1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_XADD 0xc0 /* exclusive add */ 23 24 /* alu/jmp fields */ 25 #define BPF_MOV 0xb0 /* mov reg to reg */ 26 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27 28 /* change endianness of a register */ 29 #define BPF_END 0xd0 /* flags for endianness conversion: */ 30 #define BPF_TO_LE 0x00 /* convert to little-endian */ 31 #define BPF_TO_BE 0x08 /* convert to big-endian */ 32 #define BPF_FROM_LE BPF_TO_LE 33 #define BPF_FROM_BE BPF_TO_BE 34 35 /* jmp encodings */ 36 #define BPF_JNE 0x50 /* jump != */ 37 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 38 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 41 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 42 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43 #define BPF_CALL 0x80 /* function call */ 44 #define BPF_EXIT 0x90 /* function return */ 45 46 /* Register numbers */ 47 enum { 48 BPF_REG_0 = 0, 49 BPF_REG_1, 50 BPF_REG_2, 51 BPF_REG_3, 52 BPF_REG_4, 53 BPF_REG_5, 54 BPF_REG_6, 55 BPF_REG_7, 56 BPF_REG_8, 57 BPF_REG_9, 58 BPF_REG_10, 59 __MAX_BPF_REG, 60 }; 61 62 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63 #define MAX_BPF_REG __MAX_BPF_REG 64 65 struct bpf_insn { 66 __u8 code; /* opcode */ 67 __u8 dst_reg:4; /* dest register */ 68 __u8 src_reg:4; /* source register */ 69 __s16 off; /* signed offset */ 70 __s32 imm; /* signed immediate constant */ 71 }; 72 73 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74 struct bpf_lpm_trie_key { 75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 76 __u8 data[0]; /* Arbitrary size */ 77 }; 78 79 struct bpf_cgroup_storage_key { 80 __u64 cgroup_inode_id; /* cgroup inode id */ 81 __u32 attach_type; /* program attach type */ 82 }; 83 84 /* BPF syscall commands, see bpf(2) man-page for details. */ 85 enum bpf_cmd { 86 BPF_MAP_CREATE, 87 BPF_MAP_LOOKUP_ELEM, 88 BPF_MAP_UPDATE_ELEM, 89 BPF_MAP_DELETE_ELEM, 90 BPF_MAP_GET_NEXT_KEY, 91 BPF_PROG_LOAD, 92 BPF_OBJ_PIN, 93 BPF_OBJ_GET, 94 BPF_PROG_ATTACH, 95 BPF_PROG_DETACH, 96 BPF_PROG_TEST_RUN, 97 BPF_PROG_GET_NEXT_ID, 98 BPF_MAP_GET_NEXT_ID, 99 BPF_PROG_GET_FD_BY_ID, 100 BPF_MAP_GET_FD_BY_ID, 101 BPF_OBJ_GET_INFO_BY_FD, 102 BPF_PROG_QUERY, 103 BPF_RAW_TRACEPOINT_OPEN, 104 BPF_BTF_LOAD, 105 BPF_BTF_GET_FD_BY_ID, 106 BPF_TASK_FD_QUERY, 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 108 BPF_MAP_FREEZE, 109 BPF_BTF_GET_NEXT_ID, 110 BPF_MAP_LOOKUP_BATCH, 111 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 112 BPF_MAP_UPDATE_BATCH, 113 BPF_MAP_DELETE_BATCH, 114 BPF_LINK_CREATE, 115 BPF_LINK_UPDATE, 116 BPF_LINK_GET_FD_BY_ID, 117 BPF_LINK_GET_NEXT_ID, 118 BPF_ENABLE_STATS, 119 BPF_ITER_CREATE, 120 }; 121 122 enum bpf_map_type { 123 BPF_MAP_TYPE_UNSPEC, 124 BPF_MAP_TYPE_HASH, 125 BPF_MAP_TYPE_ARRAY, 126 BPF_MAP_TYPE_PROG_ARRAY, 127 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 128 BPF_MAP_TYPE_PERCPU_HASH, 129 BPF_MAP_TYPE_PERCPU_ARRAY, 130 BPF_MAP_TYPE_STACK_TRACE, 131 BPF_MAP_TYPE_CGROUP_ARRAY, 132 BPF_MAP_TYPE_LRU_HASH, 133 BPF_MAP_TYPE_LRU_PERCPU_HASH, 134 BPF_MAP_TYPE_LPM_TRIE, 135 BPF_MAP_TYPE_ARRAY_OF_MAPS, 136 BPF_MAP_TYPE_HASH_OF_MAPS, 137 BPF_MAP_TYPE_DEVMAP, 138 BPF_MAP_TYPE_SOCKMAP, 139 BPF_MAP_TYPE_CPUMAP, 140 BPF_MAP_TYPE_XSKMAP, 141 BPF_MAP_TYPE_SOCKHASH, 142 BPF_MAP_TYPE_CGROUP_STORAGE, 143 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 144 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 145 BPF_MAP_TYPE_QUEUE, 146 BPF_MAP_TYPE_STACK, 147 BPF_MAP_TYPE_SK_STORAGE, 148 BPF_MAP_TYPE_DEVMAP_HASH, 149 BPF_MAP_TYPE_STRUCT_OPS, 150 BPF_MAP_TYPE_RINGBUF, 151 }; 152 153 /* Note that tracing related programs such as 154 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 155 * are not subject to a stable API since kernel internal data 156 * structures can change from release to release and may 157 * therefore break existing tracing BPF programs. Tracing BPF 158 * programs correspond to /a/ specific kernel which is to be 159 * analyzed, and not /a/ specific kernel /and/ all future ones. 160 */ 161 enum bpf_prog_type { 162 BPF_PROG_TYPE_UNSPEC, 163 BPF_PROG_TYPE_SOCKET_FILTER, 164 BPF_PROG_TYPE_KPROBE, 165 BPF_PROG_TYPE_SCHED_CLS, 166 BPF_PROG_TYPE_SCHED_ACT, 167 BPF_PROG_TYPE_TRACEPOINT, 168 BPF_PROG_TYPE_XDP, 169 BPF_PROG_TYPE_PERF_EVENT, 170 BPF_PROG_TYPE_CGROUP_SKB, 171 BPF_PROG_TYPE_CGROUP_SOCK, 172 BPF_PROG_TYPE_LWT_IN, 173 BPF_PROG_TYPE_LWT_OUT, 174 BPF_PROG_TYPE_LWT_XMIT, 175 BPF_PROG_TYPE_SOCK_OPS, 176 BPF_PROG_TYPE_SK_SKB, 177 BPF_PROG_TYPE_CGROUP_DEVICE, 178 BPF_PROG_TYPE_SK_MSG, 179 BPF_PROG_TYPE_RAW_TRACEPOINT, 180 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 181 BPF_PROG_TYPE_LWT_SEG6LOCAL, 182 BPF_PROG_TYPE_LIRC_MODE2, 183 BPF_PROG_TYPE_SK_REUSEPORT, 184 BPF_PROG_TYPE_FLOW_DISSECTOR, 185 BPF_PROG_TYPE_CGROUP_SYSCTL, 186 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 187 BPF_PROG_TYPE_CGROUP_SOCKOPT, 188 BPF_PROG_TYPE_TRACING, 189 BPF_PROG_TYPE_STRUCT_OPS, 190 BPF_PROG_TYPE_EXT, 191 BPF_PROG_TYPE_LSM, 192 BPF_PROG_TYPE_SK_LOOKUP, 193 }; 194 195 enum bpf_attach_type { 196 BPF_CGROUP_INET_INGRESS, 197 BPF_CGROUP_INET_EGRESS, 198 BPF_CGROUP_INET_SOCK_CREATE, 199 BPF_CGROUP_SOCK_OPS, 200 BPF_SK_SKB_STREAM_PARSER, 201 BPF_SK_SKB_STREAM_VERDICT, 202 BPF_CGROUP_DEVICE, 203 BPF_SK_MSG_VERDICT, 204 BPF_CGROUP_INET4_BIND, 205 BPF_CGROUP_INET6_BIND, 206 BPF_CGROUP_INET4_CONNECT, 207 BPF_CGROUP_INET6_CONNECT, 208 BPF_CGROUP_INET4_POST_BIND, 209 BPF_CGROUP_INET6_POST_BIND, 210 BPF_CGROUP_UDP4_SENDMSG, 211 BPF_CGROUP_UDP6_SENDMSG, 212 BPF_LIRC_MODE2, 213 BPF_FLOW_DISSECTOR, 214 BPF_CGROUP_SYSCTL, 215 BPF_CGROUP_UDP4_RECVMSG, 216 BPF_CGROUP_UDP6_RECVMSG, 217 BPF_CGROUP_GETSOCKOPT, 218 BPF_CGROUP_SETSOCKOPT, 219 BPF_TRACE_RAW_TP, 220 BPF_TRACE_FENTRY, 221 BPF_TRACE_FEXIT, 222 BPF_MODIFY_RETURN, 223 BPF_LSM_MAC, 224 BPF_TRACE_ITER, 225 BPF_CGROUP_INET4_GETPEERNAME, 226 BPF_CGROUP_INET6_GETPEERNAME, 227 BPF_CGROUP_INET4_GETSOCKNAME, 228 BPF_CGROUP_INET6_GETSOCKNAME, 229 BPF_XDP_DEVMAP, 230 BPF_CGROUP_INET_SOCK_RELEASE, 231 BPF_XDP_CPUMAP, 232 BPF_SK_LOOKUP, 233 __MAX_BPF_ATTACH_TYPE 234 }; 235 236 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 237 238 enum bpf_link_type { 239 BPF_LINK_TYPE_UNSPEC = 0, 240 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 241 BPF_LINK_TYPE_TRACING = 2, 242 BPF_LINK_TYPE_CGROUP = 3, 243 BPF_LINK_TYPE_ITER = 4, 244 BPF_LINK_TYPE_NETNS = 5, 245 246 MAX_BPF_LINK_TYPE, 247 }; 248 249 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 250 * 251 * NONE(default): No further bpf programs allowed in the subtree. 252 * 253 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 254 * the program in this cgroup yields to sub-cgroup program. 255 * 256 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 257 * that cgroup program gets run in addition to the program in this cgroup. 258 * 259 * Only one program is allowed to be attached to a cgroup with 260 * NONE or BPF_F_ALLOW_OVERRIDE flag. 261 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 262 * release old program and attach the new one. Attach flags has to match. 263 * 264 * Multiple programs are allowed to be attached to a cgroup with 265 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 266 * (those that were attached first, run first) 267 * The programs of sub-cgroup are executed first, then programs of 268 * this cgroup and then programs of parent cgroup. 269 * When children program makes decision (like picking TCP CA or sock bind) 270 * parent program has a chance to override it. 271 * 272 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 273 * programs for a cgroup. Though it's possible to replace an old program at 274 * any position by also specifying BPF_F_REPLACE flag and position itself in 275 * replace_bpf_fd attribute. Old program at this position will be released. 276 * 277 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 278 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 279 * Ex1: 280 * cgrp1 (MULTI progs A, B) -> 281 * cgrp2 (OVERRIDE prog C) -> 282 * cgrp3 (MULTI prog D) -> 283 * cgrp4 (OVERRIDE prog E) -> 284 * cgrp5 (NONE prog F) 285 * the event in cgrp5 triggers execution of F,D,A,B in that order. 286 * if prog F is detached, the execution is E,D,A,B 287 * if prog F and D are detached, the execution is E,A,B 288 * if prog F, E and D are detached, the execution is C,A,B 289 * 290 * All eligible programs are executed regardless of return code from 291 * earlier programs. 292 */ 293 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 294 #define BPF_F_ALLOW_MULTI (1U << 1) 295 #define BPF_F_REPLACE (1U << 2) 296 297 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 298 * verifier will perform strict alignment checking as if the kernel 299 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 300 * and NET_IP_ALIGN defined to 2. 301 */ 302 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 303 304 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 305 * verifier will allow any alignment whatsoever. On platforms 306 * with strict alignment requirements for loads ands stores (such 307 * as sparc and mips) the verifier validates that all loads and 308 * stores provably follow this requirement. This flag turns that 309 * checking and enforcement off. 310 * 311 * It is mostly used for testing when we want to validate the 312 * context and memory access aspects of the verifier, but because 313 * of an unaligned access the alignment check would trigger before 314 * the one we are interested in. 315 */ 316 #define BPF_F_ANY_ALIGNMENT (1U << 1) 317 318 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 319 * Verifier does sub-register def/use analysis and identifies instructions whose 320 * def only matters for low 32-bit, high 32-bit is never referenced later 321 * through implicit zero extension. Therefore verifier notifies JIT back-ends 322 * that it is safe to ignore clearing high 32-bit for these instructions. This 323 * saves some back-ends a lot of code-gen. However such optimization is not 324 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 325 * hence hasn't used verifier's analysis result. But, we really want to have a 326 * way to be able to verify the correctness of the described optimization on 327 * x86_64 on which testsuites are frequently exercised. 328 * 329 * So, this flag is introduced. Once it is set, verifier will randomize high 330 * 32-bit for those instructions who has been identified as safe to ignore them. 331 * Then, if verifier is not doing correct analysis, such randomization will 332 * regress tests to expose bugs. 333 */ 334 #define BPF_F_TEST_RND_HI32 (1U << 2) 335 336 /* The verifier internal test flag. Behavior is undefined */ 337 #define BPF_F_TEST_STATE_FREQ (1U << 3) 338 339 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 340 * two extensions: 341 * 342 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 343 * insn[0].imm: map fd map fd 344 * insn[1].imm: 0 offset into value 345 * insn[0].off: 0 0 346 * insn[1].off: 0 0 347 * ldimm64 rewrite: address of map address of map[0]+offset 348 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 349 */ 350 #define BPF_PSEUDO_MAP_FD 1 351 #define BPF_PSEUDO_MAP_VALUE 2 352 353 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 354 * offset to another bpf function 355 */ 356 #define BPF_PSEUDO_CALL 1 357 358 /* flags for BPF_MAP_UPDATE_ELEM command */ 359 enum { 360 BPF_ANY = 0, /* create new element or update existing */ 361 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 362 BPF_EXIST = 2, /* update existing element */ 363 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 364 }; 365 366 /* flags for BPF_MAP_CREATE command */ 367 enum { 368 BPF_F_NO_PREALLOC = (1U << 0), 369 /* Instead of having one common LRU list in the 370 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 371 * which can scale and perform better. 372 * Note, the LRU nodes (including free nodes) cannot be moved 373 * across different LRU lists. 374 */ 375 BPF_F_NO_COMMON_LRU = (1U << 1), 376 /* Specify numa node during map creation */ 377 BPF_F_NUMA_NODE = (1U << 2), 378 379 /* Flags for accessing BPF object from syscall side. */ 380 BPF_F_RDONLY = (1U << 3), 381 BPF_F_WRONLY = (1U << 4), 382 383 /* Flag for stack_map, store build_id+offset instead of pointer */ 384 BPF_F_STACK_BUILD_ID = (1U << 5), 385 386 /* Zero-initialize hash function seed. This should only be used for testing. */ 387 BPF_F_ZERO_SEED = (1U << 6), 388 389 /* Flags for accessing BPF object from program side. */ 390 BPF_F_RDONLY_PROG = (1U << 7), 391 BPF_F_WRONLY_PROG = (1U << 8), 392 393 /* Clone map from listener for newly accepted socket */ 394 BPF_F_CLONE = (1U << 9), 395 396 /* Enable memory-mapping BPF map */ 397 BPF_F_MMAPABLE = (1U << 10), 398 }; 399 400 /* Flags for BPF_PROG_QUERY. */ 401 402 /* Query effective (directly attached + inherited from ancestor cgroups) 403 * programs that will be executed for events within a cgroup. 404 * attach_flags with this flag are returned only for directly attached programs. 405 */ 406 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 407 408 /* type for BPF_ENABLE_STATS */ 409 enum bpf_stats_type { 410 /* enabled run_time_ns and run_cnt */ 411 BPF_STATS_RUN_TIME = 0, 412 }; 413 414 enum bpf_stack_build_id_status { 415 /* user space need an empty entry to identify end of a trace */ 416 BPF_STACK_BUILD_ID_EMPTY = 0, 417 /* with valid build_id and offset */ 418 BPF_STACK_BUILD_ID_VALID = 1, 419 /* couldn't get build_id, fallback to ip */ 420 BPF_STACK_BUILD_ID_IP = 2, 421 }; 422 423 #define BPF_BUILD_ID_SIZE 20 424 struct bpf_stack_build_id { 425 __s32 status; 426 unsigned char build_id[BPF_BUILD_ID_SIZE]; 427 union { 428 __u64 offset; 429 __u64 ip; 430 }; 431 }; 432 433 #define BPF_OBJ_NAME_LEN 16U 434 435 union bpf_attr { 436 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 437 __u32 map_type; /* one of enum bpf_map_type */ 438 __u32 key_size; /* size of key in bytes */ 439 __u32 value_size; /* size of value in bytes */ 440 __u32 max_entries; /* max number of entries in a map */ 441 __u32 map_flags; /* BPF_MAP_CREATE related 442 * flags defined above. 443 */ 444 __u32 inner_map_fd; /* fd pointing to the inner map */ 445 __u32 numa_node; /* numa node (effective only if 446 * BPF_F_NUMA_NODE is set). 447 */ 448 char map_name[BPF_OBJ_NAME_LEN]; 449 __u32 map_ifindex; /* ifindex of netdev to create on */ 450 __u32 btf_fd; /* fd pointing to a BTF type data */ 451 __u32 btf_key_type_id; /* BTF type_id of the key */ 452 __u32 btf_value_type_id; /* BTF type_id of the value */ 453 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 454 * struct stored as the 455 * map value 456 */ 457 }; 458 459 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 460 __u32 map_fd; 461 __aligned_u64 key; 462 union { 463 __aligned_u64 value; 464 __aligned_u64 next_key; 465 }; 466 __u64 flags; 467 }; 468 469 struct { /* struct used by BPF_MAP_*_BATCH commands */ 470 __aligned_u64 in_batch; /* start batch, 471 * NULL to start from beginning 472 */ 473 __aligned_u64 out_batch; /* output: next start batch */ 474 __aligned_u64 keys; 475 __aligned_u64 values; 476 __u32 count; /* input/output: 477 * input: # of key/value 478 * elements 479 * output: # of filled elements 480 */ 481 __u32 map_fd; 482 __u64 elem_flags; 483 __u64 flags; 484 } batch; 485 486 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 487 __u32 prog_type; /* one of enum bpf_prog_type */ 488 __u32 insn_cnt; 489 __aligned_u64 insns; 490 __aligned_u64 license; 491 __u32 log_level; /* verbosity level of verifier */ 492 __u32 log_size; /* size of user buffer */ 493 __aligned_u64 log_buf; /* user supplied buffer */ 494 __u32 kern_version; /* not used */ 495 __u32 prog_flags; 496 char prog_name[BPF_OBJ_NAME_LEN]; 497 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 498 /* For some prog types expected attach type must be known at 499 * load time to verify attach type specific parts of prog 500 * (context accesses, allowed helpers, etc). 501 */ 502 __u32 expected_attach_type; 503 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 504 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 505 __aligned_u64 func_info; /* func info */ 506 __u32 func_info_cnt; /* number of bpf_func_info records */ 507 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 508 __aligned_u64 line_info; /* line info */ 509 __u32 line_info_cnt; /* number of bpf_line_info records */ 510 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 511 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 512 }; 513 514 struct { /* anonymous struct used by BPF_OBJ_* commands */ 515 __aligned_u64 pathname; 516 __u32 bpf_fd; 517 __u32 file_flags; 518 }; 519 520 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 521 __u32 target_fd; /* container object to attach to */ 522 __u32 attach_bpf_fd; /* eBPF program to attach */ 523 __u32 attach_type; 524 __u32 attach_flags; 525 __u32 replace_bpf_fd; /* previously attached eBPF 526 * program to replace if 527 * BPF_F_REPLACE is used 528 */ 529 }; 530 531 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 532 __u32 prog_fd; 533 __u32 retval; 534 __u32 data_size_in; /* input: len of data_in */ 535 __u32 data_size_out; /* input/output: len of data_out 536 * returns ENOSPC if data_out 537 * is too small. 538 */ 539 __aligned_u64 data_in; 540 __aligned_u64 data_out; 541 __u32 repeat; 542 __u32 duration; 543 __u32 ctx_size_in; /* input: len of ctx_in */ 544 __u32 ctx_size_out; /* input/output: len of ctx_out 545 * returns ENOSPC if ctx_out 546 * is too small. 547 */ 548 __aligned_u64 ctx_in; 549 __aligned_u64 ctx_out; 550 } test; 551 552 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 553 union { 554 __u32 start_id; 555 __u32 prog_id; 556 __u32 map_id; 557 __u32 btf_id; 558 __u32 link_id; 559 }; 560 __u32 next_id; 561 __u32 open_flags; 562 }; 563 564 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 565 __u32 bpf_fd; 566 __u32 info_len; 567 __aligned_u64 info; 568 } info; 569 570 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 571 __u32 target_fd; /* container object to query */ 572 __u32 attach_type; 573 __u32 query_flags; 574 __u32 attach_flags; 575 __aligned_u64 prog_ids; 576 __u32 prog_cnt; 577 } query; 578 579 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 580 __u64 name; 581 __u32 prog_fd; 582 } raw_tracepoint; 583 584 struct { /* anonymous struct for BPF_BTF_LOAD */ 585 __aligned_u64 btf; 586 __aligned_u64 btf_log_buf; 587 __u32 btf_size; 588 __u32 btf_log_size; 589 __u32 btf_log_level; 590 }; 591 592 struct { 593 __u32 pid; /* input: pid */ 594 __u32 fd; /* input: fd */ 595 __u32 flags; /* input: flags */ 596 __u32 buf_len; /* input/output: buf len */ 597 __aligned_u64 buf; /* input/output: 598 * tp_name for tracepoint 599 * symbol for kprobe 600 * filename for uprobe 601 */ 602 __u32 prog_id; /* output: prod_id */ 603 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 604 __u64 probe_offset; /* output: probe_offset */ 605 __u64 probe_addr; /* output: probe_addr */ 606 } task_fd_query; 607 608 struct { /* struct used by BPF_LINK_CREATE command */ 609 __u32 prog_fd; /* eBPF program to attach */ 610 __u32 target_fd; /* object to attach to */ 611 __u32 attach_type; /* attach type */ 612 __u32 flags; /* extra flags */ 613 } link_create; 614 615 struct { /* struct used by BPF_LINK_UPDATE command */ 616 __u32 link_fd; /* link fd */ 617 /* new program fd to update link with */ 618 __u32 new_prog_fd; 619 __u32 flags; /* extra flags */ 620 /* expected link's program fd; is specified only if 621 * BPF_F_REPLACE flag is set in flags */ 622 __u32 old_prog_fd; 623 } link_update; 624 625 struct { /* struct used by BPF_ENABLE_STATS command */ 626 __u32 type; 627 } enable_stats; 628 629 struct { /* struct used by BPF_ITER_CREATE command */ 630 __u32 link_fd; 631 __u32 flags; 632 } iter_create; 633 634 } __attribute__((aligned(8))); 635 636 /* The description below is an attempt at providing documentation to eBPF 637 * developers about the multiple available eBPF helper functions. It can be 638 * parsed and used to produce a manual page. The workflow is the following, 639 * and requires the rst2man utility: 640 * 641 * $ ./scripts/bpf_helpers_doc.py \ 642 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 643 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 644 * $ man /tmp/bpf-helpers.7 645 * 646 * Note that in order to produce this external documentation, some RST 647 * formatting is used in the descriptions to get "bold" and "italics" in 648 * manual pages. Also note that the few trailing white spaces are 649 * intentional, removing them would break paragraphs for rst2man. 650 * 651 * Start of BPF helper function descriptions: 652 * 653 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 654 * Description 655 * Perform a lookup in *map* for an entry associated to *key*. 656 * Return 657 * Map value associated to *key*, or **NULL** if no entry was 658 * found. 659 * 660 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 661 * Description 662 * Add or update the value of the entry associated to *key* in 663 * *map* with *value*. *flags* is one of: 664 * 665 * **BPF_NOEXIST** 666 * The entry for *key* must not exist in the map. 667 * **BPF_EXIST** 668 * The entry for *key* must already exist in the map. 669 * **BPF_ANY** 670 * No condition on the existence of the entry for *key*. 671 * 672 * Flag value **BPF_NOEXIST** cannot be used for maps of types 673 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 674 * elements always exist), the helper would return an error. 675 * Return 676 * 0 on success, or a negative error in case of failure. 677 * 678 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 679 * Description 680 * Delete entry with *key* from *map*. 681 * Return 682 * 0 on success, or a negative error in case of failure. 683 * 684 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 685 * Description 686 * For tracing programs, safely attempt to read *size* bytes from 687 * kernel space address *unsafe_ptr* and store the data in *dst*. 688 * 689 * Generally, use **bpf_probe_read_user**\ () or 690 * **bpf_probe_read_kernel**\ () instead. 691 * Return 692 * 0 on success, or a negative error in case of failure. 693 * 694 * u64 bpf_ktime_get_ns(void) 695 * Description 696 * Return the time elapsed since system boot, in nanoseconds. 697 * Does not include time the system was suspended. 698 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 699 * Return 700 * Current *ktime*. 701 * 702 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 703 * Description 704 * This helper is a "printk()-like" facility for debugging. It 705 * prints a message defined by format *fmt* (of size *fmt_size*) 706 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 707 * available. It can take up to three additional **u64** 708 * arguments (as an eBPF helpers, the total number of arguments is 709 * limited to five). 710 * 711 * Each time the helper is called, it appends a line to the trace. 712 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 713 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 714 * The format of the trace is customizable, and the exact output 715 * one will get depends on the options set in 716 * *\/sys/kernel/debug/tracing/trace_options* (see also the 717 * *README* file under the same directory). However, it usually 718 * defaults to something like: 719 * 720 * :: 721 * 722 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 723 * 724 * In the above: 725 * 726 * * ``telnet`` is the name of the current task. 727 * * ``470`` is the PID of the current task. 728 * * ``001`` is the CPU number on which the task is 729 * running. 730 * * In ``.N..``, each character refers to a set of 731 * options (whether irqs are enabled, scheduling 732 * options, whether hard/softirqs are running, level of 733 * preempt_disabled respectively). **N** means that 734 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 735 * are set. 736 * * ``419421.045894`` is a timestamp. 737 * * ``0x00000001`` is a fake value used by BPF for the 738 * instruction pointer register. 739 * * ``<formatted msg>`` is the message formatted with 740 * *fmt*. 741 * 742 * The conversion specifiers supported by *fmt* are similar, but 743 * more limited than for printk(). They are **%d**, **%i**, 744 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 745 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 746 * of field, padding with zeroes, etc.) is available, and the 747 * helper will return **-EINVAL** (but print nothing) if it 748 * encounters an unknown specifier. 749 * 750 * Also, note that **bpf_trace_printk**\ () is slow, and should 751 * only be used for debugging purposes. For this reason, a notice 752 * bloc (spanning several lines) is printed to kernel logs and 753 * states that the helper should not be used "for production use" 754 * the first time this helper is used (or more precisely, when 755 * **trace_printk**\ () buffers are allocated). For passing values 756 * to user space, perf events should be preferred. 757 * Return 758 * The number of bytes written to the buffer, or a negative error 759 * in case of failure. 760 * 761 * u32 bpf_get_prandom_u32(void) 762 * Description 763 * Get a pseudo-random number. 764 * 765 * From a security point of view, this helper uses its own 766 * pseudo-random internal state, and cannot be used to infer the 767 * seed of other random functions in the kernel. However, it is 768 * essential to note that the generator used by the helper is not 769 * cryptographically secure. 770 * Return 771 * A random 32-bit unsigned value. 772 * 773 * u32 bpf_get_smp_processor_id(void) 774 * Description 775 * Get the SMP (symmetric multiprocessing) processor id. Note that 776 * all programs run with preemption disabled, which means that the 777 * SMP processor id is stable during all the execution of the 778 * program. 779 * Return 780 * The SMP id of the processor running the program. 781 * 782 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 783 * Description 784 * Store *len* bytes from address *from* into the packet 785 * associated to *skb*, at *offset*. *flags* are a combination of 786 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 787 * checksum for the packet after storing the bytes) and 788 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 789 * **->swhash** and *skb*\ **->l4hash** to 0). 790 * 791 * A call to this helper is susceptible to change the underlying 792 * packet buffer. Therefore, at load time, all checks on pointers 793 * previously done by the verifier are invalidated and must be 794 * performed again, if the helper is used in combination with 795 * direct packet access. 796 * Return 797 * 0 on success, or a negative error in case of failure. 798 * 799 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 800 * Description 801 * Recompute the layer 3 (e.g. IP) checksum for the packet 802 * associated to *skb*. Computation is incremental, so the helper 803 * must know the former value of the header field that was 804 * modified (*from*), the new value of this field (*to*), and the 805 * number of bytes (2 or 4) for this field, stored in *size*. 806 * Alternatively, it is possible to store the difference between 807 * the previous and the new values of the header field in *to*, by 808 * setting *from* and *size* to 0. For both methods, *offset* 809 * indicates the location of the IP checksum within the packet. 810 * 811 * This helper works in combination with **bpf_csum_diff**\ (), 812 * which does not update the checksum in-place, but offers more 813 * flexibility and can handle sizes larger than 2 or 4 for the 814 * checksum to update. 815 * 816 * A call to this helper is susceptible to change the underlying 817 * packet buffer. Therefore, at load time, all checks on pointers 818 * previously done by the verifier are invalidated and must be 819 * performed again, if the helper is used in combination with 820 * direct packet access. 821 * Return 822 * 0 on success, or a negative error in case of failure. 823 * 824 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 825 * Description 826 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 827 * packet associated to *skb*. Computation is incremental, so the 828 * helper must know the former value of the header field that was 829 * modified (*from*), the new value of this field (*to*), and the 830 * number of bytes (2 or 4) for this field, stored on the lowest 831 * four bits of *flags*. Alternatively, it is possible to store 832 * the difference between the previous and the new values of the 833 * header field in *to*, by setting *from* and the four lowest 834 * bits of *flags* to 0. For both methods, *offset* indicates the 835 * location of the IP checksum within the packet. In addition to 836 * the size of the field, *flags* can be added (bitwise OR) actual 837 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 838 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 839 * for updates resulting in a null checksum the value is set to 840 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 841 * the checksum is to be computed against a pseudo-header. 842 * 843 * This helper works in combination with **bpf_csum_diff**\ (), 844 * which does not update the checksum in-place, but offers more 845 * flexibility and can handle sizes larger than 2 or 4 for the 846 * checksum to update. 847 * 848 * A call to this helper is susceptible to change the underlying 849 * packet buffer. Therefore, at load time, all checks on pointers 850 * previously done by the verifier are invalidated and must be 851 * performed again, if the helper is used in combination with 852 * direct packet access. 853 * Return 854 * 0 on success, or a negative error in case of failure. 855 * 856 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 857 * Description 858 * This special helper is used to trigger a "tail call", or in 859 * other words, to jump into another eBPF program. The same stack 860 * frame is used (but values on stack and in registers for the 861 * caller are not accessible to the callee). This mechanism allows 862 * for program chaining, either for raising the maximum number of 863 * available eBPF instructions, or to execute given programs in 864 * conditional blocks. For security reasons, there is an upper 865 * limit to the number of successive tail calls that can be 866 * performed. 867 * 868 * Upon call of this helper, the program attempts to jump into a 869 * program referenced at index *index* in *prog_array_map*, a 870 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 871 * *ctx*, a pointer to the context. 872 * 873 * If the call succeeds, the kernel immediately runs the first 874 * instruction of the new program. This is not a function call, 875 * and it never returns to the previous program. If the call 876 * fails, then the helper has no effect, and the caller continues 877 * to run its subsequent instructions. A call can fail if the 878 * destination program for the jump does not exist (i.e. *index* 879 * is superior to the number of entries in *prog_array_map*), or 880 * if the maximum number of tail calls has been reached for this 881 * chain of programs. This limit is defined in the kernel by the 882 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 883 * which is currently set to 32. 884 * Return 885 * 0 on success, or a negative error in case of failure. 886 * 887 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 888 * Description 889 * Clone and redirect the packet associated to *skb* to another 890 * net device of index *ifindex*. Both ingress and egress 891 * interfaces can be used for redirection. The **BPF_F_INGRESS** 892 * value in *flags* is used to make the distinction (ingress path 893 * is selected if the flag is present, egress path otherwise). 894 * This is the only flag supported for now. 895 * 896 * In comparison with **bpf_redirect**\ () helper, 897 * **bpf_clone_redirect**\ () has the associated cost of 898 * duplicating the packet buffer, but this can be executed out of 899 * the eBPF program. Conversely, **bpf_redirect**\ () is more 900 * efficient, but it is handled through an action code where the 901 * redirection happens only after the eBPF program has returned. 902 * 903 * A call to this helper is susceptible to change the underlying 904 * packet buffer. Therefore, at load time, all checks on pointers 905 * previously done by the verifier are invalidated and must be 906 * performed again, if the helper is used in combination with 907 * direct packet access. 908 * Return 909 * 0 on success, or a negative error in case of failure. 910 * 911 * u64 bpf_get_current_pid_tgid(void) 912 * Return 913 * A 64-bit integer containing the current tgid and pid, and 914 * created as such: 915 * *current_task*\ **->tgid << 32 \|** 916 * *current_task*\ **->pid**. 917 * 918 * u64 bpf_get_current_uid_gid(void) 919 * Return 920 * A 64-bit integer containing the current GID and UID, and 921 * created as such: *current_gid* **<< 32 \|** *current_uid*. 922 * 923 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 924 * Description 925 * Copy the **comm** attribute of the current task into *buf* of 926 * *size_of_buf*. The **comm** attribute contains the name of 927 * the executable (excluding the path) for the current task. The 928 * *size_of_buf* must be strictly positive. On success, the 929 * helper makes sure that the *buf* is NUL-terminated. On failure, 930 * it is filled with zeroes. 931 * Return 932 * 0 on success, or a negative error in case of failure. 933 * 934 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 935 * Description 936 * Retrieve the classid for the current task, i.e. for the net_cls 937 * cgroup to which *skb* belongs. 938 * 939 * This helper can be used on TC egress path, but not on ingress. 940 * 941 * The net_cls cgroup provides an interface to tag network packets 942 * based on a user-provided identifier for all traffic coming from 943 * the tasks belonging to the related cgroup. See also the related 944 * kernel documentation, available from the Linux sources in file 945 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 946 * 947 * The Linux kernel has two versions for cgroups: there are 948 * cgroups v1 and cgroups v2. Both are available to users, who can 949 * use a mixture of them, but note that the net_cls cgroup is for 950 * cgroup v1 only. This makes it incompatible with BPF programs 951 * run on cgroups, which is a cgroup-v2-only feature (a socket can 952 * only hold data for one version of cgroups at a time). 953 * 954 * This helper is only available is the kernel was compiled with 955 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 956 * "**y**" or to "**m**". 957 * Return 958 * The classid, or 0 for the default unconfigured classid. 959 * 960 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 961 * Description 962 * Push a *vlan_tci* (VLAN tag control information) of protocol 963 * *vlan_proto* to the packet associated to *skb*, then update 964 * the checksum. Note that if *vlan_proto* is different from 965 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 966 * be **ETH_P_8021Q**. 967 * 968 * A call to this helper is susceptible to change the underlying 969 * packet buffer. Therefore, at load time, all checks on pointers 970 * previously done by the verifier are invalidated and must be 971 * performed again, if the helper is used in combination with 972 * direct packet access. 973 * Return 974 * 0 on success, or a negative error in case of failure. 975 * 976 * long bpf_skb_vlan_pop(struct sk_buff *skb) 977 * Description 978 * Pop a VLAN header from the packet associated to *skb*. 979 * 980 * A call to this helper is susceptible to change the underlying 981 * packet buffer. Therefore, at load time, all checks on pointers 982 * previously done by the verifier are invalidated and must be 983 * performed again, if the helper is used in combination with 984 * direct packet access. 985 * Return 986 * 0 on success, or a negative error in case of failure. 987 * 988 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 989 * Description 990 * Get tunnel metadata. This helper takes a pointer *key* to an 991 * empty **struct bpf_tunnel_key** of **size**, that will be 992 * filled with tunnel metadata for the packet associated to *skb*. 993 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 994 * indicates that the tunnel is based on IPv6 protocol instead of 995 * IPv4. 996 * 997 * The **struct bpf_tunnel_key** is an object that generalizes the 998 * principal parameters used by various tunneling protocols into a 999 * single struct. This way, it can be used to easily make a 1000 * decision based on the contents of the encapsulation header, 1001 * "summarized" in this struct. In particular, it holds the IP 1002 * address of the remote end (IPv4 or IPv6, depending on the case) 1003 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1004 * this struct exposes the *key*\ **->tunnel_id**, which is 1005 * generally mapped to a VNI (Virtual Network Identifier), making 1006 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1007 * () helper. 1008 * 1009 * Let's imagine that the following code is part of a program 1010 * attached to the TC ingress interface, on one end of a GRE 1011 * tunnel, and is supposed to filter out all messages coming from 1012 * remote ends with IPv4 address other than 10.0.0.1: 1013 * 1014 * :: 1015 * 1016 * int ret; 1017 * struct bpf_tunnel_key key = {}; 1018 * 1019 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1020 * if (ret < 0) 1021 * return TC_ACT_SHOT; // drop packet 1022 * 1023 * if (key.remote_ipv4 != 0x0a000001) 1024 * return TC_ACT_SHOT; // drop packet 1025 * 1026 * return TC_ACT_OK; // accept packet 1027 * 1028 * This interface can also be used with all encapsulation devices 1029 * that can operate in "collect metadata" mode: instead of having 1030 * one network device per specific configuration, the "collect 1031 * metadata" mode only requires a single device where the 1032 * configuration can be extracted from this helper. 1033 * 1034 * This can be used together with various tunnels such as VXLan, 1035 * Geneve, GRE or IP in IP (IPIP). 1036 * Return 1037 * 0 on success, or a negative error in case of failure. 1038 * 1039 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1040 * Description 1041 * Populate tunnel metadata for packet associated to *skb.* The 1042 * tunnel metadata is set to the contents of *key*, of *size*. The 1043 * *flags* can be set to a combination of the following values: 1044 * 1045 * **BPF_F_TUNINFO_IPV6** 1046 * Indicate that the tunnel is based on IPv6 protocol 1047 * instead of IPv4. 1048 * **BPF_F_ZERO_CSUM_TX** 1049 * For IPv4 packets, add a flag to tunnel metadata 1050 * indicating that checksum computation should be skipped 1051 * and checksum set to zeroes. 1052 * **BPF_F_DONT_FRAGMENT** 1053 * Add a flag to tunnel metadata indicating that the 1054 * packet should not be fragmented. 1055 * **BPF_F_SEQ_NUMBER** 1056 * Add a flag to tunnel metadata indicating that a 1057 * sequence number should be added to tunnel header before 1058 * sending the packet. This flag was added for GRE 1059 * encapsulation, but might be used with other protocols 1060 * as well in the future. 1061 * 1062 * Here is a typical usage on the transmit path: 1063 * 1064 * :: 1065 * 1066 * struct bpf_tunnel_key key; 1067 * populate key ... 1068 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1069 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1070 * 1071 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1072 * helper for additional information. 1073 * Return 1074 * 0 on success, or a negative error in case of failure. 1075 * 1076 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1077 * Description 1078 * Read the value of a perf event counter. This helper relies on a 1079 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1080 * the perf event counter is selected when *map* is updated with 1081 * perf event file descriptors. The *map* is an array whose size 1082 * is the number of available CPUs, and each cell contains a value 1083 * relative to one CPU. The value to retrieve is indicated by 1084 * *flags*, that contains the index of the CPU to look up, masked 1085 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1086 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1087 * current CPU should be retrieved. 1088 * 1089 * Note that before Linux 4.13, only hardware perf event can be 1090 * retrieved. 1091 * 1092 * Also, be aware that the newer helper 1093 * **bpf_perf_event_read_value**\ () is recommended over 1094 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1095 * quirks where error and counter value are used as a return code 1096 * (which is wrong to do since ranges may overlap). This issue is 1097 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1098 * time provides more features over the **bpf_perf_event_read**\ 1099 * () interface. Please refer to the description of 1100 * **bpf_perf_event_read_value**\ () for details. 1101 * Return 1102 * The value of the perf event counter read from the map, or a 1103 * negative error code in case of failure. 1104 * 1105 * long bpf_redirect(u32 ifindex, u64 flags) 1106 * Description 1107 * Redirect the packet to another net device of index *ifindex*. 1108 * This helper is somewhat similar to **bpf_clone_redirect**\ 1109 * (), except that the packet is not cloned, which provides 1110 * increased performance. 1111 * 1112 * Except for XDP, both ingress and egress interfaces can be used 1113 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1114 * to make the distinction (ingress path is selected if the flag 1115 * is present, egress path otherwise). Currently, XDP only 1116 * supports redirection to the egress interface, and accepts no 1117 * flag at all. 1118 * 1119 * The same effect can also be attained with the more generic 1120 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1121 * redirect target instead of providing it directly to the helper. 1122 * Return 1123 * For XDP, the helper returns **XDP_REDIRECT** on success or 1124 * **XDP_ABORTED** on error. For other program types, the values 1125 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1126 * error. 1127 * 1128 * u32 bpf_get_route_realm(struct sk_buff *skb) 1129 * Description 1130 * Retrieve the realm or the route, that is to say the 1131 * **tclassid** field of the destination for the *skb*. The 1132 * indentifier retrieved is a user-provided tag, similar to the 1133 * one used with the net_cls cgroup (see description for 1134 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1135 * held by a route (a destination entry), not by a task. 1136 * 1137 * Retrieving this identifier works with the clsact TC egress hook 1138 * (see also **tc-bpf(8)**), or alternatively on conventional 1139 * classful egress qdiscs, but not on TC ingress path. In case of 1140 * clsact TC egress hook, this has the advantage that, internally, 1141 * the destination entry has not been dropped yet in the transmit 1142 * path. Therefore, the destination entry does not need to be 1143 * artificially held via **netif_keep_dst**\ () for a classful 1144 * qdisc until the *skb* is freed. 1145 * 1146 * This helper is available only if the kernel was compiled with 1147 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1148 * Return 1149 * The realm of the route for the packet associated to *skb*, or 0 1150 * if none was found. 1151 * 1152 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1153 * Description 1154 * Write raw *data* blob into a special BPF perf event held by 1155 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1156 * event must have the following attributes: **PERF_SAMPLE_RAW** 1157 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1158 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1159 * 1160 * The *flags* are used to indicate the index in *map* for which 1161 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1162 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1163 * to indicate that the index of the current CPU core should be 1164 * used. 1165 * 1166 * The value to write, of *size*, is passed through eBPF stack and 1167 * pointed by *data*. 1168 * 1169 * The context of the program *ctx* needs also be passed to the 1170 * helper. 1171 * 1172 * On user space, a program willing to read the values needs to 1173 * call **perf_event_open**\ () on the perf event (either for 1174 * one or for all CPUs) and to store the file descriptor into the 1175 * *map*. This must be done before the eBPF program can send data 1176 * into it. An example is available in file 1177 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1178 * tree (the eBPF program counterpart is in 1179 * *samples/bpf/trace_output_kern.c*). 1180 * 1181 * **bpf_perf_event_output**\ () achieves better performance 1182 * than **bpf_trace_printk**\ () for sharing data with user 1183 * space, and is much better suitable for streaming data from eBPF 1184 * programs. 1185 * 1186 * Note that this helper is not restricted to tracing use cases 1187 * and can be used with programs attached to TC or XDP as well, 1188 * where it allows for passing data to user space listeners. Data 1189 * can be: 1190 * 1191 * * Only custom structs, 1192 * * Only the packet payload, or 1193 * * A combination of both. 1194 * Return 1195 * 0 on success, or a negative error in case of failure. 1196 * 1197 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1198 * Description 1199 * This helper was provided as an easy way to load data from a 1200 * packet. It can be used to load *len* bytes from *offset* from 1201 * the packet associated to *skb*, into the buffer pointed by 1202 * *to*. 1203 * 1204 * Since Linux 4.7, usage of this helper has mostly been replaced 1205 * by "direct packet access", enabling packet data to be 1206 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1207 * pointing respectively to the first byte of packet data and to 1208 * the byte after the last byte of packet data. However, it 1209 * remains useful if one wishes to read large quantities of data 1210 * at once from a packet into the eBPF stack. 1211 * Return 1212 * 0 on success, or a negative error in case of failure. 1213 * 1214 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1215 * Description 1216 * Walk a user or a kernel stack and return its id. To achieve 1217 * this, the helper needs *ctx*, which is a pointer to the context 1218 * on which the tracing program is executed, and a pointer to a 1219 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1220 * 1221 * The last argument, *flags*, holds the number of stack frames to 1222 * skip (from 0 to 255), masked with 1223 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1224 * a combination of the following flags: 1225 * 1226 * **BPF_F_USER_STACK** 1227 * Collect a user space stack instead of a kernel stack. 1228 * **BPF_F_FAST_STACK_CMP** 1229 * Compare stacks by hash only. 1230 * **BPF_F_REUSE_STACKID** 1231 * If two different stacks hash into the same *stackid*, 1232 * discard the old one. 1233 * 1234 * The stack id retrieved is a 32 bit long integer handle which 1235 * can be further combined with other data (including other stack 1236 * ids) and used as a key into maps. This can be useful for 1237 * generating a variety of graphs (such as flame graphs or off-cpu 1238 * graphs). 1239 * 1240 * For walking a stack, this helper is an improvement over 1241 * **bpf_probe_read**\ (), which can be used with unrolled loops 1242 * but is not efficient and consumes a lot of eBPF instructions. 1243 * Instead, **bpf_get_stackid**\ () can collect up to 1244 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1245 * this limit can be controlled with the **sysctl** program, and 1246 * that it should be manually increased in order to profile long 1247 * user stacks (such as stacks for Java programs). To do so, use: 1248 * 1249 * :: 1250 * 1251 * # sysctl kernel.perf_event_max_stack=<new value> 1252 * Return 1253 * The positive or null stack id on success, or a negative error 1254 * in case of failure. 1255 * 1256 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1257 * Description 1258 * Compute a checksum difference, from the raw buffer pointed by 1259 * *from*, of length *from_size* (that must be a multiple of 4), 1260 * towards the raw buffer pointed by *to*, of size *to_size* 1261 * (same remark). An optional *seed* can be added to the value 1262 * (this can be cascaded, the seed may come from a previous call 1263 * to the helper). 1264 * 1265 * This is flexible enough to be used in several ways: 1266 * 1267 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1268 * checksum, it can be used when pushing new data. 1269 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1270 * checksum, it can be used when removing data from a packet. 1271 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1272 * can be used to compute a diff. Note that *from_size* and 1273 * *to_size* do not need to be equal. 1274 * 1275 * This helper can be used in combination with 1276 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1277 * which one can feed in the difference computed with 1278 * **bpf_csum_diff**\ (). 1279 * Return 1280 * The checksum result, or a negative error code in case of 1281 * failure. 1282 * 1283 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1284 * Description 1285 * Retrieve tunnel options metadata for the packet associated to 1286 * *skb*, and store the raw tunnel option data to the buffer *opt* 1287 * of *size*. 1288 * 1289 * This helper can be used with encapsulation devices that can 1290 * operate in "collect metadata" mode (please refer to the related 1291 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1292 * more details). A particular example where this can be used is 1293 * in combination with the Geneve encapsulation protocol, where it 1294 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1295 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1296 * the eBPF program. This allows for full customization of these 1297 * headers. 1298 * Return 1299 * The size of the option data retrieved. 1300 * 1301 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1302 * Description 1303 * Set tunnel options metadata for the packet associated to *skb* 1304 * to the option data contained in the raw buffer *opt* of *size*. 1305 * 1306 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1307 * helper for additional information. 1308 * Return 1309 * 0 on success, or a negative error in case of failure. 1310 * 1311 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1312 * Description 1313 * Change the protocol of the *skb* to *proto*. Currently 1314 * supported are transition from IPv4 to IPv6, and from IPv6 to 1315 * IPv4. The helper takes care of the groundwork for the 1316 * transition, including resizing the socket buffer. The eBPF 1317 * program is expected to fill the new headers, if any, via 1318 * **skb_store_bytes**\ () and to recompute the checksums with 1319 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1320 * (). The main case for this helper is to perform NAT64 1321 * operations out of an eBPF program. 1322 * 1323 * Internally, the GSO type is marked as dodgy so that headers are 1324 * checked and segments are recalculated by the GSO/GRO engine. 1325 * The size for GSO target is adapted as well. 1326 * 1327 * All values for *flags* are reserved for future usage, and must 1328 * be left at zero. 1329 * 1330 * A call to this helper is susceptible to change the underlying 1331 * packet buffer. Therefore, at load time, all checks on pointers 1332 * previously done by the verifier are invalidated and must be 1333 * performed again, if the helper is used in combination with 1334 * direct packet access. 1335 * Return 1336 * 0 on success, or a negative error in case of failure. 1337 * 1338 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 1339 * Description 1340 * Change the packet type for the packet associated to *skb*. This 1341 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1342 * the eBPF program does not have a write access to *skb*\ 1343 * **->pkt_type** beside this helper. Using a helper here allows 1344 * for graceful handling of errors. 1345 * 1346 * The major use case is to change incoming *skb*s to 1347 * **PACKET_HOST** in a programmatic way instead of having to 1348 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1349 * example. 1350 * 1351 * Note that *type* only allows certain values. At this time, they 1352 * are: 1353 * 1354 * **PACKET_HOST** 1355 * Packet is for us. 1356 * **PACKET_BROADCAST** 1357 * Send packet to all. 1358 * **PACKET_MULTICAST** 1359 * Send packet to group. 1360 * **PACKET_OTHERHOST** 1361 * Send packet to someone else. 1362 * Return 1363 * 0 on success, or a negative error in case of failure. 1364 * 1365 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1366 * Description 1367 * Check whether *skb* is a descendant of the cgroup2 held by 1368 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1369 * Return 1370 * The return value depends on the result of the test, and can be: 1371 * 1372 * * 0, if the *skb* failed the cgroup2 descendant test. 1373 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1374 * * A negative error code, if an error occurred. 1375 * 1376 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1377 * Description 1378 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1379 * not set, in particular if the hash was cleared due to mangling, 1380 * recompute this hash. Later accesses to the hash can be done 1381 * directly with *skb*\ **->hash**. 1382 * 1383 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1384 * prototype with **bpf_skb_change_proto**\ (), or calling 1385 * **bpf_skb_store_bytes**\ () with the 1386 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1387 * the hash and to trigger a new computation for the next call to 1388 * **bpf_get_hash_recalc**\ (). 1389 * Return 1390 * The 32-bit hash. 1391 * 1392 * u64 bpf_get_current_task(void) 1393 * Return 1394 * A pointer to the current task struct. 1395 * 1396 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 1397 * Description 1398 * Attempt in a safe way to write *len* bytes from the buffer 1399 * *src* to *dst* in memory. It only works for threads that are in 1400 * user context, and *dst* must be a valid user space address. 1401 * 1402 * This helper should not be used to implement any kind of 1403 * security mechanism because of TOC-TOU attacks, but rather to 1404 * debug, divert, and manipulate execution of semi-cooperative 1405 * processes. 1406 * 1407 * Keep in mind that this feature is meant for experiments, and it 1408 * has a risk of crashing the system and running programs. 1409 * Therefore, when an eBPF program using this helper is attached, 1410 * a warning including PID and process name is printed to kernel 1411 * logs. 1412 * Return 1413 * 0 on success, or a negative error in case of failure. 1414 * 1415 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1416 * Description 1417 * Check whether the probe is being run is the context of a given 1418 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1419 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1420 * Return 1421 * The return value depends on the result of the test, and can be: 1422 * 1423 * * 0, if the *skb* task belongs to the cgroup2. 1424 * * 1, if the *skb* task does not belong to the cgroup2. 1425 * * A negative error code, if an error occurred. 1426 * 1427 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1428 * Description 1429 * Resize (trim or grow) the packet associated to *skb* to the 1430 * new *len*. The *flags* are reserved for future usage, and must 1431 * be left at zero. 1432 * 1433 * The basic idea is that the helper performs the needed work to 1434 * change the size of the packet, then the eBPF program rewrites 1435 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1436 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1437 * and others. This helper is a slow path utility intended for 1438 * replies with control messages. And because it is targeted for 1439 * slow path, the helper itself can afford to be slow: it 1440 * implicitly linearizes, unclones and drops offloads from the 1441 * *skb*. 1442 * 1443 * A call to this helper is susceptible to change the underlying 1444 * packet buffer. Therefore, at load time, all checks on pointers 1445 * previously done by the verifier are invalidated and must be 1446 * performed again, if the helper is used in combination with 1447 * direct packet access. 1448 * Return 1449 * 0 on success, or a negative error in case of failure. 1450 * 1451 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1452 * Description 1453 * Pull in non-linear data in case the *skb* is non-linear and not 1454 * all of *len* are part of the linear section. Make *len* bytes 1455 * from *skb* readable and writable. If a zero value is passed for 1456 * *len*, then the whole length of the *skb* is pulled. 1457 * 1458 * This helper is only needed for reading and writing with direct 1459 * packet access. 1460 * 1461 * For direct packet access, testing that offsets to access 1462 * are within packet boundaries (test on *skb*\ **->data_end**) is 1463 * susceptible to fail if offsets are invalid, or if the requested 1464 * data is in non-linear parts of the *skb*. On failure the 1465 * program can just bail out, or in the case of a non-linear 1466 * buffer, use a helper to make the data available. The 1467 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1468 * the data. Another one consists in using **bpf_skb_pull_data** 1469 * to pull in once the non-linear parts, then retesting and 1470 * eventually access the data. 1471 * 1472 * At the same time, this also makes sure the *skb* is uncloned, 1473 * which is a necessary condition for direct write. As this needs 1474 * to be an invariant for the write part only, the verifier 1475 * detects writes and adds a prologue that is calling 1476 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1477 * the very beginning in case it is indeed cloned. 1478 * 1479 * A call to this helper is susceptible to change the underlying 1480 * packet buffer. Therefore, at load time, all checks on pointers 1481 * previously done by the verifier are invalidated and must be 1482 * performed again, if the helper is used in combination with 1483 * direct packet access. 1484 * Return 1485 * 0 on success, or a negative error in case of failure. 1486 * 1487 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1488 * Description 1489 * Add the checksum *csum* into *skb*\ **->csum** in case the 1490 * driver has supplied a checksum for the entire packet into that 1491 * field. Return an error otherwise. This helper is intended to be 1492 * used in combination with **bpf_csum_diff**\ (), in particular 1493 * when the checksum needs to be updated after data has been 1494 * written into the packet through direct packet access. 1495 * Return 1496 * The checksum on success, or a negative error code in case of 1497 * failure. 1498 * 1499 * void bpf_set_hash_invalid(struct sk_buff *skb) 1500 * Description 1501 * Invalidate the current *skb*\ **->hash**. It can be used after 1502 * mangling on headers through direct packet access, in order to 1503 * indicate that the hash is outdated and to trigger a 1504 * recalculation the next time the kernel tries to access this 1505 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1506 * 1507 * long bpf_get_numa_node_id(void) 1508 * Description 1509 * Return the id of the current NUMA node. The primary use case 1510 * for this helper is the selection of sockets for the local NUMA 1511 * node, when the program is attached to sockets using the 1512 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1513 * but the helper is also available to other eBPF program types, 1514 * similarly to **bpf_get_smp_processor_id**\ (). 1515 * Return 1516 * The id of current NUMA node. 1517 * 1518 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1519 * Description 1520 * Grows headroom of packet associated to *skb* and adjusts the 1521 * offset of the MAC header accordingly, adding *len* bytes of 1522 * space. It automatically extends and reallocates memory as 1523 * required. 1524 * 1525 * This helper can be used on a layer 3 *skb* to push a MAC header 1526 * for redirection into a layer 2 device. 1527 * 1528 * All values for *flags* are reserved for future usage, and must 1529 * be left at zero. 1530 * 1531 * A call to this helper is susceptible to change the underlying 1532 * packet buffer. Therefore, at load time, all checks on pointers 1533 * previously done by the verifier are invalidated and must be 1534 * performed again, if the helper is used in combination with 1535 * direct packet access. 1536 * Return 1537 * 0 on success, or a negative error in case of failure. 1538 * 1539 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1540 * Description 1541 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1542 * it is possible to use a negative value for *delta*. This helper 1543 * can be used to prepare the packet for pushing or popping 1544 * headers. 1545 * 1546 * A call to this helper is susceptible to change the underlying 1547 * packet buffer. Therefore, at load time, all checks on pointers 1548 * previously done by the verifier are invalidated and must be 1549 * performed again, if the helper is used in combination with 1550 * direct packet access. 1551 * Return 1552 * 0 on success, or a negative error in case of failure. 1553 * 1554 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1555 * Description 1556 * Copy a NUL terminated string from an unsafe kernel address 1557 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1558 * more details. 1559 * 1560 * Generally, use **bpf_probe_read_user_str**\ () or 1561 * **bpf_probe_read_kernel_str**\ () instead. 1562 * Return 1563 * On success, the strictly positive length of the string, 1564 * including the trailing NUL character. On error, a negative 1565 * value. 1566 * 1567 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1568 * Description 1569 * If the **struct sk_buff** pointed by *skb* has a known socket, 1570 * retrieve the cookie (generated by the kernel) of this socket. 1571 * If no cookie has been set yet, generate a new cookie. Once 1572 * generated, the socket cookie remains stable for the life of the 1573 * socket. This helper can be useful for monitoring per socket 1574 * networking traffic statistics as it provides a global socket 1575 * identifier that can be assumed unique. 1576 * Return 1577 * A 8-byte long non-decreasing number on success, or 0 if the 1578 * socket field is missing inside *skb*. 1579 * 1580 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1581 * Description 1582 * Equivalent to bpf_get_socket_cookie() helper that accepts 1583 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1584 * Return 1585 * A 8-byte long non-decreasing number. 1586 * 1587 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1588 * Description 1589 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 1590 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1591 * Return 1592 * A 8-byte long non-decreasing number. 1593 * 1594 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1595 * Return 1596 * The owner UID of the socket associated to *skb*. If the socket 1597 * is **NULL**, or if it is not a full socket (i.e. if it is a 1598 * time-wait or a request socket instead), **overflowuid** value 1599 * is returned (note that **overflowuid** might also be the actual 1600 * UID value for the socket). 1601 * 1602 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 1603 * Description 1604 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1605 * to value *hash*. 1606 * Return 1607 * 0 1608 * 1609 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1610 * Description 1611 * Emulate a call to **setsockopt()** on the socket associated to 1612 * *bpf_socket*, which must be a full socket. The *level* at 1613 * which the option resides and the name *optname* of the option 1614 * must be specified, see **setsockopt(2)** for more information. 1615 * The option value of length *optlen* is pointed by *optval*. 1616 * 1617 * *bpf_socket* should be one of the following: 1618 * 1619 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1620 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1621 * and **BPF_CGROUP_INET6_CONNECT**. 1622 * 1623 * This helper actually implements a subset of **setsockopt()**. 1624 * It supports the following *level*\ s: 1625 * 1626 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1627 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1628 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1629 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 1630 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1631 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1632 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1633 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1634 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**. 1635 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1636 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1637 * Return 1638 * 0 on success, or a negative error in case of failure. 1639 * 1640 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1641 * Description 1642 * Grow or shrink the room for data in the packet associated to 1643 * *skb* by *len_diff*, and according to the selected *mode*. 1644 * 1645 * By default, the helper will reset any offloaded checksum 1646 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1647 * by the following flag: 1648 * 1649 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1650 * checksum data of the skb to CHECKSUM_NONE. 1651 * 1652 * There are two supported modes at this time: 1653 * 1654 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1655 * (room space is added or removed below the layer 2 header). 1656 * 1657 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1658 * (room space is added or removed below the layer 3 header). 1659 * 1660 * The following flags are supported at this time: 1661 * 1662 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1663 * Adjusting mss in this way is not allowed for datagrams. 1664 * 1665 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1666 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1667 * Any new space is reserved to hold a tunnel header. 1668 * Configure skb offsets and other fields accordingly. 1669 * 1670 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1671 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1672 * Use with ENCAP_L3 flags to further specify the tunnel type. 1673 * 1674 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1675 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1676 * type; *len* is the length of the inner MAC header. 1677 * 1678 * A call to this helper is susceptible to change the underlying 1679 * packet buffer. Therefore, at load time, all checks on pointers 1680 * previously done by the verifier are invalidated and must be 1681 * performed again, if the helper is used in combination with 1682 * direct packet access. 1683 * Return 1684 * 0 on success, or a negative error in case of failure. 1685 * 1686 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1687 * Description 1688 * Redirect the packet to the endpoint referenced by *map* at 1689 * index *key*. Depending on its type, this *map* can contain 1690 * references to net devices (for forwarding packets through other 1691 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1692 * but this is only implemented for native XDP (with driver 1693 * support) as of this writing). 1694 * 1695 * The lower two bits of *flags* are used as the return code if 1696 * the map lookup fails. This is so that the return value can be 1697 * one of the XDP program return codes up to **XDP_TX**, as chosen 1698 * by the caller. Any higher bits in the *flags* argument must be 1699 * unset. 1700 * 1701 * See also **bpf_redirect**\ (), which only supports redirecting 1702 * to an ifindex, but doesn't require a map to do so. 1703 * Return 1704 * **XDP_REDIRECT** on success, or the value of the two lower bits 1705 * of the *flags* argument on error. 1706 * 1707 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1708 * Description 1709 * Redirect the packet to the socket referenced by *map* (of type 1710 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1711 * egress interfaces can be used for redirection. The 1712 * **BPF_F_INGRESS** value in *flags* is used to make the 1713 * distinction (ingress path is selected if the flag is present, 1714 * egress path otherwise). This is the only flag supported for now. 1715 * Return 1716 * **SK_PASS** on success, or **SK_DROP** on error. 1717 * 1718 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1719 * Description 1720 * Add an entry to, or update a *map* referencing sockets. The 1721 * *skops* is used as a new value for the entry associated to 1722 * *key*. *flags* is one of: 1723 * 1724 * **BPF_NOEXIST** 1725 * The entry for *key* must not exist in the map. 1726 * **BPF_EXIST** 1727 * The entry for *key* must already exist in the map. 1728 * **BPF_ANY** 1729 * No condition on the existence of the entry for *key*. 1730 * 1731 * If the *map* has eBPF programs (parser and verdict), those will 1732 * be inherited by the socket being added. If the socket is 1733 * already attached to eBPF programs, this results in an error. 1734 * Return 1735 * 0 on success, or a negative error in case of failure. 1736 * 1737 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1738 * Description 1739 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1740 * *delta* (which can be positive or negative). Note that this 1741 * operation modifies the address stored in *xdp_md*\ **->data**, 1742 * so the latter must be loaded only after the helper has been 1743 * called. 1744 * 1745 * The use of *xdp_md*\ **->data_meta** is optional and programs 1746 * are not required to use it. The rationale is that when the 1747 * packet is processed with XDP (e.g. as DoS filter), it is 1748 * possible to push further meta data along with it before passing 1749 * to the stack, and to give the guarantee that an ingress eBPF 1750 * program attached as a TC classifier on the same device can pick 1751 * this up for further post-processing. Since TC works with socket 1752 * buffers, it remains possible to set from XDP the **mark** or 1753 * **priority** pointers, or other pointers for the socket buffer. 1754 * Having this scratch space generic and programmable allows for 1755 * more flexibility as the user is free to store whatever meta 1756 * data they need. 1757 * 1758 * A call to this helper is susceptible to change the underlying 1759 * packet buffer. Therefore, at load time, all checks on pointers 1760 * previously done by the verifier are invalidated and must be 1761 * performed again, if the helper is used in combination with 1762 * direct packet access. 1763 * Return 1764 * 0 on success, or a negative error in case of failure. 1765 * 1766 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1767 * Description 1768 * Read the value of a perf event counter, and store it into *buf* 1769 * of size *buf_size*. This helper relies on a *map* of type 1770 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1771 * counter is selected when *map* is updated with perf event file 1772 * descriptors. The *map* is an array whose size is the number of 1773 * available CPUs, and each cell contains a value relative to one 1774 * CPU. The value to retrieve is indicated by *flags*, that 1775 * contains the index of the CPU to look up, masked with 1776 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1777 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1778 * current CPU should be retrieved. 1779 * 1780 * This helper behaves in a way close to 1781 * **bpf_perf_event_read**\ () helper, save that instead of 1782 * just returning the value observed, it fills the *buf* 1783 * structure. This allows for additional data to be retrieved: in 1784 * particular, the enabled and running times (in *buf*\ 1785 * **->enabled** and *buf*\ **->running**, respectively) are 1786 * copied. In general, **bpf_perf_event_read_value**\ () is 1787 * recommended over **bpf_perf_event_read**\ (), which has some 1788 * ABI issues and provides fewer functionalities. 1789 * 1790 * These values are interesting, because hardware PMU (Performance 1791 * Monitoring Unit) counters are limited resources. When there are 1792 * more PMU based perf events opened than available counters, 1793 * kernel will multiplex these events so each event gets certain 1794 * percentage (but not all) of the PMU time. In case that 1795 * multiplexing happens, the number of samples or counter value 1796 * will not reflect the case compared to when no multiplexing 1797 * occurs. This makes comparison between different runs difficult. 1798 * Typically, the counter value should be normalized before 1799 * comparing to other experiments. The usual normalization is done 1800 * as follows. 1801 * 1802 * :: 1803 * 1804 * normalized_counter = counter * t_enabled / t_running 1805 * 1806 * Where t_enabled is the time enabled for event and t_running is 1807 * the time running for event since last normalization. The 1808 * enabled and running times are accumulated since the perf event 1809 * open. To achieve scaling factor between two invocations of an 1810 * eBPF program, users can use CPU id as the key (which is 1811 * typical for perf array usage model) to remember the previous 1812 * value and do the calculation inside the eBPF program. 1813 * Return 1814 * 0 on success, or a negative error in case of failure. 1815 * 1816 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1817 * Description 1818 * For en eBPF program attached to a perf event, retrieve the 1819 * value of the event counter associated to *ctx* and store it in 1820 * the structure pointed by *buf* and of size *buf_size*. Enabled 1821 * and running times are also stored in the structure (see 1822 * description of helper **bpf_perf_event_read_value**\ () for 1823 * more details). 1824 * Return 1825 * 0 on success, or a negative error in case of failure. 1826 * 1827 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1828 * Description 1829 * Emulate a call to **getsockopt()** on the socket associated to 1830 * *bpf_socket*, which must be a full socket. The *level* at 1831 * which the option resides and the name *optname* of the option 1832 * must be specified, see **getsockopt(2)** for more information. 1833 * The retrieved value is stored in the structure pointed by 1834 * *opval* and of length *optlen*. 1835 * 1836 * *bpf_socket* should be one of the following: 1837 * 1838 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1839 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1840 * and **BPF_CGROUP_INET6_CONNECT**. 1841 * 1842 * This helper actually implements a subset of **getsockopt()**. 1843 * It supports the following *level*\ s: 1844 * 1845 * * **IPPROTO_TCP**, which supports *optname* 1846 * **TCP_CONGESTION**. 1847 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1848 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1849 * Return 1850 * 0 on success, or a negative error in case of failure. 1851 * 1852 * long bpf_override_return(struct pt_regs *regs, u64 rc) 1853 * Description 1854 * Used for error injection, this helper uses kprobes to override 1855 * the return value of the probed function, and to set it to *rc*. 1856 * The first argument is the context *regs* on which the kprobe 1857 * works. 1858 * 1859 * This helper works by setting the PC (program counter) 1860 * to an override function which is run in place of the original 1861 * probed function. This means the probed function is not run at 1862 * all. The replacement function just returns with the required 1863 * value. 1864 * 1865 * This helper has security implications, and thus is subject to 1866 * restrictions. It is only available if the kernel was compiled 1867 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1868 * option, and in this case it only works on functions tagged with 1869 * **ALLOW_ERROR_INJECTION** in the kernel code. 1870 * 1871 * Also, the helper is only available for the architectures having 1872 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1873 * x86 architecture is the only one to support this feature. 1874 * Return 1875 * 0 1876 * 1877 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1878 * Description 1879 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1880 * for the full TCP socket associated to *bpf_sock_ops* to 1881 * *argval*. 1882 * 1883 * The primary use of this field is to determine if there should 1884 * be calls to eBPF programs of type 1885 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1886 * code. A program of the same type can change its value, per 1887 * connection and as necessary, when the connection is 1888 * established. This field is directly accessible for reading, but 1889 * this helper must be used for updates in order to return an 1890 * error if an eBPF program tries to set a callback that is not 1891 * supported in the current kernel. 1892 * 1893 * *argval* is a flag array which can combine these flags: 1894 * 1895 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1896 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1897 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1898 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1899 * 1900 * Therefore, this function can be used to clear a callback flag by 1901 * setting the appropriate bit to zero. e.g. to disable the RTO 1902 * callback: 1903 * 1904 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1905 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1906 * 1907 * Here are some examples of where one could call such eBPF 1908 * program: 1909 * 1910 * * When RTO fires. 1911 * * When a packet is retransmitted. 1912 * * When the connection terminates. 1913 * * When a packet is sent. 1914 * * When a packet is received. 1915 * Return 1916 * Code **-EINVAL** if the socket is not a full TCP socket; 1917 * otherwise, a positive number containing the bits that could not 1918 * be set is returned (which comes down to 0 if all bits were set 1919 * as required). 1920 * 1921 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1922 * Description 1923 * This helper is used in programs implementing policies at the 1924 * socket level. If the message *msg* is allowed to pass (i.e. if 1925 * the verdict eBPF program returns **SK_PASS**), redirect it to 1926 * the socket referenced by *map* (of type 1927 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1928 * egress interfaces can be used for redirection. The 1929 * **BPF_F_INGRESS** value in *flags* is used to make the 1930 * distinction (ingress path is selected if the flag is present, 1931 * egress path otherwise). This is the only flag supported for now. 1932 * Return 1933 * **SK_PASS** on success, or **SK_DROP** on error. 1934 * 1935 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1936 * Description 1937 * For socket policies, apply the verdict of the eBPF program to 1938 * the next *bytes* (number of bytes) of message *msg*. 1939 * 1940 * For example, this helper can be used in the following cases: 1941 * 1942 * * A single **sendmsg**\ () or **sendfile**\ () system call 1943 * contains multiple logical messages that the eBPF program is 1944 * supposed to read and for which it should apply a verdict. 1945 * * An eBPF program only cares to read the first *bytes* of a 1946 * *msg*. If the message has a large payload, then setting up 1947 * and calling the eBPF program repeatedly for all bytes, even 1948 * though the verdict is already known, would create unnecessary 1949 * overhead. 1950 * 1951 * When called from within an eBPF program, the helper sets a 1952 * counter internal to the BPF infrastructure, that is used to 1953 * apply the last verdict to the next *bytes*. If *bytes* is 1954 * smaller than the current data being processed from a 1955 * **sendmsg**\ () or **sendfile**\ () system call, the first 1956 * *bytes* will be sent and the eBPF program will be re-run with 1957 * the pointer for start of data pointing to byte number *bytes* 1958 * **+ 1**. If *bytes* is larger than the current data being 1959 * processed, then the eBPF verdict will be applied to multiple 1960 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1961 * consumed. 1962 * 1963 * Note that if a socket closes with the internal counter holding 1964 * a non-zero value, this is not a problem because data is not 1965 * being buffered for *bytes* and is sent as it is received. 1966 * Return 1967 * 0 1968 * 1969 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1970 * Description 1971 * For socket policies, prevent the execution of the verdict eBPF 1972 * program for message *msg* until *bytes* (byte number) have been 1973 * accumulated. 1974 * 1975 * This can be used when one needs a specific number of bytes 1976 * before a verdict can be assigned, even if the data spans 1977 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1978 * case would be a user calling **sendmsg**\ () repeatedly with 1979 * 1-byte long message segments. Obviously, this is bad for 1980 * performance, but it is still valid. If the eBPF program needs 1981 * *bytes* bytes to validate a header, this helper can be used to 1982 * prevent the eBPF program to be called again until *bytes* have 1983 * been accumulated. 1984 * Return 1985 * 0 1986 * 1987 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1988 * Description 1989 * For socket policies, pull in non-linear data from user space 1990 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1991 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1992 * respectively. 1993 * 1994 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1995 * *msg* it can only parse data that the (**data**, **data_end**) 1996 * pointers have already consumed. For **sendmsg**\ () hooks this 1997 * is likely the first scatterlist element. But for calls relying 1998 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1999 * be the range (**0**, **0**) because the data is shared with 2000 * user space and by default the objective is to avoid allowing 2001 * user space to modify data while (or after) eBPF verdict is 2002 * being decided. This helper can be used to pull in data and to 2003 * set the start and end pointer to given values. Data will be 2004 * copied if necessary (i.e. if data was not linear and if start 2005 * and end pointers do not point to the same chunk). 2006 * 2007 * A call to this helper is susceptible to change the underlying 2008 * packet buffer. Therefore, at load time, all checks on pointers 2009 * previously done by the verifier are invalidated and must be 2010 * performed again, if the helper is used in combination with 2011 * direct packet access. 2012 * 2013 * All values for *flags* are reserved for future usage, and must 2014 * be left at zero. 2015 * Return 2016 * 0 on success, or a negative error in case of failure. 2017 * 2018 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2019 * Description 2020 * Bind the socket associated to *ctx* to the address pointed by 2021 * *addr*, of length *addr_len*. This allows for making outgoing 2022 * connection from the desired IP address, which can be useful for 2023 * example when all processes inside a cgroup should use one 2024 * single IP address on a host that has multiple IP configured. 2025 * 2026 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2027 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2028 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2029 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2030 * behavior and lets the kernel efficiently pick up an unused 2031 * port as long as 4-tuple is unique. Passing non-zero port might 2032 * lead to degraded performance. 2033 * Return 2034 * 0 on success, or a negative error in case of failure. 2035 * 2036 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2037 * Description 2038 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2039 * possible to both shrink and grow the packet tail. 2040 * Shrink done via *delta* being a negative integer. 2041 * 2042 * A call to this helper is susceptible to change the underlying 2043 * packet buffer. Therefore, at load time, all checks on pointers 2044 * previously done by the verifier are invalidated and must be 2045 * performed again, if the helper is used in combination with 2046 * direct packet access. 2047 * Return 2048 * 0 on success, or a negative error in case of failure. 2049 * 2050 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2051 * Description 2052 * Retrieve the XFRM state (IP transform framework, see also 2053 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2054 * 2055 * The retrieved value is stored in the **struct bpf_xfrm_state** 2056 * pointed by *xfrm_state* and of length *size*. 2057 * 2058 * All values for *flags* are reserved for future usage, and must 2059 * be left at zero. 2060 * 2061 * This helper is available only if the kernel was compiled with 2062 * **CONFIG_XFRM** configuration option. 2063 * Return 2064 * 0 on success, or a negative error in case of failure. 2065 * 2066 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2067 * Description 2068 * Return a user or a kernel stack in bpf program provided buffer. 2069 * To achieve this, the helper needs *ctx*, which is a pointer 2070 * to the context on which the tracing program is executed. 2071 * To store the stacktrace, the bpf program provides *buf* with 2072 * a nonnegative *size*. 2073 * 2074 * The last argument, *flags*, holds the number of stack frames to 2075 * skip (from 0 to 255), masked with 2076 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2077 * the following flags: 2078 * 2079 * **BPF_F_USER_STACK** 2080 * Collect a user space stack instead of a kernel stack. 2081 * **BPF_F_USER_BUILD_ID** 2082 * Collect buildid+offset instead of ips for user stack, 2083 * only valid if **BPF_F_USER_STACK** is also specified. 2084 * 2085 * **bpf_get_stack**\ () can collect up to 2086 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2087 * to sufficient large buffer size. Note that 2088 * this limit can be controlled with the **sysctl** program, and 2089 * that it should be manually increased in order to profile long 2090 * user stacks (such as stacks for Java programs). To do so, use: 2091 * 2092 * :: 2093 * 2094 * # sysctl kernel.perf_event_max_stack=<new value> 2095 * Return 2096 * A non-negative value equal to or less than *size* on success, 2097 * or a negative error in case of failure. 2098 * 2099 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2100 * Description 2101 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2102 * it provides an easy way to load *len* bytes from *offset* 2103 * from the packet associated to *skb*, into the buffer pointed 2104 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2105 * a fifth argument *start_header* exists in order to select a 2106 * base offset to start from. *start_header* can be one of: 2107 * 2108 * **BPF_HDR_START_MAC** 2109 * Base offset to load data from is *skb*'s mac header. 2110 * **BPF_HDR_START_NET** 2111 * Base offset to load data from is *skb*'s network header. 2112 * 2113 * In general, "direct packet access" is the preferred method to 2114 * access packet data, however, this helper is in particular useful 2115 * in socket filters where *skb*\ **->data** does not always point 2116 * to the start of the mac header and where "direct packet access" 2117 * is not available. 2118 * Return 2119 * 0 on success, or a negative error in case of failure. 2120 * 2121 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2122 * Description 2123 * Do FIB lookup in kernel tables using parameters in *params*. 2124 * If lookup is successful and result shows packet is to be 2125 * forwarded, the neighbor tables are searched for the nexthop. 2126 * If successful (ie., FIB lookup shows forwarding and nexthop 2127 * is resolved), the nexthop address is returned in ipv4_dst 2128 * or ipv6_dst based on family, smac is set to mac address of 2129 * egress device, dmac is set to nexthop mac address, rt_metric 2130 * is set to metric from route (IPv4/IPv6 only), and ifindex 2131 * is set to the device index of the nexthop from the FIB lookup. 2132 * 2133 * *plen* argument is the size of the passed in struct. 2134 * *flags* argument can be a combination of one or more of the 2135 * following values: 2136 * 2137 * **BPF_FIB_LOOKUP_DIRECT** 2138 * Do a direct table lookup vs full lookup using FIB 2139 * rules. 2140 * **BPF_FIB_LOOKUP_OUTPUT** 2141 * Perform lookup from an egress perspective (default is 2142 * ingress). 2143 * 2144 * *ctx* is either **struct xdp_md** for XDP programs or 2145 * **struct sk_buff** tc cls_act programs. 2146 * Return 2147 * * < 0 if any input argument is invalid 2148 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2149 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2150 * packet is not forwarded or needs assist from full stack 2151 * 2152 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2153 * Description 2154 * Add an entry to, or update a sockhash *map* referencing sockets. 2155 * The *skops* is used as a new value for the entry associated to 2156 * *key*. *flags* is one of: 2157 * 2158 * **BPF_NOEXIST** 2159 * The entry for *key* must not exist in the map. 2160 * **BPF_EXIST** 2161 * The entry for *key* must already exist in the map. 2162 * **BPF_ANY** 2163 * No condition on the existence of the entry for *key*. 2164 * 2165 * If the *map* has eBPF programs (parser and verdict), those will 2166 * be inherited by the socket being added. If the socket is 2167 * already attached to eBPF programs, this results in an error. 2168 * Return 2169 * 0 on success, or a negative error in case of failure. 2170 * 2171 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2172 * Description 2173 * This helper is used in programs implementing policies at the 2174 * socket level. If the message *msg* is allowed to pass (i.e. if 2175 * the verdict eBPF program returns **SK_PASS**), redirect it to 2176 * the socket referenced by *map* (of type 2177 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2178 * egress interfaces can be used for redirection. The 2179 * **BPF_F_INGRESS** value in *flags* is used to make the 2180 * distinction (ingress path is selected if the flag is present, 2181 * egress path otherwise). This is the only flag supported for now. 2182 * Return 2183 * **SK_PASS** on success, or **SK_DROP** on error. 2184 * 2185 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2186 * Description 2187 * This helper is used in programs implementing policies at the 2188 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2189 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2190 * to the socket referenced by *map* (of type 2191 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2192 * egress interfaces can be used for redirection. The 2193 * **BPF_F_INGRESS** value in *flags* is used to make the 2194 * distinction (ingress path is selected if the flag is present, 2195 * egress otherwise). This is the only flag supported for now. 2196 * Return 2197 * **SK_PASS** on success, or **SK_DROP** on error. 2198 * 2199 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2200 * Description 2201 * Encapsulate the packet associated to *skb* within a Layer 3 2202 * protocol header. This header is provided in the buffer at 2203 * address *hdr*, with *len* its size in bytes. *type* indicates 2204 * the protocol of the header and can be one of: 2205 * 2206 * **BPF_LWT_ENCAP_SEG6** 2207 * IPv6 encapsulation with Segment Routing Header 2208 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2209 * the IPv6 header is computed by the kernel. 2210 * **BPF_LWT_ENCAP_SEG6_INLINE** 2211 * Only works if *skb* contains an IPv6 packet. Insert a 2212 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2213 * the IPv6 header. 2214 * **BPF_LWT_ENCAP_IP** 2215 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2216 * must be IPv4 or IPv6, followed by zero or more 2217 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2218 * total bytes in all prepended headers. Please note that 2219 * if **skb_is_gso**\ (*skb*) is true, no more than two 2220 * headers can be prepended, and the inner header, if 2221 * present, should be either GRE or UDP/GUE. 2222 * 2223 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2224 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2225 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2226 * **BPF_PROG_TYPE_LWT_XMIT**. 2227 * 2228 * A call to this helper is susceptible to change the underlying 2229 * packet buffer. Therefore, at load time, all checks on pointers 2230 * previously done by the verifier are invalidated and must be 2231 * performed again, if the helper is used in combination with 2232 * direct packet access. 2233 * Return 2234 * 0 on success, or a negative error in case of failure. 2235 * 2236 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2237 * Description 2238 * Store *len* bytes from address *from* into the packet 2239 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2240 * inside the outermost IPv6 Segment Routing Header can be 2241 * modified through this helper. 2242 * 2243 * A call to this helper is susceptible to change the underlying 2244 * packet buffer. Therefore, at load time, all checks on pointers 2245 * previously done by the verifier are invalidated and must be 2246 * performed again, if the helper is used in combination with 2247 * direct packet access. 2248 * Return 2249 * 0 on success, or a negative error in case of failure. 2250 * 2251 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2252 * Description 2253 * Adjust the size allocated to TLVs in the outermost IPv6 2254 * Segment Routing Header contained in the packet associated to 2255 * *skb*, at position *offset* by *delta* bytes. Only offsets 2256 * after the segments are accepted. *delta* can be as well 2257 * positive (growing) as negative (shrinking). 2258 * 2259 * A call to this helper is susceptible to change the underlying 2260 * packet buffer. Therefore, at load time, all checks on pointers 2261 * previously done by the verifier are invalidated and must be 2262 * performed again, if the helper is used in combination with 2263 * direct packet access. 2264 * Return 2265 * 0 on success, or a negative error in case of failure. 2266 * 2267 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2268 * Description 2269 * Apply an IPv6 Segment Routing action of type *action* to the 2270 * packet associated to *skb*. Each action takes a parameter 2271 * contained at address *param*, and of length *param_len* bytes. 2272 * *action* can be one of: 2273 * 2274 * **SEG6_LOCAL_ACTION_END_X** 2275 * End.X action: Endpoint with Layer-3 cross-connect. 2276 * Type of *param*: **struct in6_addr**. 2277 * **SEG6_LOCAL_ACTION_END_T** 2278 * End.T action: Endpoint with specific IPv6 table lookup. 2279 * Type of *param*: **int**. 2280 * **SEG6_LOCAL_ACTION_END_B6** 2281 * End.B6 action: Endpoint bound to an SRv6 policy. 2282 * Type of *param*: **struct ipv6_sr_hdr**. 2283 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2284 * End.B6.Encap action: Endpoint bound to an SRv6 2285 * encapsulation policy. 2286 * Type of *param*: **struct ipv6_sr_hdr**. 2287 * 2288 * A call to this helper is susceptible to change the underlying 2289 * packet buffer. Therefore, at load time, all checks on pointers 2290 * previously done by the verifier are invalidated and must be 2291 * performed again, if the helper is used in combination with 2292 * direct packet access. 2293 * Return 2294 * 0 on success, or a negative error in case of failure. 2295 * 2296 * long bpf_rc_repeat(void *ctx) 2297 * Description 2298 * This helper is used in programs implementing IR decoding, to 2299 * report a successfully decoded repeat key message. This delays 2300 * the generation of a key up event for previously generated 2301 * key down event. 2302 * 2303 * Some IR protocols like NEC have a special IR message for 2304 * repeating last button, for when a button is held down. 2305 * 2306 * The *ctx* should point to the lirc sample as passed into 2307 * the program. 2308 * 2309 * This helper is only available is the kernel was compiled with 2310 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2311 * "**y**". 2312 * Return 2313 * 0 2314 * 2315 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2316 * Description 2317 * This helper is used in programs implementing IR decoding, to 2318 * report a successfully decoded key press with *scancode*, 2319 * *toggle* value in the given *protocol*. The scancode will be 2320 * translated to a keycode using the rc keymap, and reported as 2321 * an input key down event. After a period a key up event is 2322 * generated. This period can be extended by calling either 2323 * **bpf_rc_keydown**\ () again with the same values, or calling 2324 * **bpf_rc_repeat**\ (). 2325 * 2326 * Some protocols include a toggle bit, in case the button was 2327 * released and pressed again between consecutive scancodes. 2328 * 2329 * The *ctx* should point to the lirc sample as passed into 2330 * the program. 2331 * 2332 * The *protocol* is the decoded protocol number (see 2333 * **enum rc_proto** for some predefined values). 2334 * 2335 * This helper is only available is the kernel was compiled with 2336 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2337 * "**y**". 2338 * Return 2339 * 0 2340 * 2341 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2342 * Description 2343 * Return the cgroup v2 id of the socket associated with the *skb*. 2344 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2345 * helper for cgroup v1 by providing a tag resp. identifier that 2346 * can be matched on or used for map lookups e.g. to implement 2347 * policy. The cgroup v2 id of a given path in the hierarchy is 2348 * exposed in user space through the f_handle API in order to get 2349 * to the same 64-bit id. 2350 * 2351 * This helper can be used on TC egress path, but not on ingress, 2352 * and is available only if the kernel was compiled with the 2353 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2354 * Return 2355 * The id is returned or 0 in case the id could not be retrieved. 2356 * 2357 * u64 bpf_get_current_cgroup_id(void) 2358 * Return 2359 * A 64-bit integer containing the current cgroup id based 2360 * on the cgroup within which the current task is running. 2361 * 2362 * void *bpf_get_local_storage(void *map, u64 flags) 2363 * Description 2364 * Get the pointer to the local storage area. 2365 * The type and the size of the local storage is defined 2366 * by the *map* argument. 2367 * The *flags* meaning is specific for each map type, 2368 * and has to be 0 for cgroup local storage. 2369 * 2370 * Depending on the BPF program type, a local storage area 2371 * can be shared between multiple instances of the BPF program, 2372 * running simultaneously. 2373 * 2374 * A user should care about the synchronization by himself. 2375 * For example, by using the **BPF_STX_XADD** instruction to alter 2376 * the shared data. 2377 * Return 2378 * A pointer to the local storage area. 2379 * 2380 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2381 * Description 2382 * Select a **SO_REUSEPORT** socket from a 2383 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2384 * It checks the selected socket is matching the incoming 2385 * request in the socket buffer. 2386 * Return 2387 * 0 on success, or a negative error in case of failure. 2388 * 2389 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2390 * Description 2391 * Return id of cgroup v2 that is ancestor of cgroup associated 2392 * with the *skb* at the *ancestor_level*. The root cgroup is at 2393 * *ancestor_level* zero and each step down the hierarchy 2394 * increments the level. If *ancestor_level* == level of cgroup 2395 * associated with *skb*, then return value will be same as that 2396 * of **bpf_skb_cgroup_id**\ (). 2397 * 2398 * The helper is useful to implement policies based on cgroups 2399 * that are upper in hierarchy than immediate cgroup associated 2400 * with *skb*. 2401 * 2402 * The format of returned id and helper limitations are same as in 2403 * **bpf_skb_cgroup_id**\ (). 2404 * Return 2405 * The id is returned or 0 in case the id could not be retrieved. 2406 * 2407 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2408 * Description 2409 * Look for TCP socket matching *tuple*, optionally in a child 2410 * network namespace *netns*. The return value must be checked, 2411 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2412 * 2413 * The *ctx* should point to the context of the program, such as 2414 * the skb or socket (depending on the hook in use). This is used 2415 * to determine the base network namespace for the lookup. 2416 * 2417 * *tuple_size* must be one of: 2418 * 2419 * **sizeof**\ (*tuple*\ **->ipv4**) 2420 * Look for an IPv4 socket. 2421 * **sizeof**\ (*tuple*\ **->ipv6**) 2422 * Look for an IPv6 socket. 2423 * 2424 * If the *netns* is a negative signed 32-bit integer, then the 2425 * socket lookup table in the netns associated with the *ctx* 2426 * will be used. For the TC hooks, this is the netns of the device 2427 * in the skb. For socket hooks, this is the netns of the socket. 2428 * If *netns* is any other signed 32-bit value greater than or 2429 * equal to zero then it specifies the ID of the netns relative to 2430 * the netns associated with the *ctx*. *netns* values beyond the 2431 * range of 32-bit integers are reserved for future use. 2432 * 2433 * All values for *flags* are reserved for future usage, and must 2434 * be left at zero. 2435 * 2436 * This helper is available only if the kernel was compiled with 2437 * **CONFIG_NET** configuration option. 2438 * Return 2439 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2440 * For sockets with reuseport option, the **struct bpf_sock** 2441 * result is from *reuse*\ **->socks**\ [] using the hash of the 2442 * tuple. 2443 * 2444 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2445 * Description 2446 * Look for UDP socket matching *tuple*, optionally in a child 2447 * network namespace *netns*. The return value must be checked, 2448 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2449 * 2450 * The *ctx* should point to the context of the program, such as 2451 * the skb or socket (depending on the hook in use). This is used 2452 * to determine the base network namespace for the lookup. 2453 * 2454 * *tuple_size* must be one of: 2455 * 2456 * **sizeof**\ (*tuple*\ **->ipv4**) 2457 * Look for an IPv4 socket. 2458 * **sizeof**\ (*tuple*\ **->ipv6**) 2459 * Look for an IPv6 socket. 2460 * 2461 * If the *netns* is a negative signed 32-bit integer, then the 2462 * socket lookup table in the netns associated with the *ctx* 2463 * will be used. For the TC hooks, this is the netns of the device 2464 * in the skb. For socket hooks, this is the netns of the socket. 2465 * If *netns* is any other signed 32-bit value greater than or 2466 * equal to zero then it specifies the ID of the netns relative to 2467 * the netns associated with the *ctx*. *netns* values beyond the 2468 * range of 32-bit integers are reserved for future use. 2469 * 2470 * All values for *flags* are reserved for future usage, and must 2471 * be left at zero. 2472 * 2473 * This helper is available only if the kernel was compiled with 2474 * **CONFIG_NET** configuration option. 2475 * Return 2476 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2477 * For sockets with reuseport option, the **struct bpf_sock** 2478 * result is from *reuse*\ **->socks**\ [] using the hash of the 2479 * tuple. 2480 * 2481 * long bpf_sk_release(struct bpf_sock *sock) 2482 * Description 2483 * Release the reference held by *sock*. *sock* must be a 2484 * non-**NULL** pointer that was returned from 2485 * **bpf_sk_lookup_xxx**\ (). 2486 * Return 2487 * 0 on success, or a negative error in case of failure. 2488 * 2489 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2490 * Description 2491 * Push an element *value* in *map*. *flags* is one of: 2492 * 2493 * **BPF_EXIST** 2494 * If the queue/stack is full, the oldest element is 2495 * removed to make room for this. 2496 * Return 2497 * 0 on success, or a negative error in case of failure. 2498 * 2499 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 2500 * Description 2501 * Pop an element from *map*. 2502 * Return 2503 * 0 on success, or a negative error in case of failure. 2504 * 2505 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 2506 * Description 2507 * Get an element from *map* without removing it. 2508 * Return 2509 * 0 on success, or a negative error in case of failure. 2510 * 2511 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2512 * Description 2513 * For socket policies, insert *len* bytes into *msg* at offset 2514 * *start*. 2515 * 2516 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2517 * *msg* it may want to insert metadata or options into the *msg*. 2518 * This can later be read and used by any of the lower layer BPF 2519 * hooks. 2520 * 2521 * This helper may fail if under memory pressure (a malloc 2522 * fails) in these cases BPF programs will get an appropriate 2523 * error and BPF programs will need to handle them. 2524 * Return 2525 * 0 on success, or a negative error in case of failure. 2526 * 2527 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2528 * Description 2529 * Will remove *len* bytes from a *msg* starting at byte *start*. 2530 * This may result in **ENOMEM** errors under certain situations if 2531 * an allocation and copy are required due to a full ring buffer. 2532 * However, the helper will try to avoid doing the allocation 2533 * if possible. Other errors can occur if input parameters are 2534 * invalid either due to *start* byte not being valid part of *msg* 2535 * payload and/or *pop* value being to large. 2536 * Return 2537 * 0 on success, or a negative error in case of failure. 2538 * 2539 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2540 * Description 2541 * This helper is used in programs implementing IR decoding, to 2542 * report a successfully decoded pointer movement. 2543 * 2544 * The *ctx* should point to the lirc sample as passed into 2545 * the program. 2546 * 2547 * This helper is only available is the kernel was compiled with 2548 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2549 * "**y**". 2550 * Return 2551 * 0 2552 * 2553 * long bpf_spin_lock(struct bpf_spin_lock *lock) 2554 * Description 2555 * Acquire a spinlock represented by the pointer *lock*, which is 2556 * stored as part of a value of a map. Taking the lock allows to 2557 * safely update the rest of the fields in that value. The 2558 * spinlock can (and must) later be released with a call to 2559 * **bpf_spin_unlock**\ (\ *lock*\ ). 2560 * 2561 * Spinlocks in BPF programs come with a number of restrictions 2562 * and constraints: 2563 * 2564 * * **bpf_spin_lock** objects are only allowed inside maps of 2565 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2566 * list could be extended in the future). 2567 * * BTF description of the map is mandatory. 2568 * * The BPF program can take ONE lock at a time, since taking two 2569 * or more could cause dead locks. 2570 * * Only one **struct bpf_spin_lock** is allowed per map element. 2571 * * When the lock is taken, calls (either BPF to BPF or helpers) 2572 * are not allowed. 2573 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2574 * allowed inside a spinlock-ed region. 2575 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2576 * the lock, on all execution paths, before it returns. 2577 * * The BPF program can access **struct bpf_spin_lock** only via 2578 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2579 * helpers. Loading or storing data into the **struct 2580 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2581 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2582 * of the map value must be a struct and have **struct 2583 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2584 * Nested lock inside another struct is not allowed. 2585 * * The **struct bpf_spin_lock** *lock* field in a map value must 2586 * be aligned on a multiple of 4 bytes in that value. 2587 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2588 * the **bpf_spin_lock** field to user space. 2589 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2590 * a BPF program, do not update the **bpf_spin_lock** field. 2591 * * **bpf_spin_lock** cannot be on the stack or inside a 2592 * networking packet (it can only be inside of a map values). 2593 * * **bpf_spin_lock** is available to root only. 2594 * * Tracing programs and socket filter programs cannot use 2595 * **bpf_spin_lock**\ () due to insufficient preemption checks 2596 * (but this may change in the future). 2597 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2598 * Return 2599 * 0 2600 * 2601 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 2602 * Description 2603 * Release the *lock* previously locked by a call to 2604 * **bpf_spin_lock**\ (\ *lock*\ ). 2605 * Return 2606 * 0 2607 * 2608 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2609 * Description 2610 * This helper gets a **struct bpf_sock** pointer such 2611 * that all the fields in this **bpf_sock** can be accessed. 2612 * Return 2613 * A **struct bpf_sock** pointer on success, or **NULL** in 2614 * case of failure. 2615 * 2616 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2617 * Description 2618 * This helper gets a **struct bpf_tcp_sock** pointer from a 2619 * **struct bpf_sock** pointer. 2620 * Return 2621 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2622 * case of failure. 2623 * 2624 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 2625 * Description 2626 * Set ECN (Explicit Congestion Notification) field of IP header 2627 * to **CE** (Congestion Encountered) if current value is **ECT** 2628 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2629 * and IPv4. 2630 * Return 2631 * 1 if the **CE** flag is set (either by the current helper call 2632 * or because it was already present), 0 if it is not set. 2633 * 2634 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2635 * Description 2636 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2637 * **bpf_sk_release**\ () is unnecessary and not allowed. 2638 * Return 2639 * A **struct bpf_sock** pointer on success, or **NULL** in 2640 * case of failure. 2641 * 2642 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2643 * Description 2644 * Look for TCP socket matching *tuple*, optionally in a child 2645 * network namespace *netns*. The return value must be checked, 2646 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2647 * 2648 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2649 * that it also returns timewait or request sockets. Use 2650 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2651 * full structure. 2652 * 2653 * This helper is available only if the kernel was compiled with 2654 * **CONFIG_NET** configuration option. 2655 * Return 2656 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2657 * For sockets with reuseport option, the **struct bpf_sock** 2658 * result is from *reuse*\ **->socks**\ [] using the hash of the 2659 * tuple. 2660 * 2661 * long bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2662 * Description 2663 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2664 * the listening socket in *sk*. 2665 * 2666 * *iph* points to the start of the IPv4 or IPv6 header, while 2667 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2668 * **sizeof**\ (**struct ip6hdr**). 2669 * 2670 * *th* points to the start of the TCP header, while *th_len* 2671 * contains **sizeof**\ (**struct tcphdr**). 2672 * Return 2673 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2674 * error otherwise. 2675 * 2676 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2677 * Description 2678 * Get name of sysctl in /proc/sys/ and copy it into provided by 2679 * program buffer *buf* of size *buf_len*. 2680 * 2681 * The buffer is always NUL terminated, unless it's zero-sized. 2682 * 2683 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2684 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2685 * only (e.g. "tcp_mem"). 2686 * Return 2687 * Number of character copied (not including the trailing NUL). 2688 * 2689 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2690 * truncated name in this case). 2691 * 2692 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2693 * Description 2694 * Get current value of sysctl as it is presented in /proc/sys 2695 * (incl. newline, etc), and copy it as a string into provided 2696 * by program buffer *buf* of size *buf_len*. 2697 * 2698 * The whole value is copied, no matter what file position user 2699 * space issued e.g. sys_read at. 2700 * 2701 * The buffer is always NUL terminated, unless it's zero-sized. 2702 * Return 2703 * Number of character copied (not including the trailing NUL). 2704 * 2705 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2706 * truncated name in this case). 2707 * 2708 * **-EINVAL** if current value was unavailable, e.g. because 2709 * sysctl is uninitialized and read returns -EIO for it. 2710 * 2711 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2712 * Description 2713 * Get new value being written by user space to sysctl (before 2714 * the actual write happens) and copy it as a string into 2715 * provided by program buffer *buf* of size *buf_len*. 2716 * 2717 * User space may write new value at file position > 0. 2718 * 2719 * The buffer is always NUL terminated, unless it's zero-sized. 2720 * Return 2721 * Number of character copied (not including the trailing NUL). 2722 * 2723 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2724 * truncated name in this case). 2725 * 2726 * **-EINVAL** if sysctl is being read. 2727 * 2728 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2729 * Description 2730 * Override new value being written by user space to sysctl with 2731 * value provided by program in buffer *buf* of size *buf_len*. 2732 * 2733 * *buf* should contain a string in same form as provided by user 2734 * space on sysctl write. 2735 * 2736 * User space may write new value at file position > 0. To override 2737 * the whole sysctl value file position should be set to zero. 2738 * Return 2739 * 0 on success. 2740 * 2741 * **-E2BIG** if the *buf_len* is too big. 2742 * 2743 * **-EINVAL** if sysctl is being read. 2744 * 2745 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2746 * Description 2747 * Convert the initial part of the string from buffer *buf* of 2748 * size *buf_len* to a long integer according to the given base 2749 * and save the result in *res*. 2750 * 2751 * The string may begin with an arbitrary amount of white space 2752 * (as determined by **isspace**\ (3)) followed by a single 2753 * optional '**-**' sign. 2754 * 2755 * Five least significant bits of *flags* encode base, other bits 2756 * are currently unused. 2757 * 2758 * Base must be either 8, 10, 16 or 0 to detect it automatically 2759 * similar to user space **strtol**\ (3). 2760 * Return 2761 * Number of characters consumed on success. Must be positive but 2762 * no more than *buf_len*. 2763 * 2764 * **-EINVAL** if no valid digits were found or unsupported base 2765 * was provided. 2766 * 2767 * **-ERANGE** if resulting value was out of range. 2768 * 2769 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2770 * Description 2771 * Convert the initial part of the string from buffer *buf* of 2772 * size *buf_len* to an unsigned long integer according to the 2773 * given base and save the result in *res*. 2774 * 2775 * The string may begin with an arbitrary amount of white space 2776 * (as determined by **isspace**\ (3)). 2777 * 2778 * Five least significant bits of *flags* encode base, other bits 2779 * are currently unused. 2780 * 2781 * Base must be either 8, 10, 16 or 0 to detect it automatically 2782 * similar to user space **strtoul**\ (3). 2783 * Return 2784 * Number of characters consumed on success. Must be positive but 2785 * no more than *buf_len*. 2786 * 2787 * **-EINVAL** if no valid digits were found or unsupported base 2788 * was provided. 2789 * 2790 * **-ERANGE** if resulting value was out of range. 2791 * 2792 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2793 * Description 2794 * Get a bpf-local-storage from a *sk*. 2795 * 2796 * Logically, it could be thought of getting the value from 2797 * a *map* with *sk* as the **key**. From this 2798 * perspective, the usage is not much different from 2799 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2800 * helper enforces the key must be a full socket and the map must 2801 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2802 * 2803 * Underneath, the value is stored locally at *sk* instead of 2804 * the *map*. The *map* is used as the bpf-local-storage 2805 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2806 * searched against all bpf-local-storages residing at *sk*. 2807 * 2808 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2809 * used such that a new bpf-local-storage will be 2810 * created if one does not exist. *value* can be used 2811 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2812 * the initial value of a bpf-local-storage. If *value* is 2813 * **NULL**, the new bpf-local-storage will be zero initialized. 2814 * Return 2815 * A bpf-local-storage pointer is returned on success. 2816 * 2817 * **NULL** if not found or there was an error in adding 2818 * a new bpf-local-storage. 2819 * 2820 * long bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2821 * Description 2822 * Delete a bpf-local-storage from a *sk*. 2823 * Return 2824 * 0 on success. 2825 * 2826 * **-ENOENT** if the bpf-local-storage cannot be found. 2827 * 2828 * long bpf_send_signal(u32 sig) 2829 * Description 2830 * Send signal *sig* to the process of the current task. 2831 * The signal may be delivered to any of this process's threads. 2832 * Return 2833 * 0 on success or successfully queued. 2834 * 2835 * **-EBUSY** if work queue under nmi is full. 2836 * 2837 * **-EINVAL** if *sig* is invalid. 2838 * 2839 * **-EPERM** if no permission to send the *sig*. 2840 * 2841 * **-EAGAIN** if bpf program can try again. 2842 * 2843 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2844 * Description 2845 * Try to issue a SYN cookie for the packet with corresponding 2846 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2847 * 2848 * *iph* points to the start of the IPv4 or IPv6 header, while 2849 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2850 * **sizeof**\ (**struct ip6hdr**). 2851 * 2852 * *th* points to the start of the TCP header, while *th_len* 2853 * contains the length of the TCP header. 2854 * Return 2855 * On success, lower 32 bits hold the generated SYN cookie in 2856 * followed by 16 bits which hold the MSS value for that cookie, 2857 * and the top 16 bits are unused. 2858 * 2859 * On failure, the returned value is one of the following: 2860 * 2861 * **-EINVAL** SYN cookie cannot be issued due to error 2862 * 2863 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2864 * 2865 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2866 * 2867 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2868 * 2869 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2870 * Description 2871 * Write raw *data* blob into a special BPF perf event held by 2872 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2873 * event must have the following attributes: **PERF_SAMPLE_RAW** 2874 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2875 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2876 * 2877 * The *flags* are used to indicate the index in *map* for which 2878 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2879 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2880 * to indicate that the index of the current CPU core should be 2881 * used. 2882 * 2883 * The value to write, of *size*, is passed through eBPF stack and 2884 * pointed by *data*. 2885 * 2886 * *ctx* is a pointer to in-kernel struct sk_buff. 2887 * 2888 * This helper is similar to **bpf_perf_event_output**\ () but 2889 * restricted to raw_tracepoint bpf programs. 2890 * Return 2891 * 0 on success, or a negative error in case of failure. 2892 * 2893 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2894 * Description 2895 * Safely attempt to read *size* bytes from user space address 2896 * *unsafe_ptr* and store the data in *dst*. 2897 * Return 2898 * 0 on success, or a negative error in case of failure. 2899 * 2900 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2901 * Description 2902 * Safely attempt to read *size* bytes from kernel space address 2903 * *unsafe_ptr* and store the data in *dst*. 2904 * Return 2905 * 0 on success, or a negative error in case of failure. 2906 * 2907 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2908 * Description 2909 * Copy a NUL terminated string from an unsafe user address 2910 * *unsafe_ptr* to *dst*. The *size* should include the 2911 * terminating NUL byte. In case the string length is smaller than 2912 * *size*, the target is not padded with further NUL bytes. If the 2913 * string length is larger than *size*, just *size*-1 bytes are 2914 * copied and the last byte is set to NUL. 2915 * 2916 * On success, the length of the copied string is returned. This 2917 * makes this helper useful in tracing programs for reading 2918 * strings, and more importantly to get its length at runtime. See 2919 * the following snippet: 2920 * 2921 * :: 2922 * 2923 * SEC("kprobe/sys_open") 2924 * void bpf_sys_open(struct pt_regs *ctx) 2925 * { 2926 * char buf[PATHLEN]; // PATHLEN is defined to 256 2927 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2928 * ctx->di); 2929 * 2930 * // Consume buf, for example push it to 2931 * // userspace via bpf_perf_event_output(); we 2932 * // can use res (the string length) as event 2933 * // size, after checking its boundaries. 2934 * } 2935 * 2936 * In comparison, using **bpf_probe_read_user**\ () helper here 2937 * instead to read the string would require to estimate the length 2938 * at compile time, and would often result in copying more memory 2939 * than necessary. 2940 * 2941 * Another useful use case is when parsing individual process 2942 * arguments or individual environment variables navigating 2943 * *current*\ **->mm->arg_start** and *current*\ 2944 * **->mm->env_start**: using this helper and the return value, 2945 * one can quickly iterate at the right offset of the memory area. 2946 * Return 2947 * On success, the strictly positive length of the string, 2948 * including the trailing NUL character. On error, a negative 2949 * value. 2950 * 2951 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 2952 * Description 2953 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2954 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 2955 * Return 2956 * On success, the strictly positive length of the string, including 2957 * the trailing NUL character. On error, a negative value. 2958 * 2959 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 2960 * Description 2961 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 2962 * *rcv_nxt* is the ack_seq to be sent out. 2963 * Return 2964 * 0 on success, or a negative error in case of failure. 2965 * 2966 * long bpf_send_signal_thread(u32 sig) 2967 * Description 2968 * Send signal *sig* to the thread corresponding to the current task. 2969 * Return 2970 * 0 on success or successfully queued. 2971 * 2972 * **-EBUSY** if work queue under nmi is full. 2973 * 2974 * **-EINVAL** if *sig* is invalid. 2975 * 2976 * **-EPERM** if no permission to send the *sig*. 2977 * 2978 * **-EAGAIN** if bpf program can try again. 2979 * 2980 * u64 bpf_jiffies64(void) 2981 * Description 2982 * Obtain the 64bit jiffies 2983 * Return 2984 * The 64 bit jiffies 2985 * 2986 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 2987 * Description 2988 * For an eBPF program attached to a perf event, retrieve the 2989 * branch records (**struct perf_branch_entry**) associated to *ctx* 2990 * and store it in the buffer pointed by *buf* up to size 2991 * *size* bytes. 2992 * Return 2993 * On success, number of bytes written to *buf*. On error, a 2994 * negative value. 2995 * 2996 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 2997 * instead return the number of bytes required to store all the 2998 * branch entries. If this flag is set, *buf* may be NULL. 2999 * 3000 * **-EINVAL** if arguments invalid or **size** not a multiple 3001 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3002 * 3003 * **-ENOENT** if architecture does not support branch records. 3004 * 3005 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3006 * Description 3007 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3008 * *namespace* will be returned in *nsdata*. 3009 * Return 3010 * 0 on success, or one of the following in case of failure: 3011 * 3012 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3013 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3014 * 3015 * **-ENOENT** if pidns does not exists for the current task. 3016 * 3017 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3018 * Description 3019 * Write raw *data* blob into a special BPF perf event held by 3020 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3021 * event must have the following attributes: **PERF_SAMPLE_RAW** 3022 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3023 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3024 * 3025 * The *flags* are used to indicate the index in *map* for which 3026 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3027 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3028 * to indicate that the index of the current CPU core should be 3029 * used. 3030 * 3031 * The value to write, of *size*, is passed through eBPF stack and 3032 * pointed by *data*. 3033 * 3034 * *ctx* is a pointer to in-kernel struct xdp_buff. 3035 * 3036 * This helper is similar to **bpf_perf_eventoutput**\ () but 3037 * restricted to raw_tracepoint bpf programs. 3038 * Return 3039 * 0 on success, or a negative error in case of failure. 3040 * 3041 * u64 bpf_get_netns_cookie(void *ctx) 3042 * Description 3043 * Retrieve the cookie (generated by the kernel) of the network 3044 * namespace the input *ctx* is associated with. The network 3045 * namespace cookie remains stable for its lifetime and provides 3046 * a global identifier that can be assumed unique. If *ctx* is 3047 * NULL, then the helper returns the cookie for the initial 3048 * network namespace. The cookie itself is very similar to that 3049 * of **bpf_get_socket_cookie**\ () helper, but for network 3050 * namespaces instead of sockets. 3051 * Return 3052 * A 8-byte long opaque number. 3053 * 3054 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3055 * Description 3056 * Return id of cgroup v2 that is ancestor of the cgroup associated 3057 * with the current task at the *ancestor_level*. The root cgroup 3058 * is at *ancestor_level* zero and each step down the hierarchy 3059 * increments the level. If *ancestor_level* == level of cgroup 3060 * associated with the current task, then return value will be the 3061 * same as that of **bpf_get_current_cgroup_id**\ (). 3062 * 3063 * The helper is useful to implement policies based on cgroups 3064 * that are upper in hierarchy than immediate cgroup associated 3065 * with the current task. 3066 * 3067 * The format of returned id and helper limitations are same as in 3068 * **bpf_get_current_cgroup_id**\ (). 3069 * Return 3070 * The id is returned or 0 in case the id could not be retrieved. 3071 * 3072 * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags) 3073 * Description 3074 * Helper is overloaded depending on BPF program type. This 3075 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3076 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3077 * 3078 * Assign the *sk* to the *skb*. When combined with appropriate 3079 * routing configuration to receive the packet towards the socket, 3080 * will cause *skb* to be delivered to the specified socket. 3081 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3082 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3083 * interfere with successful delivery to the socket. 3084 * 3085 * This operation is only valid from TC ingress path. 3086 * 3087 * The *flags* argument must be zero. 3088 * Return 3089 * 0 on success, or a negative error in case of failure: 3090 * 3091 * **-EINVAL** if specified *flags* are not supported. 3092 * 3093 * **-ENOENT** if the socket is unavailable for assignment. 3094 * 3095 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3096 * 3097 * **-EOPNOTSUPP** if the operation is not supported, for example 3098 * a call from outside of TC ingress. 3099 * 3100 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3101 * (reuseport). 3102 * 3103 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3104 * Description 3105 * Helper is overloaded depending on BPF program type. This 3106 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3107 * 3108 * Select the *sk* as a result of a socket lookup. 3109 * 3110 * For the operation to succeed passed socket must be compatible 3111 * with the packet description provided by the *ctx* object. 3112 * 3113 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3114 * be an exact match. While IP family (**AF_INET** or 3115 * **AF_INET6**) must be compatible, that is IPv6 sockets 3116 * that are not v6-only can be selected for IPv4 packets. 3117 * 3118 * Only TCP listeners and UDP unconnected sockets can be 3119 * selected. *sk* can also be NULL to reset any previous 3120 * selection. 3121 * 3122 * *flags* argument can combination of following values: 3123 * 3124 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3125 * socket selection, potentially done by a BPF program 3126 * that ran before us. 3127 * 3128 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3129 * load-balancing within reuseport group for the socket 3130 * being selected. 3131 * 3132 * On success *ctx->sk* will point to the selected socket. 3133 * 3134 * Return 3135 * 0 on success, or a negative errno in case of failure. 3136 * 3137 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3138 * not compatible with packet family (*ctx->family*). 3139 * 3140 * * **-EEXIST** if socket has been already selected, 3141 * potentially by another program, and 3142 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3143 * 3144 * * **-EINVAL** if unsupported flags were specified. 3145 * 3146 * * **-EPROTOTYPE** if socket L4 protocol 3147 * (*sk->protocol*) doesn't match packet protocol 3148 * (*ctx->protocol*). 3149 * 3150 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3151 * state (TCP listening or UDP unconnected). 3152 * 3153 * u64 bpf_ktime_get_boot_ns(void) 3154 * Description 3155 * Return the time elapsed since system boot, in nanoseconds. 3156 * Does include the time the system was suspended. 3157 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3158 * Return 3159 * Current *ktime*. 3160 * 3161 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3162 * Description 3163 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3164 * out the format string. 3165 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3166 * the format string itself. The *data* and *data_len* are format string 3167 * arguments. The *data* are a **u64** array and corresponding format string 3168 * values are stored in the array. For strings and pointers where pointees 3169 * are accessed, only the pointer values are stored in the *data* array. 3170 * The *data_len* is the size of *data* in bytes. 3171 * 3172 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3173 * Reading kernel memory may fail due to either invalid address or 3174 * valid address but requiring a major memory fault. If reading kernel memory 3175 * fails, the string for **%s** will be an empty string, and the ip 3176 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3177 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3178 * Return 3179 * 0 on success, or a negative error in case of failure: 3180 * 3181 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3182 * by returning 1 from bpf program. 3183 * 3184 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3185 * 3186 * **-E2BIG** if *fmt* contains too many format specifiers. 3187 * 3188 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3189 * 3190 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 3191 * Description 3192 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3193 * The *m* represents the seq_file. The *data* and *len* represent the 3194 * data to write in bytes. 3195 * Return 3196 * 0 on success, or a negative error in case of failure: 3197 * 3198 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3199 * 3200 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk) 3201 * Description 3202 * Return the cgroup v2 id of the socket *sk*. 3203 * 3204 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one 3205 * returned from **bpf_sk_lookup_xxx**\ (), 3206 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3207 * same as in **bpf_skb_cgroup_id**\ (). 3208 * 3209 * This helper is available only if the kernel was compiled with 3210 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3211 * Return 3212 * The id is returned or 0 in case the id could not be retrieved. 3213 * 3214 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level) 3215 * Description 3216 * Return id of cgroup v2 that is ancestor of cgroup associated 3217 * with the *sk* at the *ancestor_level*. The root cgroup is at 3218 * *ancestor_level* zero and each step down the hierarchy 3219 * increments the level. If *ancestor_level* == level of cgroup 3220 * associated with *sk*, then return value will be same as that 3221 * of **bpf_sk_cgroup_id**\ (). 3222 * 3223 * The helper is useful to implement policies based on cgroups 3224 * that are upper in hierarchy than immediate cgroup associated 3225 * with *sk*. 3226 * 3227 * The format of returned id and helper limitations are same as in 3228 * **bpf_sk_cgroup_id**\ (). 3229 * Return 3230 * The id is returned or 0 in case the id could not be retrieved. 3231 * 3232 * int bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 3233 * Description 3234 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3235 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3236 * of new data availability is sent. 3237 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3238 * of new data availability is sent unconditionally. 3239 * Return 3240 * 0 on success, or a negative error in case of failure. 3241 * 3242 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 3243 * Description 3244 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3245 * Return 3246 * Valid pointer with *size* bytes of memory available; NULL, 3247 * otherwise. 3248 * 3249 * void bpf_ringbuf_submit(void *data, u64 flags) 3250 * Description 3251 * Submit reserved ring buffer sample, pointed to by *data*. 3252 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3253 * of new data availability is sent. 3254 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3255 * of new data availability is sent unconditionally. 3256 * Return 3257 * Nothing. Always succeeds. 3258 * 3259 * void bpf_ringbuf_discard(void *data, u64 flags) 3260 * Description 3261 * Discard reserved ring buffer sample, pointed to by *data*. 3262 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3263 * of new data availability is sent. 3264 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3265 * of new data availability is sent unconditionally. 3266 * Return 3267 * Nothing. Always succeeds. 3268 * 3269 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 3270 * Description 3271 * Query various characteristics of provided ring buffer. What 3272 * exactly is queries is determined by *flags*: 3273 * 3274 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3275 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3276 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3277 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3278 * 3279 * Data returned is just a momentary snapshot of actual values 3280 * and could be inaccurate, so this facility should be used to 3281 * power heuristics and for reporting, not to make 100% correct 3282 * calculation. 3283 * Return 3284 * Requested value, or 0, if *flags* are not recognized. 3285 * 3286 * long bpf_csum_level(struct sk_buff *skb, u64 level) 3287 * Description 3288 * Change the skbs checksum level by one layer up or down, or 3289 * reset it entirely to none in order to have the stack perform 3290 * checksum validation. The level is applicable to the following 3291 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3292 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3293 * through **bpf_skb_adjust_room**\ () helper with passing in 3294 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3295 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3296 * the UDP header is removed. Similarly, an encap of the latter 3297 * into the former could be accompanied by a helper call to 3298 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3299 * skb is still intended to be processed in higher layers of the 3300 * stack instead of just egressing at tc. 3301 * 3302 * There are three supported level settings at this time: 3303 * 3304 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3305 * with CHECKSUM_UNNECESSARY. 3306 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3307 * with CHECKSUM_UNNECESSARY. 3308 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3309 * sets CHECKSUM_NONE to force checksum validation by the stack. 3310 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3311 * skb->csum_level. 3312 * Return 3313 * 0 on success, or a negative error in case of failure. In the 3314 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3315 * is returned or the error code -EACCES in case the skb is not 3316 * subject to CHECKSUM_UNNECESSARY. 3317 * 3318 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 3319 * Description 3320 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3321 * Return 3322 * *sk* if casting is valid, or NULL otherwise. 3323 * 3324 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 3325 * Description 3326 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3327 * Return 3328 * *sk* if casting is valid, or NULL otherwise. 3329 * 3330 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 3331 * Description 3332 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3333 * Return 3334 * *sk* if casting is valid, or NULL otherwise. 3335 * 3336 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 3337 * Description 3338 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3339 * Return 3340 * *sk* if casting is valid, or NULL otherwise. 3341 * 3342 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 3343 * Description 3344 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3345 * Return 3346 * *sk* if casting is valid, or NULL otherwise. 3347 * 3348 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 3349 * Description 3350 * Return a user or a kernel stack in bpf program provided buffer. 3351 * To achieve this, the helper needs *task*, which is a valid 3352 * pointer to struct task_struct. To store the stacktrace, the 3353 * bpf program provides *buf* with a nonnegative *size*. 3354 * 3355 * The last argument, *flags*, holds the number of stack frames to 3356 * skip (from 0 to 255), masked with 3357 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3358 * the following flags: 3359 * 3360 * **BPF_F_USER_STACK** 3361 * Collect a user space stack instead of a kernel stack. 3362 * **BPF_F_USER_BUILD_ID** 3363 * Collect buildid+offset instead of ips for user stack, 3364 * only valid if **BPF_F_USER_STACK** is also specified. 3365 * 3366 * **bpf_get_task_stack**\ () can collect up to 3367 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3368 * to sufficient large buffer size. Note that 3369 * this limit can be controlled with the **sysctl** program, and 3370 * that it should be manually increased in order to profile long 3371 * user stacks (such as stacks for Java programs). To do so, use: 3372 * 3373 * :: 3374 * 3375 * # sysctl kernel.perf_event_max_stack=<new value> 3376 * Return 3377 * A non-negative value equal to or less than *size* on success, 3378 * or a negative error in case of failure. 3379 * 3380 */ 3381 #define __BPF_FUNC_MAPPER(FN) \ 3382 FN(unspec), \ 3383 FN(map_lookup_elem), \ 3384 FN(map_update_elem), \ 3385 FN(map_delete_elem), \ 3386 FN(probe_read), \ 3387 FN(ktime_get_ns), \ 3388 FN(trace_printk), \ 3389 FN(get_prandom_u32), \ 3390 FN(get_smp_processor_id), \ 3391 FN(skb_store_bytes), \ 3392 FN(l3_csum_replace), \ 3393 FN(l4_csum_replace), \ 3394 FN(tail_call), \ 3395 FN(clone_redirect), \ 3396 FN(get_current_pid_tgid), \ 3397 FN(get_current_uid_gid), \ 3398 FN(get_current_comm), \ 3399 FN(get_cgroup_classid), \ 3400 FN(skb_vlan_push), \ 3401 FN(skb_vlan_pop), \ 3402 FN(skb_get_tunnel_key), \ 3403 FN(skb_set_tunnel_key), \ 3404 FN(perf_event_read), \ 3405 FN(redirect), \ 3406 FN(get_route_realm), \ 3407 FN(perf_event_output), \ 3408 FN(skb_load_bytes), \ 3409 FN(get_stackid), \ 3410 FN(csum_diff), \ 3411 FN(skb_get_tunnel_opt), \ 3412 FN(skb_set_tunnel_opt), \ 3413 FN(skb_change_proto), \ 3414 FN(skb_change_type), \ 3415 FN(skb_under_cgroup), \ 3416 FN(get_hash_recalc), \ 3417 FN(get_current_task), \ 3418 FN(probe_write_user), \ 3419 FN(current_task_under_cgroup), \ 3420 FN(skb_change_tail), \ 3421 FN(skb_pull_data), \ 3422 FN(csum_update), \ 3423 FN(set_hash_invalid), \ 3424 FN(get_numa_node_id), \ 3425 FN(skb_change_head), \ 3426 FN(xdp_adjust_head), \ 3427 FN(probe_read_str), \ 3428 FN(get_socket_cookie), \ 3429 FN(get_socket_uid), \ 3430 FN(set_hash), \ 3431 FN(setsockopt), \ 3432 FN(skb_adjust_room), \ 3433 FN(redirect_map), \ 3434 FN(sk_redirect_map), \ 3435 FN(sock_map_update), \ 3436 FN(xdp_adjust_meta), \ 3437 FN(perf_event_read_value), \ 3438 FN(perf_prog_read_value), \ 3439 FN(getsockopt), \ 3440 FN(override_return), \ 3441 FN(sock_ops_cb_flags_set), \ 3442 FN(msg_redirect_map), \ 3443 FN(msg_apply_bytes), \ 3444 FN(msg_cork_bytes), \ 3445 FN(msg_pull_data), \ 3446 FN(bind), \ 3447 FN(xdp_adjust_tail), \ 3448 FN(skb_get_xfrm_state), \ 3449 FN(get_stack), \ 3450 FN(skb_load_bytes_relative), \ 3451 FN(fib_lookup), \ 3452 FN(sock_hash_update), \ 3453 FN(msg_redirect_hash), \ 3454 FN(sk_redirect_hash), \ 3455 FN(lwt_push_encap), \ 3456 FN(lwt_seg6_store_bytes), \ 3457 FN(lwt_seg6_adjust_srh), \ 3458 FN(lwt_seg6_action), \ 3459 FN(rc_repeat), \ 3460 FN(rc_keydown), \ 3461 FN(skb_cgroup_id), \ 3462 FN(get_current_cgroup_id), \ 3463 FN(get_local_storage), \ 3464 FN(sk_select_reuseport), \ 3465 FN(skb_ancestor_cgroup_id), \ 3466 FN(sk_lookup_tcp), \ 3467 FN(sk_lookup_udp), \ 3468 FN(sk_release), \ 3469 FN(map_push_elem), \ 3470 FN(map_pop_elem), \ 3471 FN(map_peek_elem), \ 3472 FN(msg_push_data), \ 3473 FN(msg_pop_data), \ 3474 FN(rc_pointer_rel), \ 3475 FN(spin_lock), \ 3476 FN(spin_unlock), \ 3477 FN(sk_fullsock), \ 3478 FN(tcp_sock), \ 3479 FN(skb_ecn_set_ce), \ 3480 FN(get_listener_sock), \ 3481 FN(skc_lookup_tcp), \ 3482 FN(tcp_check_syncookie), \ 3483 FN(sysctl_get_name), \ 3484 FN(sysctl_get_current_value), \ 3485 FN(sysctl_get_new_value), \ 3486 FN(sysctl_set_new_value), \ 3487 FN(strtol), \ 3488 FN(strtoul), \ 3489 FN(sk_storage_get), \ 3490 FN(sk_storage_delete), \ 3491 FN(send_signal), \ 3492 FN(tcp_gen_syncookie), \ 3493 FN(skb_output), \ 3494 FN(probe_read_user), \ 3495 FN(probe_read_kernel), \ 3496 FN(probe_read_user_str), \ 3497 FN(probe_read_kernel_str), \ 3498 FN(tcp_send_ack), \ 3499 FN(send_signal_thread), \ 3500 FN(jiffies64), \ 3501 FN(read_branch_records), \ 3502 FN(get_ns_current_pid_tgid), \ 3503 FN(xdp_output), \ 3504 FN(get_netns_cookie), \ 3505 FN(get_current_ancestor_cgroup_id), \ 3506 FN(sk_assign), \ 3507 FN(ktime_get_boot_ns), \ 3508 FN(seq_printf), \ 3509 FN(seq_write), \ 3510 FN(sk_cgroup_id), \ 3511 FN(sk_ancestor_cgroup_id), \ 3512 FN(ringbuf_output), \ 3513 FN(ringbuf_reserve), \ 3514 FN(ringbuf_submit), \ 3515 FN(ringbuf_discard), \ 3516 FN(ringbuf_query), \ 3517 FN(csum_level), \ 3518 FN(skc_to_tcp6_sock), \ 3519 FN(skc_to_tcp_sock), \ 3520 FN(skc_to_tcp_timewait_sock), \ 3521 FN(skc_to_tcp_request_sock), \ 3522 FN(skc_to_udp6_sock), \ 3523 FN(get_task_stack), \ 3524 /* */ 3525 3526 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3527 * function eBPF program intends to call 3528 */ 3529 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3530 enum bpf_func_id { 3531 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3532 __BPF_FUNC_MAX_ID, 3533 }; 3534 #undef __BPF_ENUM_FN 3535 3536 /* All flags used by eBPF helper functions, placed here. */ 3537 3538 /* BPF_FUNC_skb_store_bytes flags. */ 3539 enum { 3540 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 3541 BPF_F_INVALIDATE_HASH = (1ULL << 1), 3542 }; 3543 3544 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3545 * First 4 bits are for passing the header field size. 3546 */ 3547 enum { 3548 BPF_F_HDR_FIELD_MASK = 0xfULL, 3549 }; 3550 3551 /* BPF_FUNC_l4_csum_replace flags. */ 3552 enum { 3553 BPF_F_PSEUDO_HDR = (1ULL << 4), 3554 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 3555 BPF_F_MARK_ENFORCE = (1ULL << 6), 3556 }; 3557 3558 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3559 enum { 3560 BPF_F_INGRESS = (1ULL << 0), 3561 }; 3562 3563 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3564 enum { 3565 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 3566 }; 3567 3568 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3569 enum { 3570 BPF_F_SKIP_FIELD_MASK = 0xffULL, 3571 BPF_F_USER_STACK = (1ULL << 8), 3572 /* flags used by BPF_FUNC_get_stackid only. */ 3573 BPF_F_FAST_STACK_CMP = (1ULL << 9), 3574 BPF_F_REUSE_STACKID = (1ULL << 10), 3575 /* flags used by BPF_FUNC_get_stack only. */ 3576 BPF_F_USER_BUILD_ID = (1ULL << 11), 3577 }; 3578 3579 /* BPF_FUNC_skb_set_tunnel_key flags. */ 3580 enum { 3581 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 3582 BPF_F_DONT_FRAGMENT = (1ULL << 2), 3583 BPF_F_SEQ_NUMBER = (1ULL << 3), 3584 }; 3585 3586 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3587 * BPF_FUNC_perf_event_read_value flags. 3588 */ 3589 enum { 3590 BPF_F_INDEX_MASK = 0xffffffffULL, 3591 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 3592 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 3593 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 3594 }; 3595 3596 /* Current network namespace */ 3597 enum { 3598 BPF_F_CURRENT_NETNS = (-1L), 3599 }; 3600 3601 /* BPF_FUNC_csum_level level values. */ 3602 enum { 3603 BPF_CSUM_LEVEL_QUERY, 3604 BPF_CSUM_LEVEL_INC, 3605 BPF_CSUM_LEVEL_DEC, 3606 BPF_CSUM_LEVEL_RESET, 3607 }; 3608 3609 /* BPF_FUNC_skb_adjust_room flags. */ 3610 enum { 3611 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 3612 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 3613 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 3614 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 3615 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 3616 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 3617 }; 3618 3619 enum { 3620 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 3621 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 3622 }; 3623 3624 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3625 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3626 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3627 3628 /* BPF_FUNC_sysctl_get_name flags. */ 3629 enum { 3630 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 3631 }; 3632 3633 /* BPF_FUNC_sk_storage_get flags */ 3634 enum { 3635 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0), 3636 }; 3637 3638 /* BPF_FUNC_read_branch_records flags. */ 3639 enum { 3640 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 3641 }; 3642 3643 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 3644 * BPF_FUNC_bpf_ringbuf_output flags. 3645 */ 3646 enum { 3647 BPF_RB_NO_WAKEUP = (1ULL << 0), 3648 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 3649 }; 3650 3651 /* BPF_FUNC_bpf_ringbuf_query flags */ 3652 enum { 3653 BPF_RB_AVAIL_DATA = 0, 3654 BPF_RB_RING_SIZE = 1, 3655 BPF_RB_CONS_POS = 2, 3656 BPF_RB_PROD_POS = 3, 3657 }; 3658 3659 /* BPF ring buffer constants */ 3660 enum { 3661 BPF_RINGBUF_BUSY_BIT = (1U << 31), 3662 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 3663 BPF_RINGBUF_HDR_SZ = 8, 3664 }; 3665 3666 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 3667 enum { 3668 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 3669 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 3670 }; 3671 3672 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 3673 enum bpf_adj_room_mode { 3674 BPF_ADJ_ROOM_NET, 3675 BPF_ADJ_ROOM_MAC, 3676 }; 3677 3678 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3679 enum bpf_hdr_start_off { 3680 BPF_HDR_START_MAC, 3681 BPF_HDR_START_NET, 3682 }; 3683 3684 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3685 enum bpf_lwt_encap_mode { 3686 BPF_LWT_ENCAP_SEG6, 3687 BPF_LWT_ENCAP_SEG6_INLINE, 3688 BPF_LWT_ENCAP_IP, 3689 }; 3690 3691 #define __bpf_md_ptr(type, name) \ 3692 union { \ 3693 type name; \ 3694 __u64 :64; \ 3695 } __attribute__((aligned(8))) 3696 3697 /* user accessible mirror of in-kernel sk_buff. 3698 * new fields can only be added to the end of this structure 3699 */ 3700 struct __sk_buff { 3701 __u32 len; 3702 __u32 pkt_type; 3703 __u32 mark; 3704 __u32 queue_mapping; 3705 __u32 protocol; 3706 __u32 vlan_present; 3707 __u32 vlan_tci; 3708 __u32 vlan_proto; 3709 __u32 priority; 3710 __u32 ingress_ifindex; 3711 __u32 ifindex; 3712 __u32 tc_index; 3713 __u32 cb[5]; 3714 __u32 hash; 3715 __u32 tc_classid; 3716 __u32 data; 3717 __u32 data_end; 3718 __u32 napi_id; 3719 3720 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3721 __u32 family; 3722 __u32 remote_ip4; /* Stored in network byte order */ 3723 __u32 local_ip4; /* Stored in network byte order */ 3724 __u32 remote_ip6[4]; /* Stored in network byte order */ 3725 __u32 local_ip6[4]; /* Stored in network byte order */ 3726 __u32 remote_port; /* Stored in network byte order */ 3727 __u32 local_port; /* stored in host byte order */ 3728 /* ... here. */ 3729 3730 __u32 data_meta; 3731 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3732 __u64 tstamp; 3733 __u32 wire_len; 3734 __u32 gso_segs; 3735 __bpf_md_ptr(struct bpf_sock *, sk); 3736 __u32 gso_size; 3737 }; 3738 3739 struct bpf_tunnel_key { 3740 __u32 tunnel_id; 3741 union { 3742 __u32 remote_ipv4; 3743 __u32 remote_ipv6[4]; 3744 }; 3745 __u8 tunnel_tos; 3746 __u8 tunnel_ttl; 3747 __u16 tunnel_ext; /* Padding, future use. */ 3748 __u32 tunnel_label; 3749 }; 3750 3751 /* user accessible mirror of in-kernel xfrm_state. 3752 * new fields can only be added to the end of this structure 3753 */ 3754 struct bpf_xfrm_state { 3755 __u32 reqid; 3756 __u32 spi; /* Stored in network byte order */ 3757 __u16 family; 3758 __u16 ext; /* Padding, future use. */ 3759 union { 3760 __u32 remote_ipv4; /* Stored in network byte order */ 3761 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3762 }; 3763 }; 3764 3765 /* Generic BPF return codes which all BPF program types may support. 3766 * The values are binary compatible with their TC_ACT_* counter-part to 3767 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3768 * programs. 3769 * 3770 * XDP is handled seprately, see XDP_*. 3771 */ 3772 enum bpf_ret_code { 3773 BPF_OK = 0, 3774 /* 1 reserved */ 3775 BPF_DROP = 2, 3776 /* 3-6 reserved */ 3777 BPF_REDIRECT = 7, 3778 /* >127 are reserved for prog type specific return codes. 3779 * 3780 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3781 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3782 * changed and should be routed based on its new L3 header. 3783 * (This is an L3 redirect, as opposed to L2 redirect 3784 * represented by BPF_REDIRECT above). 3785 */ 3786 BPF_LWT_REROUTE = 128, 3787 }; 3788 3789 struct bpf_sock { 3790 __u32 bound_dev_if; 3791 __u32 family; 3792 __u32 type; 3793 __u32 protocol; 3794 __u32 mark; 3795 __u32 priority; 3796 /* IP address also allows 1 and 2 bytes access */ 3797 __u32 src_ip4; 3798 __u32 src_ip6[4]; 3799 __u32 src_port; /* host byte order */ 3800 __u32 dst_port; /* network byte order */ 3801 __u32 dst_ip4; 3802 __u32 dst_ip6[4]; 3803 __u32 state; 3804 __s32 rx_queue_mapping; 3805 }; 3806 3807 struct bpf_tcp_sock { 3808 __u32 snd_cwnd; /* Sending congestion window */ 3809 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3810 __u32 rtt_min; 3811 __u32 snd_ssthresh; /* Slow start size threshold */ 3812 __u32 rcv_nxt; /* What we want to receive next */ 3813 __u32 snd_nxt; /* Next sequence we send */ 3814 __u32 snd_una; /* First byte we want an ack for */ 3815 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3816 __u32 ecn_flags; /* ECN status bits. */ 3817 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3818 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3819 __u32 packets_out; /* Packets which are "in flight" */ 3820 __u32 retrans_out; /* Retransmitted packets out */ 3821 __u32 total_retrans; /* Total retransmits for entire connection */ 3822 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3823 * total number of segments in. 3824 */ 3825 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3826 * total number of data segments in. 3827 */ 3828 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3829 * The total number of segments sent. 3830 */ 3831 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3832 * total number of data segments sent. 3833 */ 3834 __u32 lost_out; /* Lost packets */ 3835 __u32 sacked_out; /* SACK'd packets */ 3836 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3837 * sum(delta(rcv_nxt)), or how many bytes 3838 * were acked. 3839 */ 3840 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3841 * sum(delta(snd_una)), or how many bytes 3842 * were acked. 3843 */ 3844 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3845 * total number of DSACK blocks received 3846 */ 3847 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3848 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3849 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3850 }; 3851 3852 struct bpf_sock_tuple { 3853 union { 3854 struct { 3855 __be32 saddr; 3856 __be32 daddr; 3857 __be16 sport; 3858 __be16 dport; 3859 } ipv4; 3860 struct { 3861 __be32 saddr[4]; 3862 __be32 daddr[4]; 3863 __be16 sport; 3864 __be16 dport; 3865 } ipv6; 3866 }; 3867 }; 3868 3869 struct bpf_xdp_sock { 3870 __u32 queue_id; 3871 }; 3872 3873 #define XDP_PACKET_HEADROOM 256 3874 3875 /* User return codes for XDP prog type. 3876 * A valid XDP program must return one of these defined values. All other 3877 * return codes are reserved for future use. Unknown return codes will 3878 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3879 */ 3880 enum xdp_action { 3881 XDP_ABORTED = 0, 3882 XDP_DROP, 3883 XDP_PASS, 3884 XDP_TX, 3885 XDP_REDIRECT, 3886 }; 3887 3888 /* user accessible metadata for XDP packet hook 3889 * new fields must be added to the end of this structure 3890 */ 3891 struct xdp_md { 3892 __u32 data; 3893 __u32 data_end; 3894 __u32 data_meta; 3895 /* Below access go through struct xdp_rxq_info */ 3896 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3897 __u32 rx_queue_index; /* rxq->queue_index */ 3898 3899 __u32 egress_ifindex; /* txq->dev->ifindex */ 3900 }; 3901 3902 /* DEVMAP map-value layout 3903 * 3904 * The struct data-layout of map-value is a configuration interface. 3905 * New members can only be added to the end of this structure. 3906 */ 3907 struct bpf_devmap_val { 3908 __u32 ifindex; /* device index */ 3909 union { 3910 int fd; /* prog fd on map write */ 3911 __u32 id; /* prog id on map read */ 3912 } bpf_prog; 3913 }; 3914 3915 /* CPUMAP map-value layout 3916 * 3917 * The struct data-layout of map-value is a configuration interface. 3918 * New members can only be added to the end of this structure. 3919 */ 3920 struct bpf_cpumap_val { 3921 __u32 qsize; /* queue size to remote target CPU */ 3922 union { 3923 int fd; /* prog fd on map write */ 3924 __u32 id; /* prog id on map read */ 3925 } bpf_prog; 3926 }; 3927 3928 enum sk_action { 3929 SK_DROP = 0, 3930 SK_PASS, 3931 }; 3932 3933 /* user accessible metadata for SK_MSG packet hook, new fields must 3934 * be added to the end of this structure 3935 */ 3936 struct sk_msg_md { 3937 __bpf_md_ptr(void *, data); 3938 __bpf_md_ptr(void *, data_end); 3939 3940 __u32 family; 3941 __u32 remote_ip4; /* Stored in network byte order */ 3942 __u32 local_ip4; /* Stored in network byte order */ 3943 __u32 remote_ip6[4]; /* Stored in network byte order */ 3944 __u32 local_ip6[4]; /* Stored in network byte order */ 3945 __u32 remote_port; /* Stored in network byte order */ 3946 __u32 local_port; /* stored in host byte order */ 3947 __u32 size; /* Total size of sk_msg */ 3948 3949 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 3950 }; 3951 3952 struct sk_reuseport_md { 3953 /* 3954 * Start of directly accessible data. It begins from 3955 * the tcp/udp header. 3956 */ 3957 __bpf_md_ptr(void *, data); 3958 /* End of directly accessible data */ 3959 __bpf_md_ptr(void *, data_end); 3960 /* 3961 * Total length of packet (starting from the tcp/udp header). 3962 * Note that the directly accessible bytes (data_end - data) 3963 * could be less than this "len". Those bytes could be 3964 * indirectly read by a helper "bpf_skb_load_bytes()". 3965 */ 3966 __u32 len; 3967 /* 3968 * Eth protocol in the mac header (network byte order). e.g. 3969 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3970 */ 3971 __u32 eth_protocol; 3972 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3973 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3974 __u32 hash; /* A hash of the packet 4 tuples */ 3975 }; 3976 3977 #define BPF_TAG_SIZE 8 3978 3979 struct bpf_prog_info { 3980 __u32 type; 3981 __u32 id; 3982 __u8 tag[BPF_TAG_SIZE]; 3983 __u32 jited_prog_len; 3984 __u32 xlated_prog_len; 3985 __aligned_u64 jited_prog_insns; 3986 __aligned_u64 xlated_prog_insns; 3987 __u64 load_time; /* ns since boottime */ 3988 __u32 created_by_uid; 3989 __u32 nr_map_ids; 3990 __aligned_u64 map_ids; 3991 char name[BPF_OBJ_NAME_LEN]; 3992 __u32 ifindex; 3993 __u32 gpl_compatible:1; 3994 __u32 :31; /* alignment pad */ 3995 __u64 netns_dev; 3996 __u64 netns_ino; 3997 __u32 nr_jited_ksyms; 3998 __u32 nr_jited_func_lens; 3999 __aligned_u64 jited_ksyms; 4000 __aligned_u64 jited_func_lens; 4001 __u32 btf_id; 4002 __u32 func_info_rec_size; 4003 __aligned_u64 func_info; 4004 __u32 nr_func_info; 4005 __u32 nr_line_info; 4006 __aligned_u64 line_info; 4007 __aligned_u64 jited_line_info; 4008 __u32 nr_jited_line_info; 4009 __u32 line_info_rec_size; 4010 __u32 jited_line_info_rec_size; 4011 __u32 nr_prog_tags; 4012 __aligned_u64 prog_tags; 4013 __u64 run_time_ns; 4014 __u64 run_cnt; 4015 } __attribute__((aligned(8))); 4016 4017 struct bpf_map_info { 4018 __u32 type; 4019 __u32 id; 4020 __u32 key_size; 4021 __u32 value_size; 4022 __u32 max_entries; 4023 __u32 map_flags; 4024 char name[BPF_OBJ_NAME_LEN]; 4025 __u32 ifindex; 4026 __u32 btf_vmlinux_value_type_id; 4027 __u64 netns_dev; 4028 __u64 netns_ino; 4029 __u32 btf_id; 4030 __u32 btf_key_type_id; 4031 __u32 btf_value_type_id; 4032 } __attribute__((aligned(8))); 4033 4034 struct bpf_btf_info { 4035 __aligned_u64 btf; 4036 __u32 btf_size; 4037 __u32 id; 4038 } __attribute__((aligned(8))); 4039 4040 struct bpf_link_info { 4041 __u32 type; 4042 __u32 id; 4043 __u32 prog_id; 4044 union { 4045 struct { 4046 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 4047 __u32 tp_name_len; /* in/out: tp_name buffer len */ 4048 } raw_tracepoint; 4049 struct { 4050 __u32 attach_type; 4051 } tracing; 4052 struct { 4053 __u64 cgroup_id; 4054 __u32 attach_type; 4055 } cgroup; 4056 struct { 4057 __u32 netns_ino; 4058 __u32 attach_type; 4059 } netns; 4060 }; 4061 } __attribute__((aligned(8))); 4062 4063 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 4064 * by user and intended to be used by socket (e.g. to bind to, depends on 4065 * attach type). 4066 */ 4067 struct bpf_sock_addr { 4068 __u32 user_family; /* Allows 4-byte read, but no write. */ 4069 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4070 * Stored in network byte order. 4071 */ 4072 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4073 * Stored in network byte order. 4074 */ 4075 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 4076 * Stored in network byte order 4077 */ 4078 __u32 family; /* Allows 4-byte read, but no write */ 4079 __u32 type; /* Allows 4-byte read, but no write */ 4080 __u32 protocol; /* Allows 4-byte read, but no write */ 4081 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4082 * Stored in network byte order. 4083 */ 4084 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4085 * Stored in network byte order. 4086 */ 4087 __bpf_md_ptr(struct bpf_sock *, sk); 4088 }; 4089 4090 /* User bpf_sock_ops struct to access socket values and specify request ops 4091 * and their replies. 4092 * Some of this fields are in network (bigendian) byte order and may need 4093 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 4094 * New fields can only be added at the end of this structure 4095 */ 4096 struct bpf_sock_ops { 4097 __u32 op; 4098 union { 4099 __u32 args[4]; /* Optionally passed to bpf program */ 4100 __u32 reply; /* Returned by bpf program */ 4101 __u32 replylong[4]; /* Optionally returned by bpf prog */ 4102 }; 4103 __u32 family; 4104 __u32 remote_ip4; /* Stored in network byte order */ 4105 __u32 local_ip4; /* Stored in network byte order */ 4106 __u32 remote_ip6[4]; /* Stored in network byte order */ 4107 __u32 local_ip6[4]; /* Stored in network byte order */ 4108 __u32 remote_port; /* Stored in network byte order */ 4109 __u32 local_port; /* stored in host byte order */ 4110 __u32 is_fullsock; /* Some TCP fields are only valid if 4111 * there is a full socket. If not, the 4112 * fields read as zero. 4113 */ 4114 __u32 snd_cwnd; 4115 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 4116 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 4117 __u32 state; 4118 __u32 rtt_min; 4119 __u32 snd_ssthresh; 4120 __u32 rcv_nxt; 4121 __u32 snd_nxt; 4122 __u32 snd_una; 4123 __u32 mss_cache; 4124 __u32 ecn_flags; 4125 __u32 rate_delivered; 4126 __u32 rate_interval_us; 4127 __u32 packets_out; 4128 __u32 retrans_out; 4129 __u32 total_retrans; 4130 __u32 segs_in; 4131 __u32 data_segs_in; 4132 __u32 segs_out; 4133 __u32 data_segs_out; 4134 __u32 lost_out; 4135 __u32 sacked_out; 4136 __u32 sk_txhash; 4137 __u64 bytes_received; 4138 __u64 bytes_acked; 4139 __bpf_md_ptr(struct bpf_sock *, sk); 4140 }; 4141 4142 /* Definitions for bpf_sock_ops_cb_flags */ 4143 enum { 4144 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 4145 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 4146 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 4147 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 4148 /* Mask of all currently supported cb flags */ 4149 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF, 4150 }; 4151 4152 /* List of known BPF sock_ops operators. 4153 * New entries can only be added at the end 4154 */ 4155 enum { 4156 BPF_SOCK_OPS_VOID, 4157 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 4158 * -1 if default value should be used 4159 */ 4160 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 4161 * window (in packets) or -1 if default 4162 * value should be used 4163 */ 4164 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 4165 * active connection is initialized 4166 */ 4167 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 4168 * active connection is 4169 * established 4170 */ 4171 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 4172 * passive connection is 4173 * established 4174 */ 4175 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 4176 * needs ECN 4177 */ 4178 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 4179 * based on the path and may be 4180 * dependent on the congestion control 4181 * algorithm. In general it indicates 4182 * a congestion threshold. RTTs above 4183 * this indicate congestion 4184 */ 4185 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 4186 * Arg1: value of icsk_retransmits 4187 * Arg2: value of icsk_rto 4188 * Arg3: whether RTO has expired 4189 */ 4190 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 4191 * Arg1: sequence number of 1st byte 4192 * Arg2: # segments 4193 * Arg3: return value of 4194 * tcp_transmit_skb (0 => success) 4195 */ 4196 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 4197 * Arg1: old_state 4198 * Arg2: new_state 4199 */ 4200 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 4201 * socket transition to LISTEN state. 4202 */ 4203 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 4204 */ 4205 }; 4206 4207 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 4208 * changes between the TCP and BPF versions. Ideally this should never happen. 4209 * If it does, we need to add code to convert them before calling 4210 * the BPF sock_ops function. 4211 */ 4212 enum { 4213 BPF_TCP_ESTABLISHED = 1, 4214 BPF_TCP_SYN_SENT, 4215 BPF_TCP_SYN_RECV, 4216 BPF_TCP_FIN_WAIT1, 4217 BPF_TCP_FIN_WAIT2, 4218 BPF_TCP_TIME_WAIT, 4219 BPF_TCP_CLOSE, 4220 BPF_TCP_CLOSE_WAIT, 4221 BPF_TCP_LAST_ACK, 4222 BPF_TCP_LISTEN, 4223 BPF_TCP_CLOSING, /* Now a valid state */ 4224 BPF_TCP_NEW_SYN_RECV, 4225 4226 BPF_TCP_MAX_STATES /* Leave at the end! */ 4227 }; 4228 4229 enum { 4230 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 4231 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 4232 }; 4233 4234 struct bpf_perf_event_value { 4235 __u64 counter; 4236 __u64 enabled; 4237 __u64 running; 4238 }; 4239 4240 enum { 4241 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 4242 BPF_DEVCG_ACC_READ = (1ULL << 1), 4243 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 4244 }; 4245 4246 enum { 4247 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 4248 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 4249 }; 4250 4251 struct bpf_cgroup_dev_ctx { 4252 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 4253 __u32 access_type; 4254 __u32 major; 4255 __u32 minor; 4256 }; 4257 4258 struct bpf_raw_tracepoint_args { 4259 __u64 args[0]; 4260 }; 4261 4262 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 4263 * OUTPUT: Do lookup from egress perspective; default is ingress 4264 */ 4265 enum { 4266 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 4267 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 4268 }; 4269 4270 enum { 4271 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 4272 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 4273 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 4274 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 4275 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 4276 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 4277 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 4278 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 4279 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 4280 }; 4281 4282 struct bpf_fib_lookup { 4283 /* input: network family for lookup (AF_INET, AF_INET6) 4284 * output: network family of egress nexthop 4285 */ 4286 __u8 family; 4287 4288 /* set if lookup is to consider L4 data - e.g., FIB rules */ 4289 __u8 l4_protocol; 4290 __be16 sport; 4291 __be16 dport; 4292 4293 /* total length of packet from network header - used for MTU check */ 4294 __u16 tot_len; 4295 4296 /* input: L3 device index for lookup 4297 * output: device index from FIB lookup 4298 */ 4299 __u32 ifindex; 4300 4301 union { 4302 /* inputs to lookup */ 4303 __u8 tos; /* AF_INET */ 4304 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 4305 4306 /* output: metric of fib result (IPv4/IPv6 only) */ 4307 __u32 rt_metric; 4308 }; 4309 4310 union { 4311 __be32 ipv4_src; 4312 __u32 ipv6_src[4]; /* in6_addr; network order */ 4313 }; 4314 4315 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 4316 * network header. output: bpf_fib_lookup sets to gateway address 4317 * if FIB lookup returns gateway route 4318 */ 4319 union { 4320 __be32 ipv4_dst; 4321 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4322 }; 4323 4324 /* output */ 4325 __be16 h_vlan_proto; 4326 __be16 h_vlan_TCI; 4327 __u8 smac[6]; /* ETH_ALEN */ 4328 __u8 dmac[6]; /* ETH_ALEN */ 4329 }; 4330 4331 enum bpf_task_fd_type { 4332 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 4333 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 4334 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 4335 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 4336 BPF_FD_TYPE_UPROBE, /* filename + offset */ 4337 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 4338 }; 4339 4340 enum { 4341 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 4342 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 4343 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 4344 }; 4345 4346 struct bpf_flow_keys { 4347 __u16 nhoff; 4348 __u16 thoff; 4349 __u16 addr_proto; /* ETH_P_* of valid addrs */ 4350 __u8 is_frag; 4351 __u8 is_first_frag; 4352 __u8 is_encap; 4353 __u8 ip_proto; 4354 __be16 n_proto; 4355 __be16 sport; 4356 __be16 dport; 4357 union { 4358 struct { 4359 __be32 ipv4_src; 4360 __be32 ipv4_dst; 4361 }; 4362 struct { 4363 __u32 ipv6_src[4]; /* in6_addr; network order */ 4364 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4365 }; 4366 }; 4367 __u32 flags; 4368 __be32 flow_label; 4369 }; 4370 4371 struct bpf_func_info { 4372 __u32 insn_off; 4373 __u32 type_id; 4374 }; 4375 4376 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 4377 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 4378 4379 struct bpf_line_info { 4380 __u32 insn_off; 4381 __u32 file_name_off; 4382 __u32 line_off; 4383 __u32 line_col; 4384 }; 4385 4386 struct bpf_spin_lock { 4387 __u32 val; 4388 }; 4389 4390 struct bpf_sysctl { 4391 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 4392 * Allows 1,2,4-byte read, but no write. 4393 */ 4394 __u32 file_pos; /* Sysctl file position to read from, write to. 4395 * Allows 1,2,4-byte read an 4-byte write. 4396 */ 4397 }; 4398 4399 struct bpf_sockopt { 4400 __bpf_md_ptr(struct bpf_sock *, sk); 4401 __bpf_md_ptr(void *, optval); 4402 __bpf_md_ptr(void *, optval_end); 4403 4404 __s32 level; 4405 __s32 optname; 4406 __s32 optlen; 4407 __s32 retval; 4408 }; 4409 4410 struct bpf_pidns_info { 4411 __u32 pid; 4412 __u32 tgid; 4413 }; 4414 4415 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 4416 struct bpf_sk_lookup { 4417 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 4418 4419 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 4420 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 4421 __u32 remote_ip4; /* Network byte order */ 4422 __u32 remote_ip6[4]; /* Network byte order */ 4423 __u32 remote_port; /* Network byte order */ 4424 __u32 local_ip4; /* Network byte order */ 4425 __u32 local_ip6[4]; /* Network byte order */ 4426 __u32 local_port; /* Host byte order */ 4427 }; 4428 4429 #endif /* _UAPI__LINUX_BPF_H__ */ 4430