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