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