1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 /* BPF syscall commands, see bpf(2) man-page for details. */ 85 enum bpf_cmd { 86 BPF_MAP_CREATE, 87 BPF_MAP_LOOKUP_ELEM, 88 BPF_MAP_UPDATE_ELEM, 89 BPF_MAP_DELETE_ELEM, 90 BPF_MAP_GET_NEXT_KEY, 91 BPF_PROG_LOAD, 92 BPF_OBJ_PIN, 93 BPF_OBJ_GET, 94 BPF_PROG_ATTACH, 95 BPF_PROG_DETACH, 96 BPF_PROG_TEST_RUN, 97 BPF_PROG_GET_NEXT_ID, 98 BPF_MAP_GET_NEXT_ID, 99 BPF_PROG_GET_FD_BY_ID, 100 BPF_MAP_GET_FD_BY_ID, 101 BPF_OBJ_GET_INFO_BY_FD, 102 BPF_PROG_QUERY, 103 BPF_RAW_TRACEPOINT_OPEN, 104 BPF_BTF_LOAD, 105 BPF_BTF_GET_FD_BY_ID, 106 BPF_TASK_FD_QUERY, 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 108 BPF_MAP_FREEZE, 109 BPF_BTF_GET_NEXT_ID, 110 BPF_MAP_LOOKUP_BATCH, 111 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 112 BPF_MAP_UPDATE_BATCH, 113 BPF_MAP_DELETE_BATCH, 114 }; 115 116 enum bpf_map_type { 117 BPF_MAP_TYPE_UNSPEC, 118 BPF_MAP_TYPE_HASH, 119 BPF_MAP_TYPE_ARRAY, 120 BPF_MAP_TYPE_PROG_ARRAY, 121 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 122 BPF_MAP_TYPE_PERCPU_HASH, 123 BPF_MAP_TYPE_PERCPU_ARRAY, 124 BPF_MAP_TYPE_STACK_TRACE, 125 BPF_MAP_TYPE_CGROUP_ARRAY, 126 BPF_MAP_TYPE_LRU_HASH, 127 BPF_MAP_TYPE_LRU_PERCPU_HASH, 128 BPF_MAP_TYPE_LPM_TRIE, 129 BPF_MAP_TYPE_ARRAY_OF_MAPS, 130 BPF_MAP_TYPE_HASH_OF_MAPS, 131 BPF_MAP_TYPE_DEVMAP, 132 BPF_MAP_TYPE_SOCKMAP, 133 BPF_MAP_TYPE_CPUMAP, 134 BPF_MAP_TYPE_XSKMAP, 135 BPF_MAP_TYPE_SOCKHASH, 136 BPF_MAP_TYPE_CGROUP_STORAGE, 137 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 138 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 139 BPF_MAP_TYPE_QUEUE, 140 BPF_MAP_TYPE_STACK, 141 BPF_MAP_TYPE_SK_STORAGE, 142 BPF_MAP_TYPE_DEVMAP_HASH, 143 BPF_MAP_TYPE_STRUCT_OPS, 144 }; 145 146 /* Note that tracing related programs such as 147 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 148 * are not subject to a stable API since kernel internal data 149 * structures can change from release to release and may 150 * therefore break existing tracing BPF programs. Tracing BPF 151 * programs correspond to /a/ specific kernel which is to be 152 * analyzed, and not /a/ specific kernel /and/ all future ones. 153 */ 154 enum bpf_prog_type { 155 BPF_PROG_TYPE_UNSPEC, 156 BPF_PROG_TYPE_SOCKET_FILTER, 157 BPF_PROG_TYPE_KPROBE, 158 BPF_PROG_TYPE_SCHED_CLS, 159 BPF_PROG_TYPE_SCHED_ACT, 160 BPF_PROG_TYPE_TRACEPOINT, 161 BPF_PROG_TYPE_XDP, 162 BPF_PROG_TYPE_PERF_EVENT, 163 BPF_PROG_TYPE_CGROUP_SKB, 164 BPF_PROG_TYPE_CGROUP_SOCK, 165 BPF_PROG_TYPE_LWT_IN, 166 BPF_PROG_TYPE_LWT_OUT, 167 BPF_PROG_TYPE_LWT_XMIT, 168 BPF_PROG_TYPE_SOCK_OPS, 169 BPF_PROG_TYPE_SK_SKB, 170 BPF_PROG_TYPE_CGROUP_DEVICE, 171 BPF_PROG_TYPE_SK_MSG, 172 BPF_PROG_TYPE_RAW_TRACEPOINT, 173 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 174 BPF_PROG_TYPE_LWT_SEG6LOCAL, 175 BPF_PROG_TYPE_LIRC_MODE2, 176 BPF_PROG_TYPE_SK_REUSEPORT, 177 BPF_PROG_TYPE_FLOW_DISSECTOR, 178 BPF_PROG_TYPE_CGROUP_SYSCTL, 179 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 180 BPF_PROG_TYPE_CGROUP_SOCKOPT, 181 BPF_PROG_TYPE_TRACING, 182 BPF_PROG_TYPE_STRUCT_OPS, 183 }; 184 185 enum bpf_attach_type { 186 BPF_CGROUP_INET_INGRESS, 187 BPF_CGROUP_INET_EGRESS, 188 BPF_CGROUP_INET_SOCK_CREATE, 189 BPF_CGROUP_SOCK_OPS, 190 BPF_SK_SKB_STREAM_PARSER, 191 BPF_SK_SKB_STREAM_VERDICT, 192 BPF_CGROUP_DEVICE, 193 BPF_SK_MSG_VERDICT, 194 BPF_CGROUP_INET4_BIND, 195 BPF_CGROUP_INET6_BIND, 196 BPF_CGROUP_INET4_CONNECT, 197 BPF_CGROUP_INET6_CONNECT, 198 BPF_CGROUP_INET4_POST_BIND, 199 BPF_CGROUP_INET6_POST_BIND, 200 BPF_CGROUP_UDP4_SENDMSG, 201 BPF_CGROUP_UDP6_SENDMSG, 202 BPF_LIRC_MODE2, 203 BPF_FLOW_DISSECTOR, 204 BPF_CGROUP_SYSCTL, 205 BPF_CGROUP_UDP4_RECVMSG, 206 BPF_CGROUP_UDP6_RECVMSG, 207 BPF_CGROUP_GETSOCKOPT, 208 BPF_CGROUP_SETSOCKOPT, 209 BPF_TRACE_RAW_TP, 210 BPF_TRACE_FENTRY, 211 BPF_TRACE_FEXIT, 212 __MAX_BPF_ATTACH_TYPE 213 }; 214 215 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 216 217 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 218 * 219 * NONE(default): No further bpf programs allowed in the subtree. 220 * 221 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 222 * the program in this cgroup yields to sub-cgroup program. 223 * 224 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 225 * that cgroup program gets run in addition to the program in this cgroup. 226 * 227 * Only one program is allowed to be attached to a cgroup with 228 * NONE or BPF_F_ALLOW_OVERRIDE flag. 229 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 230 * release old program and attach the new one. Attach flags has to match. 231 * 232 * Multiple programs are allowed to be attached to a cgroup with 233 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 234 * (those that were attached first, run first) 235 * The programs of sub-cgroup are executed first, then programs of 236 * this cgroup and then programs of parent cgroup. 237 * When children program makes decision (like picking TCP CA or sock bind) 238 * parent program has a chance to override it. 239 * 240 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 241 * programs for a cgroup. Though it's possible to replace an old program at 242 * any position by also specifying BPF_F_REPLACE flag and position itself in 243 * replace_bpf_fd attribute. Old program at this position will be released. 244 * 245 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 246 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 247 * Ex1: 248 * cgrp1 (MULTI progs A, B) -> 249 * cgrp2 (OVERRIDE prog C) -> 250 * cgrp3 (MULTI prog D) -> 251 * cgrp4 (OVERRIDE prog E) -> 252 * cgrp5 (NONE prog F) 253 * the event in cgrp5 triggers execution of F,D,A,B in that order. 254 * if prog F is detached, the execution is E,D,A,B 255 * if prog F and D are detached, the execution is E,A,B 256 * if prog F, E and D are detached, the execution is C,A,B 257 * 258 * All eligible programs are executed regardless of return code from 259 * earlier programs. 260 */ 261 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 262 #define BPF_F_ALLOW_MULTI (1U << 1) 263 #define BPF_F_REPLACE (1U << 2) 264 265 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 266 * verifier will perform strict alignment checking as if the kernel 267 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 268 * and NET_IP_ALIGN defined to 2. 269 */ 270 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 271 272 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 273 * verifier will allow any alignment whatsoever. On platforms 274 * with strict alignment requirements for loads ands stores (such 275 * as sparc and mips) the verifier validates that all loads and 276 * stores provably follow this requirement. This flag turns that 277 * checking and enforcement off. 278 * 279 * It is mostly used for testing when we want to validate the 280 * context and memory access aspects of the verifier, but because 281 * of an unaligned access the alignment check would trigger before 282 * the one we are interested in. 283 */ 284 #define BPF_F_ANY_ALIGNMENT (1U << 1) 285 286 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 287 * Verifier does sub-register def/use analysis and identifies instructions whose 288 * def only matters for low 32-bit, high 32-bit is never referenced later 289 * through implicit zero extension. Therefore verifier notifies JIT back-ends 290 * that it is safe to ignore clearing high 32-bit for these instructions. This 291 * saves some back-ends a lot of code-gen. However such optimization is not 292 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 293 * hence hasn't used verifier's analysis result. But, we really want to have a 294 * way to be able to verify the correctness of the described optimization on 295 * x86_64 on which testsuites are frequently exercised. 296 * 297 * So, this flag is introduced. Once it is set, verifier will randomize high 298 * 32-bit for those instructions who has been identified as safe to ignore them. 299 * Then, if verifier is not doing correct analysis, such randomization will 300 * regress tests to expose bugs. 301 */ 302 #define BPF_F_TEST_RND_HI32 (1U << 2) 303 304 /* The verifier internal test flag. Behavior is undefined */ 305 #define BPF_F_TEST_STATE_FREQ (1U << 3) 306 307 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 308 * two extensions: 309 * 310 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 311 * insn[0].imm: map fd map fd 312 * insn[1].imm: 0 offset into value 313 * insn[0].off: 0 0 314 * insn[1].off: 0 0 315 * ldimm64 rewrite: address of map address of map[0]+offset 316 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 317 */ 318 #define BPF_PSEUDO_MAP_FD 1 319 #define BPF_PSEUDO_MAP_VALUE 2 320 321 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 322 * offset to another bpf function 323 */ 324 #define BPF_PSEUDO_CALL 1 325 326 /* flags for BPF_MAP_UPDATE_ELEM command */ 327 #define BPF_ANY 0 /* create new element or update existing */ 328 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 329 #define BPF_EXIST 2 /* update existing element */ 330 #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 331 332 /* flags for BPF_MAP_CREATE command */ 333 #define BPF_F_NO_PREALLOC (1U << 0) 334 /* Instead of having one common LRU list in the 335 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 336 * which can scale and perform better. 337 * Note, the LRU nodes (including free nodes) cannot be moved 338 * across different LRU lists. 339 */ 340 #define BPF_F_NO_COMMON_LRU (1U << 1) 341 /* Specify numa node during map creation */ 342 #define BPF_F_NUMA_NODE (1U << 2) 343 344 #define BPF_OBJ_NAME_LEN 16U 345 346 /* Flags for accessing BPF object from syscall side. */ 347 #define BPF_F_RDONLY (1U << 3) 348 #define BPF_F_WRONLY (1U << 4) 349 350 /* Flag for stack_map, store build_id+offset instead of pointer */ 351 #define BPF_F_STACK_BUILD_ID (1U << 5) 352 353 /* Zero-initialize hash function seed. This should only be used for testing. */ 354 #define BPF_F_ZERO_SEED (1U << 6) 355 356 /* Flags for accessing BPF object from program side. */ 357 #define BPF_F_RDONLY_PROG (1U << 7) 358 #define BPF_F_WRONLY_PROG (1U << 8) 359 360 /* Clone map from listener for newly accepted socket */ 361 #define BPF_F_CLONE (1U << 9) 362 363 /* Enable memory-mapping BPF map */ 364 #define BPF_F_MMAPABLE (1U << 10) 365 366 /* Flags for BPF_PROG_QUERY. */ 367 368 /* Query effective (directly attached + inherited from ancestor cgroups) 369 * programs that will be executed for events within a cgroup. 370 * attach_flags with this flag are returned only for directly attached programs. 371 */ 372 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 373 374 enum bpf_stack_build_id_status { 375 /* user space need an empty entry to identify end of a trace */ 376 BPF_STACK_BUILD_ID_EMPTY = 0, 377 /* with valid build_id and offset */ 378 BPF_STACK_BUILD_ID_VALID = 1, 379 /* couldn't get build_id, fallback to ip */ 380 BPF_STACK_BUILD_ID_IP = 2, 381 }; 382 383 #define BPF_BUILD_ID_SIZE 20 384 struct bpf_stack_build_id { 385 __s32 status; 386 unsigned char build_id[BPF_BUILD_ID_SIZE]; 387 union { 388 __u64 offset; 389 __u64 ip; 390 }; 391 }; 392 393 union bpf_attr { 394 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 395 __u32 map_type; /* one of enum bpf_map_type */ 396 __u32 key_size; /* size of key in bytes */ 397 __u32 value_size; /* size of value in bytes */ 398 __u32 max_entries; /* max number of entries in a map */ 399 __u32 map_flags; /* BPF_MAP_CREATE related 400 * flags defined above. 401 */ 402 __u32 inner_map_fd; /* fd pointing to the inner map */ 403 __u32 numa_node; /* numa node (effective only if 404 * BPF_F_NUMA_NODE is set). 405 */ 406 char map_name[BPF_OBJ_NAME_LEN]; 407 __u32 map_ifindex; /* ifindex of netdev to create on */ 408 __u32 btf_fd; /* fd pointing to a BTF type data */ 409 __u32 btf_key_type_id; /* BTF type_id of the key */ 410 __u32 btf_value_type_id; /* BTF type_id of the value */ 411 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 412 * struct stored as the 413 * map value 414 */ 415 }; 416 417 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 418 __u32 map_fd; 419 __aligned_u64 key; 420 union { 421 __aligned_u64 value; 422 __aligned_u64 next_key; 423 }; 424 __u64 flags; 425 }; 426 427 struct { /* struct used by BPF_MAP_*_BATCH commands */ 428 __aligned_u64 in_batch; /* start batch, 429 * NULL to start from beginning 430 */ 431 __aligned_u64 out_batch; /* output: next start batch */ 432 __aligned_u64 keys; 433 __aligned_u64 values; 434 __u32 count; /* input/output: 435 * input: # of key/value 436 * elements 437 * output: # of filled elements 438 */ 439 __u32 map_fd; 440 __u64 elem_flags; 441 __u64 flags; 442 } batch; 443 444 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 445 __u32 prog_type; /* one of enum bpf_prog_type */ 446 __u32 insn_cnt; 447 __aligned_u64 insns; 448 __aligned_u64 license; 449 __u32 log_level; /* verbosity level of verifier */ 450 __u32 log_size; /* size of user buffer */ 451 __aligned_u64 log_buf; /* user supplied buffer */ 452 __u32 kern_version; /* not used */ 453 __u32 prog_flags; 454 char prog_name[BPF_OBJ_NAME_LEN]; 455 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 456 /* For some prog types expected attach type must be known at 457 * load time to verify attach type specific parts of prog 458 * (context accesses, allowed helpers, etc). 459 */ 460 __u32 expected_attach_type; 461 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 462 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 463 __aligned_u64 func_info; /* func info */ 464 __u32 func_info_cnt; /* number of bpf_func_info records */ 465 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 466 __aligned_u64 line_info; /* line info */ 467 __u32 line_info_cnt; /* number of bpf_line_info records */ 468 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 469 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 470 }; 471 472 struct { /* anonymous struct used by BPF_OBJ_* commands */ 473 __aligned_u64 pathname; 474 __u32 bpf_fd; 475 __u32 file_flags; 476 }; 477 478 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 479 __u32 target_fd; /* container object to attach to */ 480 __u32 attach_bpf_fd; /* eBPF program to attach */ 481 __u32 attach_type; 482 __u32 attach_flags; 483 __u32 replace_bpf_fd; /* previously attached eBPF 484 * program to replace if 485 * BPF_F_REPLACE is used 486 */ 487 }; 488 489 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 490 __u32 prog_fd; 491 __u32 retval; 492 __u32 data_size_in; /* input: len of data_in */ 493 __u32 data_size_out; /* input/output: len of data_out 494 * returns ENOSPC if data_out 495 * is too small. 496 */ 497 __aligned_u64 data_in; 498 __aligned_u64 data_out; 499 __u32 repeat; 500 __u32 duration; 501 __u32 ctx_size_in; /* input: len of ctx_in */ 502 __u32 ctx_size_out; /* input/output: len of ctx_out 503 * returns ENOSPC if ctx_out 504 * is too small. 505 */ 506 __aligned_u64 ctx_in; 507 __aligned_u64 ctx_out; 508 } test; 509 510 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 511 union { 512 __u32 start_id; 513 __u32 prog_id; 514 __u32 map_id; 515 __u32 btf_id; 516 }; 517 __u32 next_id; 518 __u32 open_flags; 519 }; 520 521 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 522 __u32 bpf_fd; 523 __u32 info_len; 524 __aligned_u64 info; 525 } info; 526 527 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 528 __u32 target_fd; /* container object to query */ 529 __u32 attach_type; 530 __u32 query_flags; 531 __u32 attach_flags; 532 __aligned_u64 prog_ids; 533 __u32 prog_cnt; 534 } query; 535 536 struct { 537 __u64 name; 538 __u32 prog_fd; 539 } raw_tracepoint; 540 541 struct { /* anonymous struct for BPF_BTF_LOAD */ 542 __aligned_u64 btf; 543 __aligned_u64 btf_log_buf; 544 __u32 btf_size; 545 __u32 btf_log_size; 546 __u32 btf_log_level; 547 }; 548 549 struct { 550 __u32 pid; /* input: pid */ 551 __u32 fd; /* input: fd */ 552 __u32 flags; /* input: flags */ 553 __u32 buf_len; /* input/output: buf len */ 554 __aligned_u64 buf; /* input/output: 555 * tp_name for tracepoint 556 * symbol for kprobe 557 * filename for uprobe 558 */ 559 __u32 prog_id; /* output: prod_id */ 560 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 561 __u64 probe_offset; /* output: probe_offset */ 562 __u64 probe_addr; /* output: probe_addr */ 563 } task_fd_query; 564 } __attribute__((aligned(8))); 565 566 /* The description below is an attempt at providing documentation to eBPF 567 * developers about the multiple available eBPF helper functions. It can be 568 * parsed and used to produce a manual page. The workflow is the following, 569 * and requires the rst2man utility: 570 * 571 * $ ./scripts/bpf_helpers_doc.py \ 572 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 573 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 574 * $ man /tmp/bpf-helpers.7 575 * 576 * Note that in order to produce this external documentation, some RST 577 * formatting is used in the descriptions to get "bold" and "italics" in 578 * manual pages. Also note that the few trailing white spaces are 579 * intentional, removing them would break paragraphs for rst2man. 580 * 581 * Start of BPF helper function descriptions: 582 * 583 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 584 * Description 585 * Perform a lookup in *map* for an entry associated to *key*. 586 * Return 587 * Map value associated to *key*, or **NULL** if no entry was 588 * found. 589 * 590 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 591 * Description 592 * Add or update the value of the entry associated to *key* in 593 * *map* with *value*. *flags* is one of: 594 * 595 * **BPF_NOEXIST** 596 * The entry for *key* must not exist in the map. 597 * **BPF_EXIST** 598 * The entry for *key* must already exist in the map. 599 * **BPF_ANY** 600 * No condition on the existence of the entry for *key*. 601 * 602 * Flag value **BPF_NOEXIST** cannot be used for maps of types 603 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 604 * elements always exist), the helper would return an error. 605 * Return 606 * 0 on success, or a negative error in case of failure. 607 * 608 * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 609 * Description 610 * Delete entry with *key* from *map*. 611 * Return 612 * 0 on success, or a negative error in case of failure. 613 * 614 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 615 * Description 616 * For tracing programs, safely attempt to read *size* bytes from 617 * kernel space address *unsafe_ptr* and store the data in *dst*. 618 * 619 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel() 620 * instead. 621 * Return 622 * 0 on success, or a negative error in case of failure. 623 * 624 * u64 bpf_ktime_get_ns(void) 625 * Description 626 * Return the time elapsed since system boot, in nanoseconds. 627 * Return 628 * Current *ktime*. 629 * 630 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 631 * Description 632 * This helper is a "printk()-like" facility for debugging. It 633 * prints a message defined by format *fmt* (of size *fmt_size*) 634 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 635 * available. It can take up to three additional **u64** 636 * arguments (as an eBPF helpers, the total number of arguments is 637 * limited to five). 638 * 639 * Each time the helper is called, it appends a line to the trace. 640 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 641 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 642 * The format of the trace is customizable, and the exact output 643 * one will get depends on the options set in 644 * *\/sys/kernel/debug/tracing/trace_options* (see also the 645 * *README* file under the same directory). However, it usually 646 * defaults to something like: 647 * 648 * :: 649 * 650 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 651 * 652 * In the above: 653 * 654 * * ``telnet`` is the name of the current task. 655 * * ``470`` is the PID of the current task. 656 * * ``001`` is the CPU number on which the task is 657 * running. 658 * * In ``.N..``, each character refers to a set of 659 * options (whether irqs are enabled, scheduling 660 * options, whether hard/softirqs are running, level of 661 * preempt_disabled respectively). **N** means that 662 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 663 * are set. 664 * * ``419421.045894`` is a timestamp. 665 * * ``0x00000001`` is a fake value used by BPF for the 666 * instruction pointer register. 667 * * ``<formatted msg>`` is the message formatted with 668 * *fmt*. 669 * 670 * The conversion specifiers supported by *fmt* are similar, but 671 * more limited than for printk(). They are **%d**, **%i**, 672 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 673 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 674 * of field, padding with zeroes, etc.) is available, and the 675 * helper will return **-EINVAL** (but print nothing) if it 676 * encounters an unknown specifier. 677 * 678 * Also, note that **bpf_trace_printk**\ () is slow, and should 679 * only be used for debugging purposes. For this reason, a notice 680 * bloc (spanning several lines) is printed to kernel logs and 681 * states that the helper should not be used "for production use" 682 * the first time this helper is used (or more precisely, when 683 * **trace_printk**\ () buffers are allocated). For passing values 684 * to user space, perf events should be preferred. 685 * Return 686 * The number of bytes written to the buffer, or a negative error 687 * in case of failure. 688 * 689 * u32 bpf_get_prandom_u32(void) 690 * Description 691 * Get a pseudo-random number. 692 * 693 * From a security point of view, this helper uses its own 694 * pseudo-random internal state, and cannot be used to infer the 695 * seed of other random functions in the kernel. However, it is 696 * essential to note that the generator used by the helper is not 697 * cryptographically secure. 698 * Return 699 * A random 32-bit unsigned value. 700 * 701 * u32 bpf_get_smp_processor_id(void) 702 * Description 703 * Get the SMP (symmetric multiprocessing) processor id. Note that 704 * all programs run with preemption disabled, which means that the 705 * SMP processor id is stable during all the execution of the 706 * program. 707 * Return 708 * The SMP id of the processor running the program. 709 * 710 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 711 * Description 712 * Store *len* bytes from address *from* into the packet 713 * associated to *skb*, at *offset*. *flags* are a combination of 714 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 715 * checksum for the packet after storing the bytes) and 716 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 717 * **->swhash** and *skb*\ **->l4hash** to 0). 718 * 719 * A call to this helper is susceptible to change the underlying 720 * packet buffer. Therefore, at load time, all checks on pointers 721 * previously done by the verifier are invalidated and must be 722 * performed again, if the helper is used in combination with 723 * direct packet access. 724 * Return 725 * 0 on success, or a negative error in case of failure. 726 * 727 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 728 * Description 729 * Recompute the layer 3 (e.g. IP) checksum for the packet 730 * associated to *skb*. Computation is incremental, so the helper 731 * must know the former value of the header field that was 732 * modified (*from*), the new value of this field (*to*), and the 733 * number of bytes (2 or 4) for this field, stored in *size*. 734 * Alternatively, it is possible to store the difference between 735 * the previous and the new values of the header field in *to*, by 736 * setting *from* and *size* to 0. For both methods, *offset* 737 * indicates the location of the IP checksum within the packet. 738 * 739 * This helper works in combination with **bpf_csum_diff**\ (), 740 * which does not update the checksum in-place, but offers more 741 * flexibility and can handle sizes larger than 2 or 4 for the 742 * checksum to update. 743 * 744 * A call to this helper is susceptible to change the underlying 745 * packet buffer. Therefore, at load time, all checks on pointers 746 * previously done by the verifier are invalidated and must be 747 * performed again, if the helper is used in combination with 748 * direct packet access. 749 * Return 750 * 0 on success, or a negative error in case of failure. 751 * 752 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 753 * Description 754 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 755 * packet associated to *skb*. Computation is incremental, so the 756 * helper must know the former value of the header field that was 757 * modified (*from*), the new value of this field (*to*), and the 758 * number of bytes (2 or 4) for this field, stored on the lowest 759 * four bits of *flags*. Alternatively, it is possible to store 760 * the difference between the previous and the new values of the 761 * header field in *to*, by setting *from* and the four lowest 762 * bits of *flags* to 0. For both methods, *offset* indicates the 763 * location of the IP checksum within the packet. In addition to 764 * the size of the field, *flags* can be added (bitwise OR) actual 765 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 766 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 767 * for updates resulting in a null checksum the value is set to 768 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 769 * the checksum is to be computed against a pseudo-header. 770 * 771 * This helper works in combination with **bpf_csum_diff**\ (), 772 * which does not update the checksum in-place, but offers more 773 * flexibility and can handle sizes larger than 2 or 4 for the 774 * checksum to update. 775 * 776 * A call to this helper is susceptible to change the underlying 777 * packet buffer. Therefore, at load time, all checks on pointers 778 * previously done by the verifier are invalidated and must be 779 * performed again, if the helper is used in combination with 780 * direct packet access. 781 * Return 782 * 0 on success, or a negative error in case of failure. 783 * 784 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 785 * Description 786 * This special helper is used to trigger a "tail call", or in 787 * other words, to jump into another eBPF program. The same stack 788 * frame is used (but values on stack and in registers for the 789 * caller are not accessible to the callee). This mechanism allows 790 * for program chaining, either for raising the maximum number of 791 * available eBPF instructions, or to execute given programs in 792 * conditional blocks. For security reasons, there is an upper 793 * limit to the number of successive tail calls that can be 794 * performed. 795 * 796 * Upon call of this helper, the program attempts to jump into a 797 * program referenced at index *index* in *prog_array_map*, a 798 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 799 * *ctx*, a pointer to the context. 800 * 801 * If the call succeeds, the kernel immediately runs the first 802 * instruction of the new program. This is not a function call, 803 * and it never returns to the previous program. If the call 804 * fails, then the helper has no effect, and the caller continues 805 * to run its subsequent instructions. A call can fail if the 806 * destination program for the jump does not exist (i.e. *index* 807 * is superior to the number of entries in *prog_array_map*), or 808 * if the maximum number of tail calls has been reached for this 809 * chain of programs. This limit is defined in the kernel by the 810 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 811 * which is currently set to 32. 812 * Return 813 * 0 on success, or a negative error in case of failure. 814 * 815 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 816 * Description 817 * Clone and redirect the packet associated to *skb* to another 818 * net device of index *ifindex*. Both ingress and egress 819 * interfaces can be used for redirection. The **BPF_F_INGRESS** 820 * value in *flags* is used to make the distinction (ingress path 821 * is selected if the flag is present, egress path otherwise). 822 * This is the only flag supported for now. 823 * 824 * In comparison with **bpf_redirect**\ () helper, 825 * **bpf_clone_redirect**\ () has the associated cost of 826 * duplicating the packet buffer, but this can be executed out of 827 * the eBPF program. Conversely, **bpf_redirect**\ () is more 828 * efficient, but it is handled through an action code where the 829 * redirection happens only after the eBPF program has returned. 830 * 831 * A call to this helper is susceptible to change the underlying 832 * packet buffer. Therefore, at load time, all checks on pointers 833 * previously done by the verifier are invalidated and must be 834 * performed again, if the helper is used in combination with 835 * direct packet access. 836 * Return 837 * 0 on success, or a negative error in case of failure. 838 * 839 * u64 bpf_get_current_pid_tgid(void) 840 * Return 841 * A 64-bit integer containing the current tgid and pid, and 842 * created as such: 843 * *current_task*\ **->tgid << 32 \|** 844 * *current_task*\ **->pid**. 845 * 846 * u64 bpf_get_current_uid_gid(void) 847 * Return 848 * A 64-bit integer containing the current GID and UID, and 849 * created as such: *current_gid* **<< 32 \|** *current_uid*. 850 * 851 * int bpf_get_current_comm(void *buf, u32 size_of_buf) 852 * Description 853 * Copy the **comm** attribute of the current task into *buf* of 854 * *size_of_buf*. The **comm** attribute contains the name of 855 * the executable (excluding the path) for the current task. The 856 * *size_of_buf* must be strictly positive. On success, the 857 * helper makes sure that the *buf* is NUL-terminated. On failure, 858 * it is filled with zeroes. 859 * Return 860 * 0 on success, or a negative error in case of failure. 861 * 862 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 863 * Description 864 * Retrieve the classid for the current task, i.e. for the net_cls 865 * cgroup to which *skb* belongs. 866 * 867 * This helper can be used on TC egress path, but not on ingress. 868 * 869 * The net_cls cgroup provides an interface to tag network packets 870 * based on a user-provided identifier for all traffic coming from 871 * the tasks belonging to the related cgroup. See also the related 872 * kernel documentation, available from the Linux sources in file 873 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 874 * 875 * The Linux kernel has two versions for cgroups: there are 876 * cgroups v1 and cgroups v2. Both are available to users, who can 877 * use a mixture of them, but note that the net_cls cgroup is for 878 * cgroup v1 only. This makes it incompatible with BPF programs 879 * run on cgroups, which is a cgroup-v2-only feature (a socket can 880 * only hold data for one version of cgroups at a time). 881 * 882 * This helper is only available is the kernel was compiled with 883 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 884 * "**y**" or to "**m**". 885 * Return 886 * The classid, or 0 for the default unconfigured classid. 887 * 888 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 889 * Description 890 * Push a *vlan_tci* (VLAN tag control information) of protocol 891 * *vlan_proto* to the packet associated to *skb*, then update 892 * the checksum. Note that if *vlan_proto* is different from 893 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 894 * be **ETH_P_8021Q**. 895 * 896 * A call to this helper is susceptible to change the underlying 897 * packet buffer. Therefore, at load time, all checks on pointers 898 * previously done by the verifier are invalidated and must be 899 * performed again, if the helper is used in combination with 900 * direct packet access. 901 * Return 902 * 0 on success, or a negative error in case of failure. 903 * 904 * int bpf_skb_vlan_pop(struct sk_buff *skb) 905 * Description 906 * Pop a VLAN header from the packet associated to *skb*. 907 * 908 * A call to this helper is susceptible to change the underlying 909 * packet buffer. Therefore, at load time, all checks on pointers 910 * previously done by the verifier are invalidated and must be 911 * performed again, if the helper is used in combination with 912 * direct packet access. 913 * Return 914 * 0 on success, or a negative error in case of failure. 915 * 916 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 917 * Description 918 * Get tunnel metadata. This helper takes a pointer *key* to an 919 * empty **struct bpf_tunnel_key** of **size**, that will be 920 * filled with tunnel metadata for the packet associated to *skb*. 921 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 922 * indicates that the tunnel is based on IPv6 protocol instead of 923 * IPv4. 924 * 925 * The **struct bpf_tunnel_key** is an object that generalizes the 926 * principal parameters used by various tunneling protocols into a 927 * single struct. This way, it can be used to easily make a 928 * decision based on the contents of the encapsulation header, 929 * "summarized" in this struct. In particular, it holds the IP 930 * address of the remote end (IPv4 or IPv6, depending on the case) 931 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 932 * this struct exposes the *key*\ **->tunnel_id**, which is 933 * generally mapped to a VNI (Virtual Network Identifier), making 934 * it programmable together with the **bpf_skb_set_tunnel_key**\ 935 * () helper. 936 * 937 * Let's imagine that the following code is part of a program 938 * attached to the TC ingress interface, on one end of a GRE 939 * tunnel, and is supposed to filter out all messages coming from 940 * remote ends with IPv4 address other than 10.0.0.1: 941 * 942 * :: 943 * 944 * int ret; 945 * struct bpf_tunnel_key key = {}; 946 * 947 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 948 * if (ret < 0) 949 * return TC_ACT_SHOT; // drop packet 950 * 951 * if (key.remote_ipv4 != 0x0a000001) 952 * return TC_ACT_SHOT; // drop packet 953 * 954 * return TC_ACT_OK; // accept packet 955 * 956 * This interface can also be used with all encapsulation devices 957 * that can operate in "collect metadata" mode: instead of having 958 * one network device per specific configuration, the "collect 959 * metadata" mode only requires a single device where the 960 * configuration can be extracted from this helper. 961 * 962 * This can be used together with various tunnels such as VXLan, 963 * Geneve, GRE or IP in IP (IPIP). 964 * Return 965 * 0 on success, or a negative error in case of failure. 966 * 967 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 968 * Description 969 * Populate tunnel metadata for packet associated to *skb.* The 970 * tunnel metadata is set to the contents of *key*, of *size*. The 971 * *flags* can be set to a combination of the following values: 972 * 973 * **BPF_F_TUNINFO_IPV6** 974 * Indicate that the tunnel is based on IPv6 protocol 975 * instead of IPv4. 976 * **BPF_F_ZERO_CSUM_TX** 977 * For IPv4 packets, add a flag to tunnel metadata 978 * indicating that checksum computation should be skipped 979 * and checksum set to zeroes. 980 * **BPF_F_DONT_FRAGMENT** 981 * Add a flag to tunnel metadata indicating that the 982 * packet should not be fragmented. 983 * **BPF_F_SEQ_NUMBER** 984 * Add a flag to tunnel metadata indicating that a 985 * sequence number should be added to tunnel header before 986 * sending the packet. This flag was added for GRE 987 * encapsulation, but might be used with other protocols 988 * as well in the future. 989 * 990 * Here is a typical usage on the transmit path: 991 * 992 * :: 993 * 994 * struct bpf_tunnel_key key; 995 * populate key ... 996 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 997 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 998 * 999 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1000 * helper for additional information. 1001 * Return 1002 * 0 on success, or a negative error in case of failure. 1003 * 1004 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1005 * Description 1006 * Read the value of a perf event counter. This helper relies on a 1007 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1008 * the perf event counter is selected when *map* is updated with 1009 * perf event file descriptors. The *map* is an array whose size 1010 * is the number of available CPUs, and each cell contains a value 1011 * relative to one CPU. The value to retrieve is indicated by 1012 * *flags*, that contains the index of the CPU to look up, masked 1013 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1014 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1015 * current CPU should be retrieved. 1016 * 1017 * Note that before Linux 4.13, only hardware perf event can be 1018 * retrieved. 1019 * 1020 * Also, be aware that the newer helper 1021 * **bpf_perf_event_read_value**\ () is recommended over 1022 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1023 * quirks where error and counter value are used as a return code 1024 * (which is wrong to do since ranges may overlap). This issue is 1025 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1026 * time provides more features over the **bpf_perf_event_read**\ 1027 * () interface. Please refer to the description of 1028 * **bpf_perf_event_read_value**\ () for details. 1029 * Return 1030 * The value of the perf event counter read from the map, or a 1031 * negative error code in case of failure. 1032 * 1033 * int bpf_redirect(u32 ifindex, u64 flags) 1034 * Description 1035 * Redirect the packet to another net device of index *ifindex*. 1036 * This helper is somewhat similar to **bpf_clone_redirect**\ 1037 * (), except that the packet is not cloned, which provides 1038 * increased performance. 1039 * 1040 * Except for XDP, both ingress and egress interfaces can be used 1041 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1042 * to make the distinction (ingress path is selected if the flag 1043 * is present, egress path otherwise). Currently, XDP only 1044 * supports redirection to the egress interface, and accepts no 1045 * flag at all. 1046 * 1047 * The same effect can be attained with the more generic 1048 * **bpf_redirect_map**\ (), which requires specific maps to be 1049 * used but offers better performance. 1050 * Return 1051 * For XDP, the helper returns **XDP_REDIRECT** on success or 1052 * **XDP_ABORTED** on error. For other program types, the values 1053 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1054 * error. 1055 * 1056 * u32 bpf_get_route_realm(struct sk_buff *skb) 1057 * Description 1058 * Retrieve the realm or the route, that is to say the 1059 * **tclassid** field of the destination for the *skb*. The 1060 * indentifier retrieved is a user-provided tag, similar to the 1061 * one used with the net_cls cgroup (see description for 1062 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1063 * held by a route (a destination entry), not by a task. 1064 * 1065 * Retrieving this identifier works with the clsact TC egress hook 1066 * (see also **tc-bpf(8)**), or alternatively on conventional 1067 * classful egress qdiscs, but not on TC ingress path. In case of 1068 * clsact TC egress hook, this has the advantage that, internally, 1069 * the destination entry has not been dropped yet in the transmit 1070 * path. Therefore, the destination entry does not need to be 1071 * artificially held via **netif_keep_dst**\ () for a classful 1072 * qdisc until the *skb* is freed. 1073 * 1074 * This helper is available only if the kernel was compiled with 1075 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1076 * Return 1077 * The realm of the route for the packet associated to *skb*, or 0 1078 * if none was found. 1079 * 1080 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1081 * Description 1082 * Write raw *data* blob into a special BPF perf event held by 1083 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1084 * event must have the following attributes: **PERF_SAMPLE_RAW** 1085 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1086 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1087 * 1088 * The *flags* are used to indicate the index in *map* for which 1089 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1090 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1091 * to indicate that the index of the current CPU core should be 1092 * used. 1093 * 1094 * The value to write, of *size*, is passed through eBPF stack and 1095 * pointed by *data*. 1096 * 1097 * The context of the program *ctx* needs also be passed to the 1098 * helper. 1099 * 1100 * On user space, a program willing to read the values needs to 1101 * call **perf_event_open**\ () on the perf event (either for 1102 * one or for all CPUs) and to store the file descriptor into the 1103 * *map*. This must be done before the eBPF program can send data 1104 * into it. An example is available in file 1105 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1106 * tree (the eBPF program counterpart is in 1107 * *samples/bpf/trace_output_kern.c*). 1108 * 1109 * **bpf_perf_event_output**\ () achieves better performance 1110 * than **bpf_trace_printk**\ () for sharing data with user 1111 * space, and is much better suitable for streaming data from eBPF 1112 * programs. 1113 * 1114 * Note that this helper is not restricted to tracing use cases 1115 * and can be used with programs attached to TC or XDP as well, 1116 * where it allows for passing data to user space listeners. Data 1117 * can be: 1118 * 1119 * * Only custom structs, 1120 * * Only the packet payload, or 1121 * * A combination of both. 1122 * Return 1123 * 0 on success, or a negative error in case of failure. 1124 * 1125 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1126 * Description 1127 * This helper was provided as an easy way to load data from a 1128 * packet. It can be used to load *len* bytes from *offset* from 1129 * the packet associated to *skb*, into the buffer pointed by 1130 * *to*. 1131 * 1132 * Since Linux 4.7, usage of this helper has mostly been replaced 1133 * by "direct packet access", enabling packet data to be 1134 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1135 * pointing respectively to the first byte of packet data and to 1136 * the byte after the last byte of packet data. However, it 1137 * remains useful if one wishes to read large quantities of data 1138 * at once from a packet into the eBPF stack. 1139 * Return 1140 * 0 on success, or a negative error in case of failure. 1141 * 1142 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1143 * Description 1144 * Walk a user or a kernel stack and return its id. To achieve 1145 * this, the helper needs *ctx*, which is a pointer to the context 1146 * on which the tracing program is executed, and a pointer to a 1147 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1148 * 1149 * The last argument, *flags*, holds the number of stack frames to 1150 * skip (from 0 to 255), masked with 1151 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1152 * a combination of the following flags: 1153 * 1154 * **BPF_F_USER_STACK** 1155 * Collect a user space stack instead of a kernel stack. 1156 * **BPF_F_FAST_STACK_CMP** 1157 * Compare stacks by hash only. 1158 * **BPF_F_REUSE_STACKID** 1159 * If two different stacks hash into the same *stackid*, 1160 * discard the old one. 1161 * 1162 * The stack id retrieved is a 32 bit long integer handle which 1163 * can be further combined with other data (including other stack 1164 * ids) and used as a key into maps. This can be useful for 1165 * generating a variety of graphs (such as flame graphs or off-cpu 1166 * graphs). 1167 * 1168 * For walking a stack, this helper is an improvement over 1169 * **bpf_probe_read**\ (), which can be used with unrolled loops 1170 * but is not efficient and consumes a lot of eBPF instructions. 1171 * Instead, **bpf_get_stackid**\ () can collect up to 1172 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1173 * this limit can be controlled with the **sysctl** program, and 1174 * that it should be manually increased in order to profile long 1175 * user stacks (such as stacks for Java programs). To do so, use: 1176 * 1177 * :: 1178 * 1179 * # sysctl kernel.perf_event_max_stack=<new value> 1180 * Return 1181 * The positive or null stack id on success, or a negative error 1182 * in case of failure. 1183 * 1184 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1185 * Description 1186 * Compute a checksum difference, from the raw buffer pointed by 1187 * *from*, of length *from_size* (that must be a multiple of 4), 1188 * towards the raw buffer pointed by *to*, of size *to_size* 1189 * (same remark). An optional *seed* can be added to the value 1190 * (this can be cascaded, the seed may come from a previous call 1191 * to the helper). 1192 * 1193 * This is flexible enough to be used in several ways: 1194 * 1195 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1196 * checksum, it can be used when pushing new data. 1197 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1198 * checksum, it can be used when removing data from a packet. 1199 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1200 * can be used to compute a diff. Note that *from_size* and 1201 * *to_size* do not need to be equal. 1202 * 1203 * This helper can be used in combination with 1204 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1205 * which one can feed in the difference computed with 1206 * **bpf_csum_diff**\ (). 1207 * Return 1208 * The checksum result, or a negative error code in case of 1209 * failure. 1210 * 1211 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1212 * Description 1213 * Retrieve tunnel options metadata for the packet associated to 1214 * *skb*, and store the raw tunnel option data to the buffer *opt* 1215 * of *size*. 1216 * 1217 * This helper can be used with encapsulation devices that can 1218 * operate in "collect metadata" mode (please refer to the related 1219 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1220 * more details). A particular example where this can be used is 1221 * in combination with the Geneve encapsulation protocol, where it 1222 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1223 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1224 * the eBPF program. This allows for full customization of these 1225 * headers. 1226 * Return 1227 * The size of the option data retrieved. 1228 * 1229 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1230 * Description 1231 * Set tunnel options metadata for the packet associated to *skb* 1232 * to the option data contained in the raw buffer *opt* of *size*. 1233 * 1234 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1235 * helper for additional information. 1236 * Return 1237 * 0 on success, or a negative error in case of failure. 1238 * 1239 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1240 * Description 1241 * Change the protocol of the *skb* to *proto*. Currently 1242 * supported are transition from IPv4 to IPv6, and from IPv6 to 1243 * IPv4. The helper takes care of the groundwork for the 1244 * transition, including resizing the socket buffer. The eBPF 1245 * program is expected to fill the new headers, if any, via 1246 * **skb_store_bytes**\ () and to recompute the checksums with 1247 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1248 * (). The main case for this helper is to perform NAT64 1249 * operations out of an eBPF program. 1250 * 1251 * Internally, the GSO type is marked as dodgy so that headers are 1252 * checked and segments are recalculated by the GSO/GRO engine. 1253 * The size for GSO target is adapted as well. 1254 * 1255 * All values for *flags* are reserved for future usage, and must 1256 * be left at zero. 1257 * 1258 * A call to this helper is susceptible to change the underlying 1259 * packet buffer. Therefore, at load time, all checks on pointers 1260 * previously done by the verifier are invalidated and must be 1261 * performed again, if the helper is used in combination with 1262 * direct packet access. 1263 * Return 1264 * 0 on success, or a negative error in case of failure. 1265 * 1266 * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 1267 * Description 1268 * Change the packet type for the packet associated to *skb*. This 1269 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1270 * the eBPF program does not have a write access to *skb*\ 1271 * **->pkt_type** beside this helper. Using a helper here allows 1272 * for graceful handling of errors. 1273 * 1274 * The major use case is to change incoming *skb*s to 1275 * **PACKET_HOST** in a programmatic way instead of having to 1276 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1277 * example. 1278 * 1279 * Note that *type* only allows certain values. At this time, they 1280 * are: 1281 * 1282 * **PACKET_HOST** 1283 * Packet is for us. 1284 * **PACKET_BROADCAST** 1285 * Send packet to all. 1286 * **PACKET_MULTICAST** 1287 * Send packet to group. 1288 * **PACKET_OTHERHOST** 1289 * Send packet to someone else. 1290 * Return 1291 * 0 on success, or a negative error in case of failure. 1292 * 1293 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1294 * Description 1295 * Check whether *skb* is a descendant of the cgroup2 held by 1296 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1297 * Return 1298 * The return value depends on the result of the test, and can be: 1299 * 1300 * * 0, if the *skb* failed the cgroup2 descendant test. 1301 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1302 * * A negative error code, if an error occurred. 1303 * 1304 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1305 * Description 1306 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1307 * not set, in particular if the hash was cleared due to mangling, 1308 * recompute this hash. Later accesses to the hash can be done 1309 * directly with *skb*\ **->hash**. 1310 * 1311 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1312 * prototype with **bpf_skb_change_proto**\ (), or calling 1313 * **bpf_skb_store_bytes**\ () with the 1314 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1315 * the hash and to trigger a new computation for the next call to 1316 * **bpf_get_hash_recalc**\ (). 1317 * Return 1318 * The 32-bit hash. 1319 * 1320 * u64 bpf_get_current_task(void) 1321 * Return 1322 * A pointer to the current task struct. 1323 * 1324 * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1325 * Description 1326 * Attempt in a safe way to write *len* bytes from the buffer 1327 * *src* to *dst* in memory. It only works for threads that are in 1328 * user context, and *dst* must be a valid user space address. 1329 * 1330 * This helper should not be used to implement any kind of 1331 * security mechanism because of TOC-TOU attacks, but rather to 1332 * debug, divert, and manipulate execution of semi-cooperative 1333 * processes. 1334 * 1335 * Keep in mind that this feature is meant for experiments, and it 1336 * has a risk of crashing the system and running programs. 1337 * Therefore, when an eBPF program using this helper is attached, 1338 * a warning including PID and process name is printed to kernel 1339 * logs. 1340 * Return 1341 * 0 on success, or a negative error in case of failure. 1342 * 1343 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1344 * Description 1345 * Check whether the probe is being run is the context of a given 1346 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1347 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1348 * Return 1349 * The return value depends on the result of the test, and can be: 1350 * 1351 * * 0, if the *skb* task belongs to the cgroup2. 1352 * * 1, if the *skb* task does not belong to the cgroup2. 1353 * * A negative error code, if an error occurred. 1354 * 1355 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1356 * Description 1357 * Resize (trim or grow) the packet associated to *skb* to the 1358 * new *len*. The *flags* are reserved for future usage, and must 1359 * be left at zero. 1360 * 1361 * The basic idea is that the helper performs the needed work to 1362 * change the size of the packet, then the eBPF program rewrites 1363 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1364 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1365 * and others. This helper is a slow path utility intended for 1366 * replies with control messages. And because it is targeted for 1367 * slow path, the helper itself can afford to be slow: it 1368 * implicitly linearizes, unclones and drops offloads from the 1369 * *skb*. 1370 * 1371 * A call to this helper is susceptible to change the underlying 1372 * packet buffer. Therefore, at load time, all checks on pointers 1373 * previously done by the verifier are invalidated and must be 1374 * performed again, if the helper is used in combination with 1375 * direct packet access. 1376 * Return 1377 * 0 on success, or a negative error in case of failure. 1378 * 1379 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1380 * Description 1381 * Pull in non-linear data in case the *skb* is non-linear and not 1382 * all of *len* are part of the linear section. Make *len* bytes 1383 * from *skb* readable and writable. If a zero value is passed for 1384 * *len*, then the whole length of the *skb* is pulled. 1385 * 1386 * This helper is only needed for reading and writing with direct 1387 * packet access. 1388 * 1389 * For direct packet access, testing that offsets to access 1390 * are within packet boundaries (test on *skb*\ **->data_end**) is 1391 * susceptible to fail if offsets are invalid, or if the requested 1392 * data is in non-linear parts of the *skb*. On failure the 1393 * program can just bail out, or in the case of a non-linear 1394 * buffer, use a helper to make the data available. The 1395 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1396 * the data. Another one consists in using **bpf_skb_pull_data** 1397 * to pull in once the non-linear parts, then retesting and 1398 * eventually access the data. 1399 * 1400 * At the same time, this also makes sure the *skb* is uncloned, 1401 * which is a necessary condition for direct write. As this needs 1402 * to be an invariant for the write part only, the verifier 1403 * detects writes and adds a prologue that is calling 1404 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1405 * the very beginning in case it is indeed cloned. 1406 * 1407 * A call to this helper is susceptible to change the underlying 1408 * packet buffer. Therefore, at load time, all checks on pointers 1409 * previously done by the verifier are invalidated and must be 1410 * performed again, if the helper is used in combination with 1411 * direct packet access. 1412 * Return 1413 * 0 on success, or a negative error in case of failure. 1414 * 1415 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1416 * Description 1417 * Add the checksum *csum* into *skb*\ **->csum** in case the 1418 * driver has supplied a checksum for the entire packet into that 1419 * field. Return an error otherwise. This helper is intended to be 1420 * used in combination with **bpf_csum_diff**\ (), in particular 1421 * when the checksum needs to be updated after data has been 1422 * written into the packet through direct packet access. 1423 * Return 1424 * The checksum on success, or a negative error code in case of 1425 * failure. 1426 * 1427 * void bpf_set_hash_invalid(struct sk_buff *skb) 1428 * Description 1429 * Invalidate the current *skb*\ **->hash**. It can be used after 1430 * mangling on headers through direct packet access, in order to 1431 * indicate that the hash is outdated and to trigger a 1432 * recalculation the next time the kernel tries to access this 1433 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1434 * 1435 * int bpf_get_numa_node_id(void) 1436 * Description 1437 * Return the id of the current NUMA node. The primary use case 1438 * for this helper is the selection of sockets for the local NUMA 1439 * node, when the program is attached to sockets using the 1440 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1441 * but the helper is also available to other eBPF program types, 1442 * similarly to **bpf_get_smp_processor_id**\ (). 1443 * Return 1444 * The id of current NUMA node. 1445 * 1446 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1447 * Description 1448 * Grows headroom of packet associated to *skb* and adjusts the 1449 * offset of the MAC header accordingly, adding *len* bytes of 1450 * space. It automatically extends and reallocates memory as 1451 * required. 1452 * 1453 * This helper can be used on a layer 3 *skb* to push a MAC header 1454 * for redirection into a layer 2 device. 1455 * 1456 * All values for *flags* are reserved for future usage, and must 1457 * be left at zero. 1458 * 1459 * A call to this helper is susceptible to change the underlying 1460 * packet buffer. Therefore, at load time, all checks on pointers 1461 * previously done by the verifier are invalidated and must be 1462 * performed again, if the helper is used in combination with 1463 * direct packet access. 1464 * Return 1465 * 0 on success, or a negative error in case of failure. 1466 * 1467 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1468 * Description 1469 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1470 * it is possible to use a negative value for *delta*. This helper 1471 * can be used to prepare the packet for pushing or popping 1472 * headers. 1473 * 1474 * A call to this helper is susceptible to change the underlying 1475 * packet buffer. Therefore, at load time, all checks on pointers 1476 * previously done by the verifier are invalidated and must be 1477 * performed again, if the helper is used in combination with 1478 * direct packet access. 1479 * Return 1480 * 0 on success, or a negative error in case of failure. 1481 * 1482 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1483 * Description 1484 * Copy a NUL terminated string from an unsafe kernel address 1485 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for 1486 * more details. 1487 * 1488 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str() 1489 * instead. 1490 * Return 1491 * On success, the strictly positive length of the string, 1492 * including the trailing NUL character. On error, a negative 1493 * value. 1494 * 1495 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1496 * Description 1497 * If the **struct sk_buff** pointed by *skb* has a known socket, 1498 * retrieve the cookie (generated by the kernel) of this socket. 1499 * If no cookie has been set yet, generate a new cookie. Once 1500 * generated, the socket cookie remains stable for the life of the 1501 * socket. This helper can be useful for monitoring per socket 1502 * networking traffic statistics as it provides a global socket 1503 * identifier that can be assumed unique. 1504 * Return 1505 * A 8-byte long non-decreasing number on success, or 0 if the 1506 * socket field is missing inside *skb*. 1507 * 1508 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1509 * Description 1510 * Equivalent to bpf_get_socket_cookie() helper that accepts 1511 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1512 * Return 1513 * A 8-byte long non-decreasing number. 1514 * 1515 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1516 * Description 1517 * Equivalent to bpf_get_socket_cookie() helper that accepts 1518 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1519 * Return 1520 * A 8-byte long non-decreasing number. 1521 * 1522 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1523 * Return 1524 * The owner UID of the socket associated to *skb*. If the socket 1525 * is **NULL**, or if it is not a full socket (i.e. if it is a 1526 * time-wait or a request socket instead), **overflowuid** value 1527 * is returned (note that **overflowuid** might also be the actual 1528 * UID value for the socket). 1529 * 1530 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1531 * Description 1532 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1533 * to value *hash*. 1534 * Return 1535 * 0 1536 * 1537 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1538 * Description 1539 * Emulate a call to **setsockopt()** on the socket associated to 1540 * *bpf_socket*, which must be a full socket. The *level* at 1541 * which the option resides and the name *optname* of the option 1542 * must be specified, see **setsockopt(2)** for more information. 1543 * The option value of length *optlen* is pointed by *optval*. 1544 * 1545 * This helper actually implements a subset of **setsockopt()**. 1546 * It supports the following *level*\ s: 1547 * 1548 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1549 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1550 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 1551 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1552 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1553 * **TCP_BPF_SNDCWND_CLAMP**. 1554 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1555 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1556 * Return 1557 * 0 on success, or a negative error in case of failure. 1558 * 1559 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1560 * Description 1561 * Grow or shrink the room for data in the packet associated to 1562 * *skb* by *len_diff*, and according to the selected *mode*. 1563 * 1564 * There are two supported modes at this time: 1565 * 1566 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1567 * (room space is added or removed below the layer 2 header). 1568 * 1569 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1570 * (room space is added or removed below the layer 3 header). 1571 * 1572 * The following flags are supported at this time: 1573 * 1574 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1575 * Adjusting mss in this way is not allowed for datagrams. 1576 * 1577 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1578 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1579 * Any new space is reserved to hold a tunnel header. 1580 * Configure skb offsets and other fields accordingly. 1581 * 1582 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1583 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1584 * Use with ENCAP_L3 flags to further specify the tunnel type. 1585 * 1586 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1587 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1588 * type; *len* is the length of the inner MAC header. 1589 * 1590 * A call to this helper is susceptible to change the underlying 1591 * packet buffer. Therefore, at load time, all checks on pointers 1592 * previously done by the verifier are invalidated and must be 1593 * performed again, if the helper is used in combination with 1594 * direct packet access. 1595 * Return 1596 * 0 on success, or a negative error in case of failure. 1597 * 1598 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1599 * Description 1600 * Redirect the packet to the endpoint referenced by *map* at 1601 * index *key*. Depending on its type, this *map* can contain 1602 * references to net devices (for forwarding packets through other 1603 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1604 * but this is only implemented for native XDP (with driver 1605 * support) as of this writing). 1606 * 1607 * The lower two bits of *flags* are used as the return code if 1608 * the map lookup fails. This is so that the return value can be 1609 * one of the XDP program return codes up to XDP_TX, as chosen by 1610 * the caller. Any higher bits in the *flags* argument must be 1611 * unset. 1612 * 1613 * When used to redirect packets to net devices, this helper 1614 * provides a high performance increase over **bpf_redirect**\ (). 1615 * This is due to various implementation details of the underlying 1616 * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1617 * () tries to send packet as a "bulk" to the device. 1618 * Return 1619 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1620 * 1621 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1622 * Description 1623 * Redirect the packet to the socket referenced by *map* (of type 1624 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1625 * egress interfaces can be used for redirection. The 1626 * **BPF_F_INGRESS** value in *flags* is used to make the 1627 * distinction (ingress path is selected if the flag is present, 1628 * egress path otherwise). This is the only flag supported for now. 1629 * Return 1630 * **SK_PASS** on success, or **SK_DROP** on error. 1631 * 1632 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1633 * Description 1634 * Add an entry to, or update a *map* referencing sockets. The 1635 * *skops* is used as a new value for the entry associated to 1636 * *key*. *flags* is one of: 1637 * 1638 * **BPF_NOEXIST** 1639 * The entry for *key* must not exist in the map. 1640 * **BPF_EXIST** 1641 * The entry for *key* must already exist in the map. 1642 * **BPF_ANY** 1643 * No condition on the existence of the entry for *key*. 1644 * 1645 * If the *map* has eBPF programs (parser and verdict), those will 1646 * be inherited by the socket being added. If the socket is 1647 * already attached to eBPF programs, this results in an error. 1648 * Return 1649 * 0 on success, or a negative error in case of failure. 1650 * 1651 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1652 * Description 1653 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1654 * *delta* (which can be positive or negative). Note that this 1655 * operation modifies the address stored in *xdp_md*\ **->data**, 1656 * so the latter must be loaded only after the helper has been 1657 * called. 1658 * 1659 * The use of *xdp_md*\ **->data_meta** is optional and programs 1660 * are not required to use it. The rationale is that when the 1661 * packet is processed with XDP (e.g. as DoS filter), it is 1662 * possible to push further meta data along with it before passing 1663 * to the stack, and to give the guarantee that an ingress eBPF 1664 * program attached as a TC classifier on the same device can pick 1665 * this up for further post-processing. Since TC works with socket 1666 * buffers, it remains possible to set from XDP the **mark** or 1667 * **priority** pointers, or other pointers for the socket buffer. 1668 * Having this scratch space generic and programmable allows for 1669 * more flexibility as the user is free to store whatever meta 1670 * data they need. 1671 * 1672 * A call to this helper is susceptible to change the underlying 1673 * packet buffer. Therefore, at load time, all checks on pointers 1674 * previously done by the verifier are invalidated and must be 1675 * performed again, if the helper is used in combination with 1676 * direct packet access. 1677 * Return 1678 * 0 on success, or a negative error in case of failure. 1679 * 1680 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1681 * Description 1682 * Read the value of a perf event counter, and store it into *buf* 1683 * of size *buf_size*. This helper relies on a *map* of type 1684 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1685 * counter is selected when *map* is updated with perf event file 1686 * descriptors. The *map* is an array whose size is the number of 1687 * available CPUs, and each cell contains a value relative to one 1688 * CPU. The value to retrieve is indicated by *flags*, that 1689 * contains the index of the CPU to look up, masked with 1690 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1691 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1692 * current CPU should be retrieved. 1693 * 1694 * This helper behaves in a way close to 1695 * **bpf_perf_event_read**\ () helper, save that instead of 1696 * just returning the value observed, it fills the *buf* 1697 * structure. This allows for additional data to be retrieved: in 1698 * particular, the enabled and running times (in *buf*\ 1699 * **->enabled** and *buf*\ **->running**, respectively) are 1700 * copied. In general, **bpf_perf_event_read_value**\ () is 1701 * recommended over **bpf_perf_event_read**\ (), which has some 1702 * ABI issues and provides fewer functionalities. 1703 * 1704 * These values are interesting, because hardware PMU (Performance 1705 * Monitoring Unit) counters are limited resources. When there are 1706 * more PMU based perf events opened than available counters, 1707 * kernel will multiplex these events so each event gets certain 1708 * percentage (but not all) of the PMU time. In case that 1709 * multiplexing happens, the number of samples or counter value 1710 * will not reflect the case compared to when no multiplexing 1711 * occurs. This makes comparison between different runs difficult. 1712 * Typically, the counter value should be normalized before 1713 * comparing to other experiments. The usual normalization is done 1714 * as follows. 1715 * 1716 * :: 1717 * 1718 * normalized_counter = counter * t_enabled / t_running 1719 * 1720 * Where t_enabled is the time enabled for event and t_running is 1721 * the time running for event since last normalization. The 1722 * enabled and running times are accumulated since the perf event 1723 * open. To achieve scaling factor between two invocations of an 1724 * eBPF program, users can can use CPU id as the key (which is 1725 * typical for perf array usage model) to remember the previous 1726 * value and do the calculation inside the eBPF program. 1727 * Return 1728 * 0 on success, or a negative error in case of failure. 1729 * 1730 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1731 * Description 1732 * For en eBPF program attached to a perf event, retrieve the 1733 * value of the event counter associated to *ctx* and store it in 1734 * the structure pointed by *buf* and of size *buf_size*. Enabled 1735 * and running times are also stored in the structure (see 1736 * description of helper **bpf_perf_event_read_value**\ () for 1737 * more details). 1738 * Return 1739 * 0 on success, or a negative error in case of failure. 1740 * 1741 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen) 1742 * Description 1743 * Emulate a call to **getsockopt()** on the socket associated to 1744 * *bpf_socket*, which must be a full socket. The *level* at 1745 * which the option resides and the name *optname* of the option 1746 * must be specified, see **getsockopt(2)** for more information. 1747 * The retrieved value is stored in the structure pointed by 1748 * *opval* and of length *optlen*. 1749 * 1750 * This helper actually implements a subset of **getsockopt()**. 1751 * It supports the following *level*\ s: 1752 * 1753 * * **IPPROTO_TCP**, which supports *optname* 1754 * **TCP_CONGESTION**. 1755 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1756 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1757 * Return 1758 * 0 on success, or a negative error in case of failure. 1759 * 1760 * int bpf_override_return(struct pt_regs *regs, u64 rc) 1761 * Description 1762 * Used for error injection, this helper uses kprobes to override 1763 * the return value of the probed function, and to set it to *rc*. 1764 * The first argument is the context *regs* on which the kprobe 1765 * works. 1766 * 1767 * This helper works by setting setting the PC (program counter) 1768 * to an override function which is run in place of the original 1769 * probed function. This means the probed function is not run at 1770 * all. The replacement function just returns with the required 1771 * value. 1772 * 1773 * This helper has security implications, and thus is subject to 1774 * restrictions. It is only available if the kernel was compiled 1775 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1776 * option, and in this case it only works on functions tagged with 1777 * **ALLOW_ERROR_INJECTION** in the kernel code. 1778 * 1779 * Also, the helper is only available for the architectures having 1780 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1781 * x86 architecture is the only one to support this feature. 1782 * Return 1783 * 0 1784 * 1785 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1786 * Description 1787 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1788 * for the full TCP socket associated to *bpf_sock_ops* to 1789 * *argval*. 1790 * 1791 * The primary use of this field is to determine if there should 1792 * be calls to eBPF programs of type 1793 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1794 * code. A program of the same type can change its value, per 1795 * connection and as necessary, when the connection is 1796 * established. This field is directly accessible for reading, but 1797 * this helper must be used for updates in order to return an 1798 * error if an eBPF program tries to set a callback that is not 1799 * supported in the current kernel. 1800 * 1801 * *argval* is a flag array which can combine these flags: 1802 * 1803 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1804 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1805 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1806 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1807 * 1808 * Therefore, this function can be used to clear a callback flag by 1809 * setting the appropriate bit to zero. e.g. to disable the RTO 1810 * callback: 1811 * 1812 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1813 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1814 * 1815 * Here are some examples of where one could call such eBPF 1816 * program: 1817 * 1818 * * When RTO fires. 1819 * * When a packet is retransmitted. 1820 * * When the connection terminates. 1821 * * When a packet is sent. 1822 * * When a packet is received. 1823 * Return 1824 * Code **-EINVAL** if the socket is not a full TCP socket; 1825 * otherwise, a positive number containing the bits that could not 1826 * be set is returned (which comes down to 0 if all bits were set 1827 * as required). 1828 * 1829 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1830 * Description 1831 * This helper is used in programs implementing policies at the 1832 * socket level. If the message *msg* is allowed to pass (i.e. if 1833 * the verdict eBPF program returns **SK_PASS**), redirect it to 1834 * the socket referenced by *map* (of type 1835 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1836 * egress interfaces can be used for redirection. The 1837 * **BPF_F_INGRESS** value in *flags* is used to make the 1838 * distinction (ingress path is selected if the flag is present, 1839 * egress path otherwise). This is the only flag supported for now. 1840 * Return 1841 * **SK_PASS** on success, or **SK_DROP** on error. 1842 * 1843 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1844 * Description 1845 * For socket policies, apply the verdict of the eBPF program to 1846 * the next *bytes* (number of bytes) of message *msg*. 1847 * 1848 * For example, this helper can be used in the following cases: 1849 * 1850 * * A single **sendmsg**\ () or **sendfile**\ () system call 1851 * contains multiple logical messages that the eBPF program is 1852 * supposed to read and for which it should apply a verdict. 1853 * * An eBPF program only cares to read the first *bytes* of a 1854 * *msg*. If the message has a large payload, then setting up 1855 * and calling the eBPF program repeatedly for all bytes, even 1856 * though the verdict is already known, would create unnecessary 1857 * overhead. 1858 * 1859 * When called from within an eBPF program, the helper sets a 1860 * counter internal to the BPF infrastructure, that is used to 1861 * apply the last verdict to the next *bytes*. If *bytes* is 1862 * smaller than the current data being processed from a 1863 * **sendmsg**\ () or **sendfile**\ () system call, the first 1864 * *bytes* will be sent and the eBPF program will be re-run with 1865 * the pointer for start of data pointing to byte number *bytes* 1866 * **+ 1**. If *bytes* is larger than the current data being 1867 * processed, then the eBPF verdict will be applied to multiple 1868 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1869 * consumed. 1870 * 1871 * Note that if a socket closes with the internal counter holding 1872 * a non-zero value, this is not a problem because data is not 1873 * being buffered for *bytes* and is sent as it is received. 1874 * Return 1875 * 0 1876 * 1877 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1878 * Description 1879 * For socket policies, prevent the execution of the verdict eBPF 1880 * program for message *msg* until *bytes* (byte number) have been 1881 * accumulated. 1882 * 1883 * This can be used when one needs a specific number of bytes 1884 * before a verdict can be assigned, even if the data spans 1885 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1886 * case would be a user calling **sendmsg**\ () repeatedly with 1887 * 1-byte long message segments. Obviously, this is bad for 1888 * performance, but it is still valid. If the eBPF program needs 1889 * *bytes* bytes to validate a header, this helper can be used to 1890 * prevent the eBPF program to be called again until *bytes* have 1891 * been accumulated. 1892 * Return 1893 * 0 1894 * 1895 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1896 * Description 1897 * For socket policies, pull in non-linear data from user space 1898 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1899 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1900 * respectively. 1901 * 1902 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1903 * *msg* it can only parse data that the (**data**, **data_end**) 1904 * pointers have already consumed. For **sendmsg**\ () hooks this 1905 * is likely the first scatterlist element. But for calls relying 1906 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1907 * be the range (**0**, **0**) because the data is shared with 1908 * user space and by default the objective is to avoid allowing 1909 * user space to modify data while (or after) eBPF verdict is 1910 * being decided. This helper can be used to pull in data and to 1911 * set the start and end pointer to given values. Data will be 1912 * copied if necessary (i.e. if data was not linear and if start 1913 * and end pointers do not point to the same chunk). 1914 * 1915 * A call to this helper is susceptible to change the underlying 1916 * packet buffer. Therefore, at load time, all checks on pointers 1917 * previously done by the verifier are invalidated and must be 1918 * performed again, if the helper is used in combination with 1919 * direct packet access. 1920 * 1921 * All values for *flags* are reserved for future usage, and must 1922 * be left at zero. 1923 * Return 1924 * 0 on success, or a negative error in case of failure. 1925 * 1926 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 1927 * Description 1928 * Bind the socket associated to *ctx* to the address pointed by 1929 * *addr*, of length *addr_len*. This allows for making outgoing 1930 * connection from the desired IP address, which can be useful for 1931 * example when all processes inside a cgroup should use one 1932 * single IP address on a host that has multiple IP configured. 1933 * 1934 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1935 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1936 * **AF_INET6**). Looking for a free port to bind to can be 1937 * expensive, therefore binding to port is not permitted by the 1938 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 1939 * must be set to zero. 1940 * Return 1941 * 0 on success, or a negative error in case of failure. 1942 * 1943 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 1944 * Description 1945 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1946 * only possible to shrink the packet as of this writing, 1947 * therefore *delta* must be a negative integer. 1948 * 1949 * A call to this helper is susceptible to change the underlying 1950 * packet buffer. Therefore, at load time, all checks on pointers 1951 * previously done by the verifier are invalidated and must be 1952 * performed again, if the helper is used in combination with 1953 * direct packet access. 1954 * Return 1955 * 0 on success, or a negative error in case of failure. 1956 * 1957 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 1958 * Description 1959 * Retrieve the XFRM state (IP transform framework, see also 1960 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1961 * 1962 * The retrieved value is stored in the **struct bpf_xfrm_state** 1963 * pointed by *xfrm_state* and of length *size*. 1964 * 1965 * All values for *flags* are reserved for future usage, and must 1966 * be left at zero. 1967 * 1968 * This helper is available only if the kernel was compiled with 1969 * **CONFIG_XFRM** configuration option. 1970 * Return 1971 * 0 on success, or a negative error in case of failure. 1972 * 1973 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 1974 * Description 1975 * Return a user or a kernel stack in bpf program provided buffer. 1976 * To achieve this, the helper needs *ctx*, which is a pointer 1977 * to the context on which the tracing program is executed. 1978 * To store the stacktrace, the bpf program provides *buf* with 1979 * a nonnegative *size*. 1980 * 1981 * The last argument, *flags*, holds the number of stack frames to 1982 * skip (from 0 to 255), masked with 1983 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1984 * the following flags: 1985 * 1986 * **BPF_F_USER_STACK** 1987 * Collect a user space stack instead of a kernel stack. 1988 * **BPF_F_USER_BUILD_ID** 1989 * Collect buildid+offset instead of ips for user stack, 1990 * only valid if **BPF_F_USER_STACK** is also specified. 1991 * 1992 * **bpf_get_stack**\ () can collect up to 1993 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1994 * to sufficient large buffer size. Note that 1995 * this limit can be controlled with the **sysctl** program, and 1996 * that it should be manually increased in order to profile long 1997 * user stacks (such as stacks for Java programs). To do so, use: 1998 * 1999 * :: 2000 * 2001 * # sysctl kernel.perf_event_max_stack=<new value> 2002 * Return 2003 * A non-negative value equal to or less than *size* on success, 2004 * or a negative error in case of failure. 2005 * 2006 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2007 * Description 2008 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2009 * it provides an easy way to load *len* bytes from *offset* 2010 * from the packet associated to *skb*, into the buffer pointed 2011 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2012 * a fifth argument *start_header* exists in order to select a 2013 * base offset to start from. *start_header* can be one of: 2014 * 2015 * **BPF_HDR_START_MAC** 2016 * Base offset to load data from is *skb*'s mac header. 2017 * **BPF_HDR_START_NET** 2018 * Base offset to load data from is *skb*'s network header. 2019 * 2020 * In general, "direct packet access" is the preferred method to 2021 * access packet data, however, this helper is in particular useful 2022 * in socket filters where *skb*\ **->data** does not always point 2023 * to the start of the mac header and where "direct packet access" 2024 * is not available. 2025 * Return 2026 * 0 on success, or a negative error in case of failure. 2027 * 2028 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2029 * Description 2030 * Do FIB lookup in kernel tables using parameters in *params*. 2031 * If lookup is successful and result shows packet is to be 2032 * forwarded, the neighbor tables are searched for the nexthop. 2033 * If successful (ie., FIB lookup shows forwarding and nexthop 2034 * is resolved), the nexthop address is returned in ipv4_dst 2035 * or ipv6_dst based on family, smac is set to mac address of 2036 * egress device, dmac is set to nexthop mac address, rt_metric 2037 * is set to metric from route (IPv4/IPv6 only), and ifindex 2038 * is set to the device index of the nexthop from the FIB lookup. 2039 * 2040 * *plen* argument is the size of the passed in struct. 2041 * *flags* argument can be a combination of one or more of the 2042 * following values: 2043 * 2044 * **BPF_FIB_LOOKUP_DIRECT** 2045 * Do a direct table lookup vs full lookup using FIB 2046 * rules. 2047 * **BPF_FIB_LOOKUP_OUTPUT** 2048 * Perform lookup from an egress perspective (default is 2049 * ingress). 2050 * 2051 * *ctx* is either **struct xdp_md** for XDP programs or 2052 * **struct sk_buff** tc cls_act programs. 2053 * Return 2054 * * < 0 if any input argument is invalid 2055 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2056 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2057 * packet is not forwarded or needs assist from full stack 2058 * 2059 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2060 * Description 2061 * Add an entry to, or update a sockhash *map* referencing sockets. 2062 * The *skops* is used as a new value for the entry associated to 2063 * *key*. *flags* is one of: 2064 * 2065 * **BPF_NOEXIST** 2066 * The entry for *key* must not exist in the map. 2067 * **BPF_EXIST** 2068 * The entry for *key* must already exist in the map. 2069 * **BPF_ANY** 2070 * No condition on the existence of the entry for *key*. 2071 * 2072 * If the *map* has eBPF programs (parser and verdict), those will 2073 * be inherited by the socket being added. If the socket is 2074 * already attached to eBPF programs, this results in an error. 2075 * Return 2076 * 0 on success, or a negative error in case of failure. 2077 * 2078 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2079 * Description 2080 * This helper is used in programs implementing policies at the 2081 * socket level. If the message *msg* is allowed to pass (i.e. if 2082 * the verdict eBPF program returns **SK_PASS**), redirect it to 2083 * the socket referenced by *map* (of type 2084 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2085 * egress interfaces can be used for redirection. The 2086 * **BPF_F_INGRESS** value in *flags* is used to make the 2087 * distinction (ingress path is selected if the flag is present, 2088 * egress path otherwise). This is the only flag supported for now. 2089 * Return 2090 * **SK_PASS** on success, or **SK_DROP** on error. 2091 * 2092 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2093 * Description 2094 * This helper is used in programs implementing policies at the 2095 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2096 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2097 * to the socket referenced by *map* (of type 2098 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2099 * egress interfaces can be used for redirection. The 2100 * **BPF_F_INGRESS** value in *flags* is used to make the 2101 * distinction (ingress path is selected if the flag is present, 2102 * egress otherwise). This is the only flag supported for now. 2103 * Return 2104 * **SK_PASS** on success, or **SK_DROP** on error. 2105 * 2106 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2107 * Description 2108 * Encapsulate the packet associated to *skb* within a Layer 3 2109 * protocol header. This header is provided in the buffer at 2110 * address *hdr*, with *len* its size in bytes. *type* indicates 2111 * the protocol of the header and can be one of: 2112 * 2113 * **BPF_LWT_ENCAP_SEG6** 2114 * IPv6 encapsulation with Segment Routing Header 2115 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2116 * the IPv6 header is computed by the kernel. 2117 * **BPF_LWT_ENCAP_SEG6_INLINE** 2118 * Only works if *skb* contains an IPv6 packet. Insert a 2119 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2120 * the IPv6 header. 2121 * **BPF_LWT_ENCAP_IP** 2122 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2123 * must be IPv4 or IPv6, followed by zero or more 2124 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2125 * total bytes in all prepended headers. Please note that 2126 * if **skb_is_gso**\ (*skb*) is true, no more than two 2127 * headers can be prepended, and the inner header, if 2128 * present, should be either GRE or UDP/GUE. 2129 * 2130 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2131 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2132 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2133 * **BPF_PROG_TYPE_LWT_XMIT**. 2134 * 2135 * A call to this helper is susceptible to change the underlying 2136 * packet buffer. Therefore, at load time, all checks on pointers 2137 * previously done by the verifier are invalidated and must be 2138 * performed again, if the helper is used in combination with 2139 * direct packet access. 2140 * Return 2141 * 0 on success, or a negative error in case of failure. 2142 * 2143 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2144 * Description 2145 * Store *len* bytes from address *from* into the packet 2146 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2147 * inside the outermost IPv6 Segment Routing Header can be 2148 * modified through this helper. 2149 * 2150 * A call to this helper is susceptible to change the underlying 2151 * packet buffer. Therefore, at load time, all checks on pointers 2152 * previously done by the verifier are invalidated and must be 2153 * performed again, if the helper is used in combination with 2154 * direct packet access. 2155 * Return 2156 * 0 on success, or a negative error in case of failure. 2157 * 2158 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2159 * Description 2160 * Adjust the size allocated to TLVs in the outermost IPv6 2161 * Segment Routing Header contained in the packet associated to 2162 * *skb*, at position *offset* by *delta* bytes. Only offsets 2163 * after the segments are accepted. *delta* can be as well 2164 * positive (growing) as negative (shrinking). 2165 * 2166 * A call to this helper is susceptible to change the underlying 2167 * packet buffer. Therefore, at load time, all checks on pointers 2168 * previously done by the verifier are invalidated and must be 2169 * performed again, if the helper is used in combination with 2170 * direct packet access. 2171 * Return 2172 * 0 on success, or a negative error in case of failure. 2173 * 2174 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2175 * Description 2176 * Apply an IPv6 Segment Routing action of type *action* to the 2177 * packet associated to *skb*. Each action takes a parameter 2178 * contained at address *param*, and of length *param_len* bytes. 2179 * *action* can be one of: 2180 * 2181 * **SEG6_LOCAL_ACTION_END_X** 2182 * End.X action: Endpoint with Layer-3 cross-connect. 2183 * Type of *param*: **struct in6_addr**. 2184 * **SEG6_LOCAL_ACTION_END_T** 2185 * End.T action: Endpoint with specific IPv6 table lookup. 2186 * Type of *param*: **int**. 2187 * **SEG6_LOCAL_ACTION_END_B6** 2188 * End.B6 action: Endpoint bound to an SRv6 policy. 2189 * Type of *param*: **struct ipv6_sr_hdr**. 2190 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2191 * End.B6.Encap action: Endpoint bound to an SRv6 2192 * encapsulation policy. 2193 * Type of *param*: **struct ipv6_sr_hdr**. 2194 * 2195 * A call to this helper is susceptible to change the underlying 2196 * packet buffer. Therefore, at load time, all checks on pointers 2197 * previously done by the verifier are invalidated and must be 2198 * performed again, if the helper is used in combination with 2199 * direct packet access. 2200 * Return 2201 * 0 on success, or a negative error in case of failure. 2202 * 2203 * int bpf_rc_repeat(void *ctx) 2204 * Description 2205 * This helper is used in programs implementing IR decoding, to 2206 * report a successfully decoded repeat key message. This delays 2207 * the generation of a key up event for previously generated 2208 * key down event. 2209 * 2210 * Some IR protocols like NEC have a special IR message for 2211 * repeating last button, for when a button is held down. 2212 * 2213 * The *ctx* should point to the lirc sample as passed into 2214 * the program. 2215 * 2216 * This helper is only available is the kernel was compiled with 2217 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2218 * "**y**". 2219 * Return 2220 * 0 2221 * 2222 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2223 * Description 2224 * This helper is used in programs implementing IR decoding, to 2225 * report a successfully decoded key press with *scancode*, 2226 * *toggle* value in the given *protocol*. The scancode will be 2227 * translated to a keycode using the rc keymap, and reported as 2228 * an input key down event. After a period a key up event is 2229 * generated. This period can be extended by calling either 2230 * **bpf_rc_keydown**\ () again with the same values, or calling 2231 * **bpf_rc_repeat**\ (). 2232 * 2233 * Some protocols include a toggle bit, in case the button was 2234 * released and pressed again between consecutive scancodes. 2235 * 2236 * The *ctx* should point to the lirc sample as passed into 2237 * the program. 2238 * 2239 * The *protocol* is the decoded protocol number (see 2240 * **enum rc_proto** for some predefined values). 2241 * 2242 * This helper is only available is the kernel was compiled with 2243 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2244 * "**y**". 2245 * Return 2246 * 0 2247 * 2248 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2249 * Description 2250 * Return the cgroup v2 id of the socket associated with the *skb*. 2251 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2252 * helper for cgroup v1 by providing a tag resp. identifier that 2253 * can be matched on or used for map lookups e.g. to implement 2254 * policy. The cgroup v2 id of a given path in the hierarchy is 2255 * exposed in user space through the f_handle API in order to get 2256 * to the same 64-bit id. 2257 * 2258 * This helper can be used on TC egress path, but not on ingress, 2259 * and is available only if the kernel was compiled with the 2260 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2261 * Return 2262 * The id is returned or 0 in case the id could not be retrieved. 2263 * 2264 * u64 bpf_get_current_cgroup_id(void) 2265 * Return 2266 * A 64-bit integer containing the current cgroup id based 2267 * on the cgroup within which the current task is running. 2268 * 2269 * void *bpf_get_local_storage(void *map, u64 flags) 2270 * Description 2271 * Get the pointer to the local storage area. 2272 * The type and the size of the local storage is defined 2273 * by the *map* argument. 2274 * The *flags* meaning is specific for each map type, 2275 * and has to be 0 for cgroup local storage. 2276 * 2277 * Depending on the BPF program type, a local storage area 2278 * can be shared between multiple instances of the BPF program, 2279 * running simultaneously. 2280 * 2281 * A user should care about the synchronization by himself. 2282 * For example, by using the **BPF_STX_XADD** instruction to alter 2283 * the shared data. 2284 * Return 2285 * A pointer to the local storage area. 2286 * 2287 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2288 * Description 2289 * Select a **SO_REUSEPORT** socket from a 2290 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2291 * It checks the selected socket is matching the incoming 2292 * request in the socket buffer. 2293 * Return 2294 * 0 on success, or a negative error in case of failure. 2295 * 2296 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2297 * Description 2298 * Return id of cgroup v2 that is ancestor of cgroup associated 2299 * with the *skb* at the *ancestor_level*. The root cgroup is at 2300 * *ancestor_level* zero and each step down the hierarchy 2301 * increments the level. If *ancestor_level* == level of cgroup 2302 * associated with *skb*, then return value will be same as that 2303 * of **bpf_skb_cgroup_id**\ (). 2304 * 2305 * The helper is useful to implement policies based on cgroups 2306 * that are upper in hierarchy than immediate cgroup associated 2307 * with *skb*. 2308 * 2309 * The format of returned id and helper limitations are same as in 2310 * **bpf_skb_cgroup_id**\ (). 2311 * Return 2312 * The id is returned or 0 in case the id could not be retrieved. 2313 * 2314 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2315 * Description 2316 * Look for TCP socket matching *tuple*, optionally in a child 2317 * network namespace *netns*. The return value must be checked, 2318 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2319 * 2320 * The *ctx* should point to the context of the program, such as 2321 * the skb or socket (depending on the hook in use). This is used 2322 * to determine the base network namespace for the lookup. 2323 * 2324 * *tuple_size* must be one of: 2325 * 2326 * **sizeof**\ (*tuple*\ **->ipv4**) 2327 * Look for an IPv4 socket. 2328 * **sizeof**\ (*tuple*\ **->ipv6**) 2329 * Look for an IPv6 socket. 2330 * 2331 * If the *netns* is a negative signed 32-bit integer, then the 2332 * socket lookup table in the netns associated with the *ctx* will 2333 * will be used. For the TC hooks, this is the netns of the device 2334 * in the skb. For socket hooks, this is the netns of the socket. 2335 * If *netns* is any other signed 32-bit value greater than or 2336 * equal to zero then it specifies the ID of the netns relative to 2337 * the netns associated with the *ctx*. *netns* values beyond the 2338 * range of 32-bit integers are reserved for future use. 2339 * 2340 * All values for *flags* are reserved for future usage, and must 2341 * be left at zero. 2342 * 2343 * This helper is available only if the kernel was compiled with 2344 * **CONFIG_NET** configuration option. 2345 * Return 2346 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2347 * For sockets with reuseport option, the **struct bpf_sock** 2348 * result is from *reuse*\ **->socks**\ [] using the hash of the 2349 * tuple. 2350 * 2351 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2352 * Description 2353 * Look for UDP socket matching *tuple*, optionally in a child 2354 * network namespace *netns*. The return value must be checked, 2355 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2356 * 2357 * The *ctx* should point to the context of the program, such as 2358 * the skb or socket (depending on the hook in use). This is used 2359 * to determine the base network namespace for the lookup. 2360 * 2361 * *tuple_size* must be one of: 2362 * 2363 * **sizeof**\ (*tuple*\ **->ipv4**) 2364 * Look for an IPv4 socket. 2365 * **sizeof**\ (*tuple*\ **->ipv6**) 2366 * Look for an IPv6 socket. 2367 * 2368 * If the *netns* is a negative signed 32-bit integer, then the 2369 * socket lookup table in the netns associated with the *ctx* will 2370 * will be used. For the TC hooks, this is the netns of the device 2371 * in the skb. For socket hooks, this is the netns of the socket. 2372 * If *netns* is any other signed 32-bit value greater than or 2373 * equal to zero then it specifies the ID of the netns relative to 2374 * the netns associated with the *ctx*. *netns* values beyond the 2375 * range of 32-bit integers are reserved for future use. 2376 * 2377 * All values for *flags* are reserved for future usage, and must 2378 * be left at zero. 2379 * 2380 * This helper is available only if the kernel was compiled with 2381 * **CONFIG_NET** configuration option. 2382 * Return 2383 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2384 * For sockets with reuseport option, the **struct bpf_sock** 2385 * result is from *reuse*\ **->socks**\ [] using the hash of the 2386 * tuple. 2387 * 2388 * int bpf_sk_release(struct bpf_sock *sock) 2389 * Description 2390 * Release the reference held by *sock*. *sock* must be a 2391 * non-**NULL** pointer that was returned from 2392 * **bpf_sk_lookup_xxx**\ (). 2393 * Return 2394 * 0 on success, or a negative error in case of failure. 2395 * 2396 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2397 * Description 2398 * Push an element *value* in *map*. *flags* is one of: 2399 * 2400 * **BPF_EXIST** 2401 * If the queue/stack is full, the oldest element is 2402 * removed to make room for this. 2403 * Return 2404 * 0 on success, or a negative error in case of failure. 2405 * 2406 * int bpf_map_pop_elem(struct bpf_map *map, void *value) 2407 * Description 2408 * Pop an element from *map*. 2409 * Return 2410 * 0 on success, or a negative error in case of failure. 2411 * 2412 * int bpf_map_peek_elem(struct bpf_map *map, void *value) 2413 * Description 2414 * Get an element from *map* without removing it. 2415 * Return 2416 * 0 on success, or a negative error in case of failure. 2417 * 2418 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2419 * Description 2420 * For socket policies, insert *len* bytes into *msg* at offset 2421 * *start*. 2422 * 2423 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2424 * *msg* it may want to insert metadata or options into the *msg*. 2425 * This can later be read and used by any of the lower layer BPF 2426 * hooks. 2427 * 2428 * This helper may fail if under memory pressure (a malloc 2429 * fails) in these cases BPF programs will get an appropriate 2430 * error and BPF programs will need to handle them. 2431 * Return 2432 * 0 on success, or a negative error in case of failure. 2433 * 2434 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2435 * Description 2436 * Will remove *len* bytes from a *msg* starting at byte *start*. 2437 * This may result in **ENOMEM** errors under certain situations if 2438 * an allocation and copy are required due to a full ring buffer. 2439 * However, the helper will try to avoid doing the allocation 2440 * if possible. Other errors can occur if input parameters are 2441 * invalid either due to *start* byte not being valid part of *msg* 2442 * payload and/or *pop* value being to large. 2443 * Return 2444 * 0 on success, or a negative error in case of failure. 2445 * 2446 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2447 * Description 2448 * This helper is used in programs implementing IR decoding, to 2449 * report a successfully decoded pointer movement. 2450 * 2451 * The *ctx* should point to the lirc sample as passed into 2452 * the program. 2453 * 2454 * This helper is only available is the kernel was compiled with 2455 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2456 * "**y**". 2457 * Return 2458 * 0 2459 * 2460 * int bpf_spin_lock(struct bpf_spin_lock *lock) 2461 * Description 2462 * Acquire a spinlock represented by the pointer *lock*, which is 2463 * stored as part of a value of a map. Taking the lock allows to 2464 * safely update the rest of the fields in that value. The 2465 * spinlock can (and must) later be released with a call to 2466 * **bpf_spin_unlock**\ (\ *lock*\ ). 2467 * 2468 * Spinlocks in BPF programs come with a number of restrictions 2469 * and constraints: 2470 * 2471 * * **bpf_spin_lock** objects are only allowed inside maps of 2472 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2473 * list could be extended in the future). 2474 * * BTF description of the map is mandatory. 2475 * * The BPF program can take ONE lock at a time, since taking two 2476 * or more could cause dead locks. 2477 * * Only one **struct bpf_spin_lock** is allowed per map element. 2478 * * When the lock is taken, calls (either BPF to BPF or helpers) 2479 * are not allowed. 2480 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2481 * allowed inside a spinlock-ed region. 2482 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2483 * the lock, on all execution paths, before it returns. 2484 * * The BPF program can access **struct bpf_spin_lock** only via 2485 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2486 * helpers. Loading or storing data into the **struct 2487 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2488 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2489 * of the map value must be a struct and have **struct 2490 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2491 * Nested lock inside another struct is not allowed. 2492 * * The **struct bpf_spin_lock** *lock* field in a map value must 2493 * be aligned on a multiple of 4 bytes in that value. 2494 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2495 * the **bpf_spin_lock** field to user space. 2496 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2497 * a BPF program, do not update the **bpf_spin_lock** field. 2498 * * **bpf_spin_lock** cannot be on the stack or inside a 2499 * networking packet (it can only be inside of a map values). 2500 * * **bpf_spin_lock** is available to root only. 2501 * * Tracing programs and socket filter programs cannot use 2502 * **bpf_spin_lock**\ () due to insufficient preemption checks 2503 * (but this may change in the future). 2504 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2505 * Return 2506 * 0 2507 * 2508 * int bpf_spin_unlock(struct bpf_spin_lock *lock) 2509 * Description 2510 * Release the *lock* previously locked by a call to 2511 * **bpf_spin_lock**\ (\ *lock*\ ). 2512 * Return 2513 * 0 2514 * 2515 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2516 * Description 2517 * This helper gets a **struct bpf_sock** pointer such 2518 * that all the fields in this **bpf_sock** can be accessed. 2519 * Return 2520 * A **struct bpf_sock** pointer on success, or **NULL** in 2521 * case of failure. 2522 * 2523 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2524 * Description 2525 * This helper gets a **struct bpf_tcp_sock** pointer from a 2526 * **struct bpf_sock** pointer. 2527 * Return 2528 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2529 * case of failure. 2530 * 2531 * int bpf_skb_ecn_set_ce(struct sk_buff *skb) 2532 * Description 2533 * Set ECN (Explicit Congestion Notification) field of IP header 2534 * to **CE** (Congestion Encountered) if current value is **ECT** 2535 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2536 * and IPv4. 2537 * Return 2538 * 1 if the **CE** flag is set (either by the current helper call 2539 * or because it was already present), 0 if it is not set. 2540 * 2541 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2542 * Description 2543 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2544 * **bpf_sk_release**\ () is unnecessary and not allowed. 2545 * Return 2546 * A **struct bpf_sock** pointer on success, or **NULL** in 2547 * case of failure. 2548 * 2549 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2550 * Description 2551 * Look for TCP socket matching *tuple*, optionally in a child 2552 * network namespace *netns*. The return value must be checked, 2553 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2554 * 2555 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2556 * that it also returns timewait or request sockets. Use 2557 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2558 * full structure. 2559 * 2560 * This helper is available only if the kernel was compiled with 2561 * **CONFIG_NET** configuration option. 2562 * Return 2563 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2564 * For sockets with reuseport option, the **struct bpf_sock** 2565 * result is from *reuse*\ **->socks**\ [] using the hash of the 2566 * tuple. 2567 * 2568 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2569 * Description 2570 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2571 * the listening socket in *sk*. 2572 * 2573 * *iph* points to the start of the IPv4 or IPv6 header, while 2574 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2575 * **sizeof**\ (**struct ip6hdr**). 2576 * 2577 * *th* points to the start of the TCP header, while *th_len* 2578 * contains **sizeof**\ (**struct tcphdr**). 2579 * 2580 * Return 2581 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2582 * error otherwise. 2583 * 2584 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2585 * Description 2586 * Get name of sysctl in /proc/sys/ and copy it into provided by 2587 * program buffer *buf* of size *buf_len*. 2588 * 2589 * The buffer is always NUL terminated, unless it's zero-sized. 2590 * 2591 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2592 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2593 * only (e.g. "tcp_mem"). 2594 * Return 2595 * Number of character copied (not including the trailing NUL). 2596 * 2597 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2598 * truncated name in this case). 2599 * 2600 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2601 * Description 2602 * Get current value of sysctl as it is presented in /proc/sys 2603 * (incl. newline, etc), and copy it as a string into provided 2604 * by program buffer *buf* of size *buf_len*. 2605 * 2606 * The whole value is copied, no matter what file position user 2607 * space issued e.g. sys_read at. 2608 * 2609 * The buffer is always NUL terminated, unless it's zero-sized. 2610 * Return 2611 * Number of character copied (not including the trailing NUL). 2612 * 2613 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2614 * truncated name in this case). 2615 * 2616 * **-EINVAL** if current value was unavailable, e.g. because 2617 * sysctl is uninitialized and read returns -EIO for it. 2618 * 2619 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2620 * Description 2621 * Get new value being written by user space to sysctl (before 2622 * the actual write happens) and copy it as a string into 2623 * provided by program buffer *buf* of size *buf_len*. 2624 * 2625 * User space may write new value at file position > 0. 2626 * 2627 * The buffer is always NUL terminated, unless it's zero-sized. 2628 * Return 2629 * Number of character copied (not including the trailing NUL). 2630 * 2631 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2632 * truncated name in this case). 2633 * 2634 * **-EINVAL** if sysctl is being read. 2635 * 2636 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2637 * Description 2638 * Override new value being written by user space to sysctl with 2639 * value provided by program in buffer *buf* of size *buf_len*. 2640 * 2641 * *buf* should contain a string in same form as provided by user 2642 * space on sysctl write. 2643 * 2644 * User space may write new value at file position > 0. To override 2645 * the whole sysctl value file position should be set to zero. 2646 * Return 2647 * 0 on success. 2648 * 2649 * **-E2BIG** if the *buf_len* is too big. 2650 * 2651 * **-EINVAL** if sysctl is being read. 2652 * 2653 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2654 * Description 2655 * Convert the initial part of the string from buffer *buf* of 2656 * size *buf_len* to a long integer according to the given base 2657 * and save the result in *res*. 2658 * 2659 * The string may begin with an arbitrary amount of white space 2660 * (as determined by **isspace**\ (3)) followed by a single 2661 * optional '**-**' sign. 2662 * 2663 * Five least significant bits of *flags* encode base, other bits 2664 * are currently unused. 2665 * 2666 * Base must be either 8, 10, 16 or 0 to detect it automatically 2667 * similar to user space **strtol**\ (3). 2668 * Return 2669 * Number of characters consumed on success. Must be positive but 2670 * no more than *buf_len*. 2671 * 2672 * **-EINVAL** if no valid digits were found or unsupported base 2673 * was provided. 2674 * 2675 * **-ERANGE** if resulting value was out of range. 2676 * 2677 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2678 * Description 2679 * Convert the initial part of the string from buffer *buf* of 2680 * size *buf_len* to an unsigned long integer according to the 2681 * given base and save the result in *res*. 2682 * 2683 * The string may begin with an arbitrary amount of white space 2684 * (as determined by **isspace**\ (3)). 2685 * 2686 * Five least significant bits of *flags* encode base, other bits 2687 * are currently unused. 2688 * 2689 * Base must be either 8, 10, 16 or 0 to detect it automatically 2690 * similar to user space **strtoul**\ (3). 2691 * Return 2692 * Number of characters consumed on success. Must be positive but 2693 * no more than *buf_len*. 2694 * 2695 * **-EINVAL** if no valid digits were found or unsupported base 2696 * was provided. 2697 * 2698 * **-ERANGE** if resulting value was out of range. 2699 * 2700 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2701 * Description 2702 * Get a bpf-local-storage from a *sk*. 2703 * 2704 * Logically, it could be thought of getting the value from 2705 * a *map* with *sk* as the **key**. From this 2706 * perspective, the usage is not much different from 2707 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2708 * helper enforces the key must be a full socket and the map must 2709 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2710 * 2711 * Underneath, the value is stored locally at *sk* instead of 2712 * the *map*. The *map* is used as the bpf-local-storage 2713 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2714 * searched against all bpf-local-storages residing at *sk*. 2715 * 2716 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2717 * used such that a new bpf-local-storage will be 2718 * created if one does not exist. *value* can be used 2719 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2720 * the initial value of a bpf-local-storage. If *value* is 2721 * **NULL**, the new bpf-local-storage will be zero initialized. 2722 * Return 2723 * A bpf-local-storage pointer is returned on success. 2724 * 2725 * **NULL** if not found or there was an error in adding 2726 * a new bpf-local-storage. 2727 * 2728 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2729 * Description 2730 * Delete a bpf-local-storage from a *sk*. 2731 * Return 2732 * 0 on success. 2733 * 2734 * **-ENOENT** if the bpf-local-storage cannot be found. 2735 * 2736 * int bpf_send_signal(u32 sig) 2737 * Description 2738 * Send signal *sig* to the process of the current task. 2739 * The signal may be delivered to any of this process's threads. 2740 * Return 2741 * 0 on success or successfully queued. 2742 * 2743 * **-EBUSY** if work queue under nmi is full. 2744 * 2745 * **-EINVAL** if *sig* is invalid. 2746 * 2747 * **-EPERM** if no permission to send the *sig*. 2748 * 2749 * **-EAGAIN** if bpf program can try again. 2750 * 2751 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2752 * Description 2753 * Try to issue a SYN cookie for the packet with corresponding 2754 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2755 * 2756 * *iph* points to the start of the IPv4 or IPv6 header, while 2757 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2758 * **sizeof**\ (**struct ip6hdr**). 2759 * 2760 * *th* points to the start of the TCP header, while *th_len* 2761 * contains the length of the TCP header. 2762 * 2763 * Return 2764 * On success, lower 32 bits hold the generated SYN cookie in 2765 * followed by 16 bits which hold the MSS value for that cookie, 2766 * and the top 16 bits are unused. 2767 * 2768 * On failure, the returned value is one of the following: 2769 * 2770 * **-EINVAL** SYN cookie cannot be issued due to error 2771 * 2772 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2773 * 2774 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2775 * 2776 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2777 * 2778 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2779 * Description 2780 * Write raw *data* blob into a special BPF perf event held by 2781 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2782 * event must have the following attributes: **PERF_SAMPLE_RAW** 2783 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2784 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2785 * 2786 * The *flags* are used to indicate the index in *map* for which 2787 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2788 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2789 * to indicate that the index of the current CPU core should be 2790 * used. 2791 * 2792 * The value to write, of *size*, is passed through eBPF stack and 2793 * pointed by *data*. 2794 * 2795 * *ctx* is a pointer to in-kernel struct sk_buff. 2796 * 2797 * This helper is similar to **bpf_perf_event_output**\ () but 2798 * restricted to raw_tracepoint bpf programs. 2799 * Return 2800 * 0 on success, or a negative error in case of failure. 2801 * 2802 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2803 * Description 2804 * Safely attempt to read *size* bytes from user space address 2805 * *unsafe_ptr* and store the data in *dst*. 2806 * Return 2807 * 0 on success, or a negative error in case of failure. 2808 * 2809 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2810 * Description 2811 * Safely attempt to read *size* bytes from kernel space address 2812 * *unsafe_ptr* and store the data in *dst*. 2813 * Return 2814 * 0 on success, or a negative error in case of failure. 2815 * 2816 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2817 * Description 2818 * Copy a NUL terminated string from an unsafe user address 2819 * *unsafe_ptr* to *dst*. The *size* should include the 2820 * terminating NUL byte. In case the string length is smaller than 2821 * *size*, the target is not padded with further NUL bytes. If the 2822 * string length is larger than *size*, just *size*-1 bytes are 2823 * copied and the last byte is set to NUL. 2824 * 2825 * On success, the length of the copied string is returned. This 2826 * makes this helper useful in tracing programs for reading 2827 * strings, and more importantly to get its length at runtime. See 2828 * the following snippet: 2829 * 2830 * :: 2831 * 2832 * SEC("kprobe/sys_open") 2833 * void bpf_sys_open(struct pt_regs *ctx) 2834 * { 2835 * char buf[PATHLEN]; // PATHLEN is defined to 256 2836 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2837 * ctx->di); 2838 * 2839 * // Consume buf, for example push it to 2840 * // userspace via bpf_perf_event_output(); we 2841 * // can use res (the string length) as event 2842 * // size, after checking its boundaries. 2843 * } 2844 * 2845 * In comparison, using **bpf_probe_read_user()** helper here 2846 * instead to read the string would require to estimate the length 2847 * at compile time, and would often result in copying more memory 2848 * than necessary. 2849 * 2850 * Another useful use case is when parsing individual process 2851 * arguments or individual environment variables navigating 2852 * *current*\ **->mm->arg_start** and *current*\ 2853 * **->mm->env_start**: using this helper and the return value, 2854 * one can quickly iterate at the right offset of the memory area. 2855 * Return 2856 * On success, the strictly positive length of the string, 2857 * including the trailing NUL character. On error, a negative 2858 * value. 2859 * 2860 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 2861 * Description 2862 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2863 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply. 2864 * Return 2865 * On success, the strictly positive length of the string, including 2866 * the trailing NUL character. On error, a negative value. 2867 * 2868 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 2869 * Description 2870 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock. 2871 * *rcv_nxt* is the ack_seq to be sent out. 2872 * Return 2873 * 0 on success, or a negative error in case of failure. 2874 * 2875 * int bpf_send_signal_thread(u32 sig) 2876 * Description 2877 * Send signal *sig* to the thread corresponding to the current task. 2878 * Return 2879 * 0 on success or successfully queued. 2880 * 2881 * **-EBUSY** if work queue under nmi is full. 2882 * 2883 * **-EINVAL** if *sig* is invalid. 2884 * 2885 * **-EPERM** if no permission to send the *sig*. 2886 * 2887 * **-EAGAIN** if bpf program can try again. 2888 */ 2889 #define __BPF_FUNC_MAPPER(FN) \ 2890 FN(unspec), \ 2891 FN(map_lookup_elem), \ 2892 FN(map_update_elem), \ 2893 FN(map_delete_elem), \ 2894 FN(probe_read), \ 2895 FN(ktime_get_ns), \ 2896 FN(trace_printk), \ 2897 FN(get_prandom_u32), \ 2898 FN(get_smp_processor_id), \ 2899 FN(skb_store_bytes), \ 2900 FN(l3_csum_replace), \ 2901 FN(l4_csum_replace), \ 2902 FN(tail_call), \ 2903 FN(clone_redirect), \ 2904 FN(get_current_pid_tgid), \ 2905 FN(get_current_uid_gid), \ 2906 FN(get_current_comm), \ 2907 FN(get_cgroup_classid), \ 2908 FN(skb_vlan_push), \ 2909 FN(skb_vlan_pop), \ 2910 FN(skb_get_tunnel_key), \ 2911 FN(skb_set_tunnel_key), \ 2912 FN(perf_event_read), \ 2913 FN(redirect), \ 2914 FN(get_route_realm), \ 2915 FN(perf_event_output), \ 2916 FN(skb_load_bytes), \ 2917 FN(get_stackid), \ 2918 FN(csum_diff), \ 2919 FN(skb_get_tunnel_opt), \ 2920 FN(skb_set_tunnel_opt), \ 2921 FN(skb_change_proto), \ 2922 FN(skb_change_type), \ 2923 FN(skb_under_cgroup), \ 2924 FN(get_hash_recalc), \ 2925 FN(get_current_task), \ 2926 FN(probe_write_user), \ 2927 FN(current_task_under_cgroup), \ 2928 FN(skb_change_tail), \ 2929 FN(skb_pull_data), \ 2930 FN(csum_update), \ 2931 FN(set_hash_invalid), \ 2932 FN(get_numa_node_id), \ 2933 FN(skb_change_head), \ 2934 FN(xdp_adjust_head), \ 2935 FN(probe_read_str), \ 2936 FN(get_socket_cookie), \ 2937 FN(get_socket_uid), \ 2938 FN(set_hash), \ 2939 FN(setsockopt), \ 2940 FN(skb_adjust_room), \ 2941 FN(redirect_map), \ 2942 FN(sk_redirect_map), \ 2943 FN(sock_map_update), \ 2944 FN(xdp_adjust_meta), \ 2945 FN(perf_event_read_value), \ 2946 FN(perf_prog_read_value), \ 2947 FN(getsockopt), \ 2948 FN(override_return), \ 2949 FN(sock_ops_cb_flags_set), \ 2950 FN(msg_redirect_map), \ 2951 FN(msg_apply_bytes), \ 2952 FN(msg_cork_bytes), \ 2953 FN(msg_pull_data), \ 2954 FN(bind), \ 2955 FN(xdp_adjust_tail), \ 2956 FN(skb_get_xfrm_state), \ 2957 FN(get_stack), \ 2958 FN(skb_load_bytes_relative), \ 2959 FN(fib_lookup), \ 2960 FN(sock_hash_update), \ 2961 FN(msg_redirect_hash), \ 2962 FN(sk_redirect_hash), \ 2963 FN(lwt_push_encap), \ 2964 FN(lwt_seg6_store_bytes), \ 2965 FN(lwt_seg6_adjust_srh), \ 2966 FN(lwt_seg6_action), \ 2967 FN(rc_repeat), \ 2968 FN(rc_keydown), \ 2969 FN(skb_cgroup_id), \ 2970 FN(get_current_cgroup_id), \ 2971 FN(get_local_storage), \ 2972 FN(sk_select_reuseport), \ 2973 FN(skb_ancestor_cgroup_id), \ 2974 FN(sk_lookup_tcp), \ 2975 FN(sk_lookup_udp), \ 2976 FN(sk_release), \ 2977 FN(map_push_elem), \ 2978 FN(map_pop_elem), \ 2979 FN(map_peek_elem), \ 2980 FN(msg_push_data), \ 2981 FN(msg_pop_data), \ 2982 FN(rc_pointer_rel), \ 2983 FN(spin_lock), \ 2984 FN(spin_unlock), \ 2985 FN(sk_fullsock), \ 2986 FN(tcp_sock), \ 2987 FN(skb_ecn_set_ce), \ 2988 FN(get_listener_sock), \ 2989 FN(skc_lookup_tcp), \ 2990 FN(tcp_check_syncookie), \ 2991 FN(sysctl_get_name), \ 2992 FN(sysctl_get_current_value), \ 2993 FN(sysctl_get_new_value), \ 2994 FN(sysctl_set_new_value), \ 2995 FN(strtol), \ 2996 FN(strtoul), \ 2997 FN(sk_storage_get), \ 2998 FN(sk_storage_delete), \ 2999 FN(send_signal), \ 3000 FN(tcp_gen_syncookie), \ 3001 FN(skb_output), \ 3002 FN(probe_read_user), \ 3003 FN(probe_read_kernel), \ 3004 FN(probe_read_user_str), \ 3005 FN(probe_read_kernel_str), \ 3006 FN(tcp_send_ack), \ 3007 FN(send_signal_thread), 3008 3009 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3010 * function eBPF program intends to call 3011 */ 3012 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3013 enum bpf_func_id { 3014 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3015 __BPF_FUNC_MAX_ID, 3016 }; 3017 #undef __BPF_ENUM_FN 3018 3019 /* All flags used by eBPF helper functions, placed here. */ 3020 3021 /* BPF_FUNC_skb_store_bytes flags. */ 3022 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 3023 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 3024 3025 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3026 * First 4 bits are for passing the header field size. 3027 */ 3028 #define BPF_F_HDR_FIELD_MASK 0xfULL 3029 3030 /* BPF_FUNC_l4_csum_replace flags. */ 3031 #define BPF_F_PSEUDO_HDR (1ULL << 4) 3032 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 3033 #define BPF_F_MARK_ENFORCE (1ULL << 6) 3034 3035 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3036 #define BPF_F_INGRESS (1ULL << 0) 3037 3038 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3039 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 3040 3041 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3042 #define BPF_F_SKIP_FIELD_MASK 0xffULL 3043 #define BPF_F_USER_STACK (1ULL << 8) 3044 /* flags used by BPF_FUNC_get_stackid only. */ 3045 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 3046 #define BPF_F_REUSE_STACKID (1ULL << 10) 3047 /* flags used by BPF_FUNC_get_stack only. */ 3048 #define BPF_F_USER_BUILD_ID (1ULL << 11) 3049 3050 /* BPF_FUNC_skb_set_tunnel_key flags. */ 3051 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 3052 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 3053 #define BPF_F_SEQ_NUMBER (1ULL << 3) 3054 3055 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3056 * BPF_FUNC_perf_event_read_value flags. 3057 */ 3058 #define BPF_F_INDEX_MASK 0xffffffffULL 3059 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 3060 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 3061 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 3062 3063 /* Current network namespace */ 3064 #define BPF_F_CURRENT_NETNS (-1L) 3065 3066 /* BPF_FUNC_skb_adjust_room flags. */ 3067 #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 3068 3069 #define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff 3070 #define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 3071 3072 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 3073 #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 3074 #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 3075 #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 3076 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3077 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3078 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3079 3080 /* BPF_FUNC_sysctl_get_name flags. */ 3081 #define BPF_F_SYSCTL_BASE_NAME (1ULL << 0) 3082 3083 /* BPF_FUNC_sk_storage_get flags */ 3084 #define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0) 3085 3086 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 3087 enum bpf_adj_room_mode { 3088 BPF_ADJ_ROOM_NET, 3089 BPF_ADJ_ROOM_MAC, 3090 }; 3091 3092 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3093 enum bpf_hdr_start_off { 3094 BPF_HDR_START_MAC, 3095 BPF_HDR_START_NET, 3096 }; 3097 3098 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3099 enum bpf_lwt_encap_mode { 3100 BPF_LWT_ENCAP_SEG6, 3101 BPF_LWT_ENCAP_SEG6_INLINE, 3102 BPF_LWT_ENCAP_IP, 3103 }; 3104 3105 #define __bpf_md_ptr(type, name) \ 3106 union { \ 3107 type name; \ 3108 __u64 :64; \ 3109 } __attribute__((aligned(8))) 3110 3111 /* user accessible mirror of in-kernel sk_buff. 3112 * new fields can only be added to the end of this structure 3113 */ 3114 struct __sk_buff { 3115 __u32 len; 3116 __u32 pkt_type; 3117 __u32 mark; 3118 __u32 queue_mapping; 3119 __u32 protocol; 3120 __u32 vlan_present; 3121 __u32 vlan_tci; 3122 __u32 vlan_proto; 3123 __u32 priority; 3124 __u32 ingress_ifindex; 3125 __u32 ifindex; 3126 __u32 tc_index; 3127 __u32 cb[5]; 3128 __u32 hash; 3129 __u32 tc_classid; 3130 __u32 data; 3131 __u32 data_end; 3132 __u32 napi_id; 3133 3134 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3135 __u32 family; 3136 __u32 remote_ip4; /* Stored in network byte order */ 3137 __u32 local_ip4; /* Stored in network byte order */ 3138 __u32 remote_ip6[4]; /* Stored in network byte order */ 3139 __u32 local_ip6[4]; /* Stored in network byte order */ 3140 __u32 remote_port; /* Stored in network byte order */ 3141 __u32 local_port; /* stored in host byte order */ 3142 /* ... here. */ 3143 3144 __u32 data_meta; 3145 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3146 __u64 tstamp; 3147 __u32 wire_len; 3148 __u32 gso_segs; 3149 __bpf_md_ptr(struct bpf_sock *, sk); 3150 }; 3151 3152 struct bpf_tunnel_key { 3153 __u32 tunnel_id; 3154 union { 3155 __u32 remote_ipv4; 3156 __u32 remote_ipv6[4]; 3157 }; 3158 __u8 tunnel_tos; 3159 __u8 tunnel_ttl; 3160 __u16 tunnel_ext; /* Padding, future use. */ 3161 __u32 tunnel_label; 3162 }; 3163 3164 /* user accessible mirror of in-kernel xfrm_state. 3165 * new fields can only be added to the end of this structure 3166 */ 3167 struct bpf_xfrm_state { 3168 __u32 reqid; 3169 __u32 spi; /* Stored in network byte order */ 3170 __u16 family; 3171 __u16 ext; /* Padding, future use. */ 3172 union { 3173 __u32 remote_ipv4; /* Stored in network byte order */ 3174 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3175 }; 3176 }; 3177 3178 /* Generic BPF return codes which all BPF program types may support. 3179 * The values are binary compatible with their TC_ACT_* counter-part to 3180 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3181 * programs. 3182 * 3183 * XDP is handled seprately, see XDP_*. 3184 */ 3185 enum bpf_ret_code { 3186 BPF_OK = 0, 3187 /* 1 reserved */ 3188 BPF_DROP = 2, 3189 /* 3-6 reserved */ 3190 BPF_REDIRECT = 7, 3191 /* >127 are reserved for prog type specific return codes. 3192 * 3193 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3194 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3195 * changed and should be routed based on its new L3 header. 3196 * (This is an L3 redirect, as opposed to L2 redirect 3197 * represented by BPF_REDIRECT above). 3198 */ 3199 BPF_LWT_REROUTE = 128, 3200 }; 3201 3202 struct bpf_sock { 3203 __u32 bound_dev_if; 3204 __u32 family; 3205 __u32 type; 3206 __u32 protocol; 3207 __u32 mark; 3208 __u32 priority; 3209 /* IP address also allows 1 and 2 bytes access */ 3210 __u32 src_ip4; 3211 __u32 src_ip6[4]; 3212 __u32 src_port; /* host byte order */ 3213 __u32 dst_port; /* network byte order */ 3214 __u32 dst_ip4; 3215 __u32 dst_ip6[4]; 3216 __u32 state; 3217 }; 3218 3219 struct bpf_tcp_sock { 3220 __u32 snd_cwnd; /* Sending congestion window */ 3221 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3222 __u32 rtt_min; 3223 __u32 snd_ssthresh; /* Slow start size threshold */ 3224 __u32 rcv_nxt; /* What we want to receive next */ 3225 __u32 snd_nxt; /* Next sequence we send */ 3226 __u32 snd_una; /* First byte we want an ack for */ 3227 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3228 __u32 ecn_flags; /* ECN status bits. */ 3229 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3230 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3231 __u32 packets_out; /* Packets which are "in flight" */ 3232 __u32 retrans_out; /* Retransmitted packets out */ 3233 __u32 total_retrans; /* Total retransmits for entire connection */ 3234 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3235 * total number of segments in. 3236 */ 3237 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3238 * total number of data segments in. 3239 */ 3240 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3241 * The total number of segments sent. 3242 */ 3243 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3244 * total number of data segments sent. 3245 */ 3246 __u32 lost_out; /* Lost packets */ 3247 __u32 sacked_out; /* SACK'd packets */ 3248 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3249 * sum(delta(rcv_nxt)), or how many bytes 3250 * were acked. 3251 */ 3252 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3253 * sum(delta(snd_una)), or how many bytes 3254 * were acked. 3255 */ 3256 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3257 * total number of DSACK blocks received 3258 */ 3259 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3260 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3261 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3262 }; 3263 3264 struct bpf_sock_tuple { 3265 union { 3266 struct { 3267 __be32 saddr; 3268 __be32 daddr; 3269 __be16 sport; 3270 __be16 dport; 3271 } ipv4; 3272 struct { 3273 __be32 saddr[4]; 3274 __be32 daddr[4]; 3275 __be16 sport; 3276 __be16 dport; 3277 } ipv6; 3278 }; 3279 }; 3280 3281 struct bpf_xdp_sock { 3282 __u32 queue_id; 3283 }; 3284 3285 #define XDP_PACKET_HEADROOM 256 3286 3287 /* User return codes for XDP prog type. 3288 * A valid XDP program must return one of these defined values. All other 3289 * return codes are reserved for future use. Unknown return codes will 3290 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3291 */ 3292 enum xdp_action { 3293 XDP_ABORTED = 0, 3294 XDP_DROP, 3295 XDP_PASS, 3296 XDP_TX, 3297 XDP_REDIRECT, 3298 }; 3299 3300 /* user accessible metadata for XDP packet hook 3301 * new fields must be added to the end of this structure 3302 */ 3303 struct xdp_md { 3304 __u32 data; 3305 __u32 data_end; 3306 __u32 data_meta; 3307 /* Below access go through struct xdp_rxq_info */ 3308 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3309 __u32 rx_queue_index; /* rxq->queue_index */ 3310 }; 3311 3312 enum sk_action { 3313 SK_DROP = 0, 3314 SK_PASS, 3315 }; 3316 3317 /* user accessible metadata for SK_MSG packet hook, new fields must 3318 * be added to the end of this structure 3319 */ 3320 struct sk_msg_md { 3321 __bpf_md_ptr(void *, data); 3322 __bpf_md_ptr(void *, data_end); 3323 3324 __u32 family; 3325 __u32 remote_ip4; /* Stored in network byte order */ 3326 __u32 local_ip4; /* Stored in network byte order */ 3327 __u32 remote_ip6[4]; /* Stored in network byte order */ 3328 __u32 local_ip6[4]; /* Stored in network byte order */ 3329 __u32 remote_port; /* Stored in network byte order */ 3330 __u32 local_port; /* stored in host byte order */ 3331 __u32 size; /* Total size of sk_msg */ 3332 }; 3333 3334 struct sk_reuseport_md { 3335 /* 3336 * Start of directly accessible data. It begins from 3337 * the tcp/udp header. 3338 */ 3339 __bpf_md_ptr(void *, data); 3340 /* End of directly accessible data */ 3341 __bpf_md_ptr(void *, data_end); 3342 /* 3343 * Total length of packet (starting from the tcp/udp header). 3344 * Note that the directly accessible bytes (data_end - data) 3345 * could be less than this "len". Those bytes could be 3346 * indirectly read by a helper "bpf_skb_load_bytes()". 3347 */ 3348 __u32 len; 3349 /* 3350 * Eth protocol in the mac header (network byte order). e.g. 3351 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3352 */ 3353 __u32 eth_protocol; 3354 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3355 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3356 __u32 hash; /* A hash of the packet 4 tuples */ 3357 }; 3358 3359 #define BPF_TAG_SIZE 8 3360 3361 struct bpf_prog_info { 3362 __u32 type; 3363 __u32 id; 3364 __u8 tag[BPF_TAG_SIZE]; 3365 __u32 jited_prog_len; 3366 __u32 xlated_prog_len; 3367 __aligned_u64 jited_prog_insns; 3368 __aligned_u64 xlated_prog_insns; 3369 __u64 load_time; /* ns since boottime */ 3370 __u32 created_by_uid; 3371 __u32 nr_map_ids; 3372 __aligned_u64 map_ids; 3373 char name[BPF_OBJ_NAME_LEN]; 3374 __u32 ifindex; 3375 __u32 gpl_compatible:1; 3376 __u32 :31; /* alignment pad */ 3377 __u64 netns_dev; 3378 __u64 netns_ino; 3379 __u32 nr_jited_ksyms; 3380 __u32 nr_jited_func_lens; 3381 __aligned_u64 jited_ksyms; 3382 __aligned_u64 jited_func_lens; 3383 __u32 btf_id; 3384 __u32 func_info_rec_size; 3385 __aligned_u64 func_info; 3386 __u32 nr_func_info; 3387 __u32 nr_line_info; 3388 __aligned_u64 line_info; 3389 __aligned_u64 jited_line_info; 3390 __u32 nr_jited_line_info; 3391 __u32 line_info_rec_size; 3392 __u32 jited_line_info_rec_size; 3393 __u32 nr_prog_tags; 3394 __aligned_u64 prog_tags; 3395 __u64 run_time_ns; 3396 __u64 run_cnt; 3397 } __attribute__((aligned(8))); 3398 3399 struct bpf_map_info { 3400 __u32 type; 3401 __u32 id; 3402 __u32 key_size; 3403 __u32 value_size; 3404 __u32 max_entries; 3405 __u32 map_flags; 3406 char name[BPF_OBJ_NAME_LEN]; 3407 __u32 ifindex; 3408 __u32 btf_vmlinux_value_type_id; 3409 __u64 netns_dev; 3410 __u64 netns_ino; 3411 __u32 btf_id; 3412 __u32 btf_key_type_id; 3413 __u32 btf_value_type_id; 3414 } __attribute__((aligned(8))); 3415 3416 struct bpf_btf_info { 3417 __aligned_u64 btf; 3418 __u32 btf_size; 3419 __u32 id; 3420 } __attribute__((aligned(8))); 3421 3422 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 3423 * by user and intended to be used by socket (e.g. to bind to, depends on 3424 * attach attach type). 3425 */ 3426 struct bpf_sock_addr { 3427 __u32 user_family; /* Allows 4-byte read, but no write. */ 3428 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3429 * Stored in network byte order. 3430 */ 3431 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3432 * Stored in network byte order. 3433 */ 3434 __u32 user_port; /* Allows 4-byte read and write. 3435 * Stored in network byte order 3436 */ 3437 __u32 family; /* Allows 4-byte read, but no write */ 3438 __u32 type; /* Allows 4-byte read, but no write */ 3439 __u32 protocol; /* Allows 4-byte read, but no write */ 3440 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 3441 * Stored in network byte order. 3442 */ 3443 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 3444 * Stored in network byte order. 3445 */ 3446 __bpf_md_ptr(struct bpf_sock *, sk); 3447 }; 3448 3449 /* User bpf_sock_ops struct to access socket values and specify request ops 3450 * and their replies. 3451 * Some of this fields are in network (bigendian) byte order and may need 3452 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 3453 * New fields can only be added at the end of this structure 3454 */ 3455 struct bpf_sock_ops { 3456 __u32 op; 3457 union { 3458 __u32 args[4]; /* Optionally passed to bpf program */ 3459 __u32 reply; /* Returned by bpf program */ 3460 __u32 replylong[4]; /* Optionally returned by bpf prog */ 3461 }; 3462 __u32 family; 3463 __u32 remote_ip4; /* Stored in network byte order */ 3464 __u32 local_ip4; /* Stored in network byte order */ 3465 __u32 remote_ip6[4]; /* Stored in network byte order */ 3466 __u32 local_ip6[4]; /* Stored in network byte order */ 3467 __u32 remote_port; /* Stored in network byte order */ 3468 __u32 local_port; /* stored in host byte order */ 3469 __u32 is_fullsock; /* Some TCP fields are only valid if 3470 * there is a full socket. If not, the 3471 * fields read as zero. 3472 */ 3473 __u32 snd_cwnd; 3474 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3475 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 3476 __u32 state; 3477 __u32 rtt_min; 3478 __u32 snd_ssthresh; 3479 __u32 rcv_nxt; 3480 __u32 snd_nxt; 3481 __u32 snd_una; 3482 __u32 mss_cache; 3483 __u32 ecn_flags; 3484 __u32 rate_delivered; 3485 __u32 rate_interval_us; 3486 __u32 packets_out; 3487 __u32 retrans_out; 3488 __u32 total_retrans; 3489 __u32 segs_in; 3490 __u32 data_segs_in; 3491 __u32 segs_out; 3492 __u32 data_segs_out; 3493 __u32 lost_out; 3494 __u32 sacked_out; 3495 __u32 sk_txhash; 3496 __u64 bytes_received; 3497 __u64 bytes_acked; 3498 __bpf_md_ptr(struct bpf_sock *, sk); 3499 }; 3500 3501 /* Definitions for bpf_sock_ops_cb_flags */ 3502 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3503 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3504 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3505 #define BPF_SOCK_OPS_RTT_CB_FLAG (1<<3) 3506 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0xF /* Mask of all currently 3507 * supported cb flags 3508 */ 3509 3510 /* List of known BPF sock_ops operators. 3511 * New entries can only be added at the end 3512 */ 3513 enum { 3514 BPF_SOCK_OPS_VOID, 3515 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 3516 * -1 if default value should be used 3517 */ 3518 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 3519 * window (in packets) or -1 if default 3520 * value should be used 3521 */ 3522 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 3523 * active connection is initialized 3524 */ 3525 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 3526 * active connection is 3527 * established 3528 */ 3529 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 3530 * passive connection is 3531 * established 3532 */ 3533 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 3534 * needs ECN 3535 */ 3536 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3537 * based on the path and may be 3538 * dependent on the congestion control 3539 * algorithm. In general it indicates 3540 * a congestion threshold. RTTs above 3541 * this indicate congestion 3542 */ 3543 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3544 * Arg1: value of icsk_retransmits 3545 * Arg2: value of icsk_rto 3546 * Arg3: whether RTO has expired 3547 */ 3548 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3549 * Arg1: sequence number of 1st byte 3550 * Arg2: # segments 3551 * Arg3: return value of 3552 * tcp_transmit_skb (0 => success) 3553 */ 3554 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3555 * Arg1: old_state 3556 * Arg2: new_state 3557 */ 3558 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3559 * socket transition to LISTEN state. 3560 */ 3561 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 3562 */ 3563 }; 3564 3565 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3566 * changes between the TCP and BPF versions. Ideally this should never happen. 3567 * If it does, we need to add code to convert them before calling 3568 * the BPF sock_ops function. 3569 */ 3570 enum { 3571 BPF_TCP_ESTABLISHED = 1, 3572 BPF_TCP_SYN_SENT, 3573 BPF_TCP_SYN_RECV, 3574 BPF_TCP_FIN_WAIT1, 3575 BPF_TCP_FIN_WAIT2, 3576 BPF_TCP_TIME_WAIT, 3577 BPF_TCP_CLOSE, 3578 BPF_TCP_CLOSE_WAIT, 3579 BPF_TCP_LAST_ACK, 3580 BPF_TCP_LISTEN, 3581 BPF_TCP_CLOSING, /* Now a valid state */ 3582 BPF_TCP_NEW_SYN_RECV, 3583 3584 BPF_TCP_MAX_STATES /* Leave at the end! */ 3585 }; 3586 3587 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 3588 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3589 3590 struct bpf_perf_event_value { 3591 __u64 counter; 3592 __u64 enabled; 3593 __u64 running; 3594 }; 3595 3596 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3597 #define BPF_DEVCG_ACC_READ (1ULL << 1) 3598 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3599 3600 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3601 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3602 3603 struct bpf_cgroup_dev_ctx { 3604 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 3605 __u32 access_type; 3606 __u32 major; 3607 __u32 minor; 3608 }; 3609 3610 struct bpf_raw_tracepoint_args { 3611 __u64 args[0]; 3612 }; 3613 3614 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 3615 * OUTPUT: Do lookup from egress perspective; default is ingress 3616 */ 3617 #define BPF_FIB_LOOKUP_DIRECT (1U << 0) 3618 #define BPF_FIB_LOOKUP_OUTPUT (1U << 1) 3619 3620 enum { 3621 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 3622 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 3623 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 3624 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 3625 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 3626 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 3627 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 3628 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 3629 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 3630 }; 3631 3632 struct bpf_fib_lookup { 3633 /* input: network family for lookup (AF_INET, AF_INET6) 3634 * output: network family of egress nexthop 3635 */ 3636 __u8 family; 3637 3638 /* set if lookup is to consider L4 data - e.g., FIB rules */ 3639 __u8 l4_protocol; 3640 __be16 sport; 3641 __be16 dport; 3642 3643 /* total length of packet from network header - used for MTU check */ 3644 __u16 tot_len; 3645 3646 /* input: L3 device index for lookup 3647 * output: device index from FIB lookup 3648 */ 3649 __u32 ifindex; 3650 3651 union { 3652 /* inputs to lookup */ 3653 __u8 tos; /* AF_INET */ 3654 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 3655 3656 /* output: metric of fib result (IPv4/IPv6 only) */ 3657 __u32 rt_metric; 3658 }; 3659 3660 union { 3661 __be32 ipv4_src; 3662 __u32 ipv6_src[4]; /* in6_addr; network order */ 3663 }; 3664 3665 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3666 * network header. output: bpf_fib_lookup sets to gateway address 3667 * if FIB lookup returns gateway route 3668 */ 3669 union { 3670 __be32 ipv4_dst; 3671 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3672 }; 3673 3674 /* output */ 3675 __be16 h_vlan_proto; 3676 __be16 h_vlan_TCI; 3677 __u8 smac[6]; /* ETH_ALEN */ 3678 __u8 dmac[6]; /* ETH_ALEN */ 3679 }; 3680 3681 enum bpf_task_fd_type { 3682 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 3683 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 3684 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 3685 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 3686 BPF_FD_TYPE_UPROBE, /* filename + offset */ 3687 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 3688 }; 3689 3690 #define BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0) 3691 #define BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1) 3692 #define BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2) 3693 3694 struct bpf_flow_keys { 3695 __u16 nhoff; 3696 __u16 thoff; 3697 __u16 addr_proto; /* ETH_P_* of valid addrs */ 3698 __u8 is_frag; 3699 __u8 is_first_frag; 3700 __u8 is_encap; 3701 __u8 ip_proto; 3702 __be16 n_proto; 3703 __be16 sport; 3704 __be16 dport; 3705 union { 3706 struct { 3707 __be32 ipv4_src; 3708 __be32 ipv4_dst; 3709 }; 3710 struct { 3711 __u32 ipv6_src[4]; /* in6_addr; network order */ 3712 __u32 ipv6_dst[4]; /* in6_addr; network order */ 3713 }; 3714 }; 3715 __u32 flags; 3716 __be32 flow_label; 3717 }; 3718 3719 struct bpf_func_info { 3720 __u32 insn_off; 3721 __u32 type_id; 3722 }; 3723 3724 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3725 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3726 3727 struct bpf_line_info { 3728 __u32 insn_off; 3729 __u32 file_name_off; 3730 __u32 line_off; 3731 __u32 line_col; 3732 }; 3733 3734 struct bpf_spin_lock { 3735 __u32 val; 3736 }; 3737 3738 struct bpf_sysctl { 3739 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 3740 * Allows 1,2,4-byte read, but no write. 3741 */ 3742 __u32 file_pos; /* Sysctl file position to read from, write to. 3743 * Allows 1,2,4-byte read an 4-byte write. 3744 */ 3745 }; 3746 3747 struct bpf_sockopt { 3748 __bpf_md_ptr(struct bpf_sock *, sk); 3749 __bpf_md_ptr(void *, optval); 3750 __bpf_md_ptr(void *, optval_end); 3751 3752 __s32 level; 3753 __s32 optname; 3754 __s32 optlen; 3755 __s32 retval; 3756 }; 3757 3758 #endif /* _UAPI__LINUX_BPF_H__ */ 3759