this indicate congestion 4575 */ 4576 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 4577 * Arg1: value of icsk_retransmits 4578 * Arg2: value of icsk_rto 4579 * Arg3: whether RTO has expired 4580 */ 4581 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 4582 * Arg1: sequence number of 1st byte 4583 * Arg2: # segments 4584 * Arg3: return value of 4585 * tcp_transmit_skb (0 => success) 4586 */ 4587 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 4588 * Arg1: old_state 4589 * Arg2: new_state 4590 */ 4591 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 4592 * socket transition to LISTEN state. 4593 */ 4594 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 4595 */ 4596 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 4597 * It will be called to handle 4598 * the packets received at 4599 * an already established 4600 * connection. 4601 * 4602 * sock_ops->skb_data: 4603 * Referring to the received skb. 4604 * It covers the TCP header only. 4605 * 4606 * bpf_load_hdr_opt() can also 4607 * be used to search for a 4608 * particular option. 4609 */ 4610 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 4611 * header option later in 4612 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4613 * Arg1: bool want_cookie. (in 4614 * writing SYNACK only) 4615 * 4616 * sock_ops->skb_data: 4617 * Not available because no header has 4618 * been written yet. 4619 * 4620 * sock_ops->skb_tcp_flags: 4621 * The tcp_flags of the 4622 * outgoing skb. (e.g. SYN, ACK, FIN). 4623 * 4624 * bpf_reserve_hdr_opt() should 4625 * be used to reserve space. 4626 */ 4627 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 4628 * Arg1: bool want_cookie. (in 4629 * writing SYNACK only) 4630 * 4631 * sock_ops->skb_data: 4632 * Referring to the outgoing skb. 4633 * It covers the TCP header 4634 * that has already been written 4635 * by the kernel and the 4636 * earlier bpf-progs. 4637 * 4638 * sock_ops->skb_tcp_flags: 4639 * The tcp_flags of the outgoing 4640 * skb. (e.g. SYN, ACK, FIN). 4641 * 4642 * bpf_store_hdr_opt() should 4643 * be used to write the 4644 * option. 4645 * 4646 * bpf_load_hdr_opt() can also 4647 * be used to search for a 4648 * particular option that 4649 * has already been written 4650 * by the kernel or the 4651 * earlier bpf-progs. 4652 */ 4653 }; 4654 4655 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 4656 * changes between the TCP and BPF versions. Ideally this should never happen. 4657 * If it does, we need to add code to convert them before calling 4658 * the BPF sock_ops function. 4659 */ 4660 enum { 4661 BPF_TCP_ESTABLISHED = 1, 4662 BPF_TCP_SYN_SENT, 4663 BPF_TCP_SYN_RECV, 4664 BPF_TCP_FIN_WAIT1, 4665 BPF_TCP_FIN_WAIT2, 4666 BPF_TCP_TIME_WAIT, 4667 BPF_TCP_CLOSE, 4668 BPF_TCP_CLOSE_WAIT, 4669 BPF_TCP_LAST_ACK, 4670 BPF_TCP_LISTEN, 4671 BPF_TCP_CLOSING, /* Now a valid state */ 4672 BPF_TCP_NEW_SYN_RECV, 4673 4674 BPF_TCP_MAX_STATES /* Leave at the end! */ 4675 }; 4676 4677 enum { 4678 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 4679 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 4680 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 4681 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 4682 /* Copy the SYN pkt to optval 4683 * 4684 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 4685 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 4686 * to only getting from the saved_syn. It can either get the 4687 * syn packet from: 4688 * 4689 * 1. the just-received SYN packet (only available when writing the 4690 * SYNACK). It will be useful when it is not necessary to 4691 * save the SYN packet for latter use. It is also the only way 4692 * to get the SYN during syncookie mode because the syn 4693 * packet cannot be saved during syncookie. 4694 * 4695 * OR 4696 * 4697 * 2. the earlier saved syn which was done by 4698 * bpf_setsockopt(TCP_SAVE_SYN). 4699 * 4700 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 4701 * SYN packet is obtained. 4702 * 4703 * If the bpf-prog does not need the IP[46] header, the 4704 * bpf-prog can avoid parsing the IP header by using 4705 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 4706 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 4707 * 4708 * >0: Total number of bytes copied 4709 * -ENOSPC: Not enough space in optval. Only optlen number of 4710 * bytes is copied. 4711 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 4712 * is not saved by setsockopt(TCP_SAVE_SYN). 4713 */ 4714 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 4715 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 4716 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 4717 }; 4718 4719 enum { 4720 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 4721 }; 4722 4723 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 4724 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 4725 */ 4726 enum { 4727 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 4728 * total option spaces 4729 * required for an established 4730 * sk in order to calculate the 4731 * MSS. No skb is actually 4732 * sent. 4733 */ 4734 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 4735 * when sending a SYN. 4736 */ 4737 }; 4738 4739 struct bpf_perf_event_value { 4740 __u64 counter; 4741 __u64 enabled; 4742 __u64 running; 4743 }; 4744 4745 enum { 4746 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 4747 BPF_DEVCG_ACC_READ = (1ULL << 1), 4748 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 4749 }; 4750 4751 enum { 4752 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 4753 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 4754 }; 4755 4756 struct bpf_cgroup_dev_ctx { 4757 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 4758 __u32 access_type; 4759 __u32 major; 4760 __u32 minor; 4761 }; 4762 4763 struct bpf_raw_tracepoint_args { 4764 __u64 args[0]; 4765 }; 4766 4767 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 4768 * OUTPUT: Do lookup from egress perspective; default is ingress 4769 */ 4770 enum { 4771 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 4772 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 4773 }; 4774 4775 enum { 4776 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 4777 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 4778 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 4779 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 4780 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 4781 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 4782 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 4783 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 4784 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 4785 }; 4786 4787 struct bpf_fib_lookup { 4788 /* input: network family for lookup (AF_INET, AF_INET6) 4789 * output: network family of egress nexthop 4790 */ 4791 __u8 family; 4792 4793 /* set if lookup is to consider L4 data - e.g., FIB rules */ 4794 __u8 l4_protocol; 4795 __be16 sport; 4796 __be16 dport; 4797 4798 /* total length of packet from network header - used for MTU check */ 4799 __u16 tot_len; 4800 4801 /* input: L3 device index for lookup 4802 * output: device index from FIB lookup 4803 */ 4804 __u32 ifindex; 4805 4806 union { 4807 /* inputs to lookup */ 4808 __u8 tos; /* AF_INET */ 4809 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 4810 4811 /* output: metric of fib result (IPv4/IPv6 only) */ 4812 __u32 rt_metric; 4813 }; 4814 4815 union { 4816 __be32 ipv4_src; 4817 __u32 ipv6_src[4]; /* in6_addr; network order */ 4818 }; 4819 4820 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 4821 * network header. output: bpf_fib_lookup sets to gateway address 4822 * if FIB lookup returns gateway route 4823 */ 4824 union { 4825 __be32 ipv4_dst; 4826 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4827 }; 4828 4829 /* output */ 4830 __be16 h_vlan_proto; 4831 __be16 h_vlan_TCI; 4832 __u8 smac[6]; /* ETH_ALEN */ 4833 __u8 dmac[6]; /* ETH_ALEN */ 4834 }; 4835 4836 enum bpf_task_fd_type { 4837 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 4838 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 4839 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 4840 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 4841 BPF_FD_TYPE_UPROBE, /* filename + offset */ 4842 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 4843 }; 4844 4845 enum { 4846 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 4847 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 4848 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 4849 }; 4850 4851 struct bpf_flow_keys { 4852 __u16 nhoff; 4853 __u16 thoff; 4854 __u16 addr_proto; /* ETH_P_* of valid addrs */ 4855 __u8 is_frag; 4856 __u8 is_first_frag; 4857 __u8 is_encap; 4858 __u8 ip_proto; 4859 __be16 n_proto; 4860 __be16 sport; 4861 __be16 dport; 4862 union { 4863 struct { 4864 __be32 ipv4_src; 4865 __be32 ipv4_dst; 4866 }; 4867 struct { 4868 __u32 ipv6_src[4]; /* in6_addr; network order */ 4869 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4870 }; 4871 }; 4872 __u32 flags; 4873 __be32 flow_label; 4874 }; 4875 4876 struct bpf_func_info { 4877 __u32 insn_off; 4878 __u32 type_id; 4879 }; 4880 4881 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 4882 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 4883 4884 struct bpf_line_info { 4885 __u32 insn_off; 4886 __u32 file_name_off; 4887 __u32 line_off; 4888 __u32 line_col; 4889 }; 4890 4891 struct bpf_spin_lock { 4892 __u32 val; 4893 }; 4894 4895 struct bpf_sysctl { 4896 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 4897 * Allows 1,2,4-byte read, but no write. 4898 */ 4899 __u32 file_pos; /* Sysctl file position to read from, write to. 4900 * Allows 1,2,4-byte read an 4-byte write. 4901 */ 4902 }; 4903 4904 struct bpf_sockopt { 4905 __bpf_md_ptr(struct bpf_sock *, sk); 4906 __bpf_md_ptr(void *, optval); 4907 __bpf_md_ptr(void *, optval_end); 4908 4909 __s32 level; 4910 __s32 optname; 4911 __s32 optlen; 4912 __s32 retval; 4913 }; 4914 4915 struct bpf_pidns_info { 4916 __u32 pid; 4917 __u32 tgid; 4918 }; 4919 4920 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 4921 struct bpf_sk_lookup { 4922 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 4923 4924 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 4925 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 4926 __u32 remote_ip4; /* Network byte order */ 4927 __u32 remote_ip6[4]; /* Network byte order */ 4928 __u32 remote_port; /* Network byte order */ 4929 __u32 local_ip4; /* Network byte order */ 4930 __u32 local_ip6[4]; /* Network byte order */ 4931 __u32 local_port; /* Host byte order */ 4932 }; 4933 4934 /* 4935 * struct btf_ptr is used for typed pointer representation; the 4936 * type id is used to render the pointer data as the appropriate type 4937 * via the bpf_snprintf_btf() helper described above. A flags field - 4938 * potentially to specify additional details about the BTF pointer 4939 * (rather than its mode of display) - is included for future use. 4940 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 4941 */ 4942 struct btf_ptr { 4943 void *ptr; 4944 __u32 type_id; 4945 __u32 flags; /* BTF ptr flags; unused at present. */ 4946 }; 4947 4948 /* 4949 * Flags to control bpf_snprintf_btf() behaviour. 4950 * - BTF_F_COMPACT: no formatting around type information 4951 * - BTF_F_NONAME: no struct/union member names/types 4952 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 4953 * equivalent to %px. 4954 * - BTF_F_ZERO: show zero-valued struct/union members; they 4955 * are not displayed by default 4956 */ 4957 enum { 4958 BTF_F_COMPACT = (1ULL << 0), 4959 BTF_F_NONAME = (1ULL << 1), 4960 BTF_F_PTR_RAW = (1ULL << 2), 4961 BTF_F_ZERO = (1ULL << 3), 4962 }; 4963 4964 #endif /* _UAPI__LINUX_BPF_H__ */ 4965