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 BPF_LINK_DETACH, 121 }; 122 123 enum bpf_map_type { 124 BPF_MAP_TYPE_UNSPEC, 125 BPF_MAP_TYPE_HASH, 126 BPF_MAP_TYPE_ARRAY, 127 BPF_MAP_TYPE_PROG_ARRAY, 128 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 129 BPF_MAP_TYPE_PERCPU_HASH, 130 BPF_MAP_TYPE_PERCPU_ARRAY, 131 BPF_MAP_TYPE_STACK_TRACE, 132 BPF_MAP_TYPE_CGROUP_ARRAY, 133 BPF_MAP_TYPE_LRU_HASH, 134 BPF_MAP_TYPE_LRU_PERCPU_HASH, 135 BPF_MAP_TYPE_LPM_TRIE, 136 BPF_MAP_TYPE_ARRAY_OF_MAPS, 137 BPF_MAP_TYPE_HASH_OF_MAPS, 138 BPF_MAP_TYPE_DEVMAP, 139 BPF_MAP_TYPE_SOCKMAP, 140 BPF_MAP_TYPE_CPUMAP, 141 BPF_MAP_TYPE_XSKMAP, 142 BPF_MAP_TYPE_SOCKHASH, 143 BPF_MAP_TYPE_CGROUP_STORAGE, 144 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 145 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 146 BPF_MAP_TYPE_QUEUE, 147 BPF_MAP_TYPE_STACK, 148 BPF_MAP_TYPE_SK_STORAGE, 149 BPF_MAP_TYPE_DEVMAP_HASH, 150 BPF_MAP_TYPE_STRUCT_OPS, 151 BPF_MAP_TYPE_RINGBUF, 152 }; 153 154 /* Note that tracing related programs such as 155 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 156 * are not subject to a stable API since kernel internal data 157 * structures can change from release to release and may 158 * therefore break existing tracing BPF programs. Tracing BPF 159 * programs correspond to /a/ specific kernel which is to be 160 * analyzed, and not /a/ specific kernel /and/ all future ones. 161 */ 162 enum bpf_prog_type { 163 BPF_PROG_TYPE_UNSPEC, 164 BPF_PROG_TYPE_SOCKET_FILTER, 165 BPF_PROG_TYPE_KPROBE, 166 BPF_PROG_TYPE_SCHED_CLS, 167 BPF_PROG_TYPE_SCHED_ACT, 168 BPF_PROG_TYPE_TRACEPOINT, 169 BPF_PROG_TYPE_XDP, 170 BPF_PROG_TYPE_PERF_EVENT, 171 BPF_PROG_TYPE_CGROUP_SKB, 172 BPF_PROG_TYPE_CGROUP_SOCK, 173 BPF_PROG_TYPE_LWT_IN, 174 BPF_PROG_TYPE_LWT_OUT, 175 BPF_PROG_TYPE_LWT_XMIT, 176 BPF_PROG_TYPE_SOCK_OPS, 177 BPF_PROG_TYPE_SK_SKB, 178 BPF_PROG_TYPE_CGROUP_DEVICE, 179 BPF_PROG_TYPE_SK_MSG, 180 BPF_PROG_TYPE_RAW_TRACEPOINT, 181 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 182 BPF_PROG_TYPE_LWT_SEG6LOCAL, 183 BPF_PROG_TYPE_LIRC_MODE2, 184 BPF_PROG_TYPE_SK_REUSEPORT, 185 BPF_PROG_TYPE_FLOW_DISSECTOR, 186 BPF_PROG_TYPE_CGROUP_SYSCTL, 187 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 188 BPF_PROG_TYPE_CGROUP_SOCKOPT, 189 BPF_PROG_TYPE_TRACING, 190 BPF_PROG_TYPE_STRUCT_OPS, 191 BPF_PROG_TYPE_EXT, 192 BPF_PROG_TYPE_LSM, 193 BPF_PROG_TYPE_SK_LOOKUP, 194 }; 195 196 enum bpf_attach_type { 197 BPF_CGROUP_INET_INGRESS, 198 BPF_CGROUP_INET_EGRESS, 199 BPF_CGROUP_INET_SOCK_CREATE, 200 BPF_CGROUP_SOCK_OPS, 201 BPF_SK_SKB_STREAM_PARSER, 202 BPF_SK_SKB_STREAM_VERDICT, 203 BPF_CGROUP_DEVICE, 204 BPF_SK_MSG_VERDICT, 205 BPF_CGROUP_INET4_BIND, 206 BPF_CGROUP_INET6_BIND, 207 BPF_CGROUP_INET4_CONNECT, 208 BPF_CGROUP_INET6_CONNECT, 209 BPF_CGROUP_INET4_POST_BIND, 210 BPF_CGROUP_INET6_POST_BIND, 211 BPF_CGROUP_UDP4_SENDMSG, 212 BPF_CGROUP_UDP6_SENDMSG, 213 BPF_LIRC_MODE2, 214 BPF_FLOW_DISSECTOR, 215 BPF_CGROUP_SYSCTL, 216 BPF_CGROUP_UDP4_RECVMSG, 217 BPF_CGROUP_UDP6_RECVMSG, 218 BPF_CGROUP_GETSOCKOPT, 219 BPF_CGROUP_SETSOCKOPT, 220 BPF_TRACE_RAW_TP, 221 BPF_TRACE_FENTRY, 222 BPF_TRACE_FEXIT, 223 BPF_MODIFY_RETURN, 224 BPF_LSM_MAC, 225 BPF_TRACE_ITER, 226 BPF_CGROUP_INET4_GETPEERNAME, 227 BPF_CGROUP_INET6_GETPEERNAME, 228 BPF_CGROUP_INET4_GETSOCKNAME, 229 BPF_CGROUP_INET6_GETSOCKNAME, 230 BPF_XDP_DEVMAP, 231 BPF_CGROUP_INET_SOCK_RELEASE, 232 BPF_XDP_CPUMAP, 233 BPF_SK_LOOKUP, 234 BPF_XDP, 235 __MAX_BPF_ATTACH_TYPE 236 }; 237 238 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 239 240 enum bpf_link_type { 241 BPF_LINK_TYPE_UNSPEC = 0, 242 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 243 BPF_LINK_TYPE_TRACING = 2, 244 BPF_LINK_TYPE_CGROUP = 3, 245 BPF_LINK_TYPE_ITER = 4, 246 BPF_LINK_TYPE_NETNS = 5, 247 BPF_LINK_TYPE_XDP = 6, 248 249 MAX_BPF_LINK_TYPE, 250 }; 251 252 enum bpf_iter_link_info { 253 BPF_ITER_LINK_UNSPEC = 0, 254 BPF_ITER_LINK_MAP_FD = 1, 255 256 MAX_BPF_ITER_LINK_INFO, 257 }; 258 259 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 260 * 261 * NONE(default): No further bpf programs allowed in the subtree. 262 * 263 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 264 * the program in this cgroup yields to sub-cgroup program. 265 * 266 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 267 * that cgroup program gets run in addition to the program in this cgroup. 268 * 269 * Only one program is allowed to be attached to a cgroup with 270 * NONE or BPF_F_ALLOW_OVERRIDE flag. 271 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 272 * release old program and attach the new one. Attach flags has to match. 273 * 274 * Multiple programs are allowed to be attached to a cgroup with 275 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 276 * (those that were attached first, run first) 277 * The programs of sub-cgroup are executed first, then programs of 278 * this cgroup and then programs of parent cgroup. 279 * When children program makes decision (like picking TCP CA or sock bind) 280 * parent program has a chance to override it. 281 * 282 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 283 * programs for a cgroup. Though it's possible to replace an old program at 284 * any position by also specifying BPF_F_REPLACE flag and position itself in 285 * replace_bpf_fd attribute. Old program at this position will be released. 286 * 287 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 288 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 289 * Ex1: 290 * cgrp1 (MULTI progs A, B) -> 291 * cgrp2 (OVERRIDE prog C) -> 292 * cgrp3 (MULTI prog D) -> 293 * cgrp4 (OVERRIDE prog E) -> 294 * cgrp5 (NONE prog F) 295 * the event in cgrp5 triggers execution of F,D,A,B in that order. 296 * if prog F is detached, the execution is E,D,A,B 297 * if prog F and D are detached, the execution is E,A,B 298 * if prog F, E and D are detached, the execution is C,A,B 299 * 300 * All eligible programs are executed regardless of return code from 301 * earlier programs. 302 */ 303 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 304 #define BPF_F_ALLOW_MULTI (1U << 1) 305 #define BPF_F_REPLACE (1U << 2) 306 307 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 308 * verifier will perform strict alignment checking as if the kernel 309 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 310 * and NET_IP_ALIGN defined to 2. 311 */ 312 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 313 314 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 315 * verifier will allow any alignment whatsoever. On platforms 316 * with strict alignment requirements for loads ands stores (such 317 * as sparc and mips) the verifier validates that all loads and 318 * stores provably follow this requirement. This flag turns that 319 * checking and enforcement off. 320 * 321 * It is mostly used for testing when we want to validate the 322 * context and memory access aspects of the verifier, but because 323 * of an unaligned access the alignment check would trigger before 324 * the one we are interested in. 325 */ 326 #define BPF_F_ANY_ALIGNMENT (1U << 1) 327 328 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 329 * Verifier does sub-register def/use analysis and identifies instructions whose 330 * def only matters for low 32-bit, high 32-bit is never referenced later 331 * through implicit zero extension. Therefore verifier notifies JIT back-ends 332 * that it is safe to ignore clearing high 32-bit for these instructions. This 333 * saves some back-ends a lot of code-gen. However such optimization is not 334 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 335 * hence hasn't used verifier's analysis result. But, we really want to have a 336 * way to be able to verify the correctness of the described optimization on 337 * x86_64 on which testsuites are frequently exercised. 338 * 339 * So, this flag is introduced. Once it is set, verifier will randomize high 340 * 32-bit for those instructions who has been identified as safe to ignore them. 341 * Then, if verifier is not doing correct analysis, such randomization will 342 * regress tests to expose bugs. 343 */ 344 #define BPF_F_TEST_RND_HI32 (1U << 2) 345 346 /* The verifier internal test flag. Behavior is undefined */ 347 #define BPF_F_TEST_STATE_FREQ (1U << 3) 348 349 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 350 * two extensions: 351 * 352 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 353 * insn[0].imm: map fd map fd 354 * insn[1].imm: 0 offset into value 355 * insn[0].off: 0 0 356 * insn[1].off: 0 0 357 * ldimm64 rewrite: address of map address of map[0]+offset 358 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 359 */ 360 #define BPF_PSEUDO_MAP_FD 1 361 #define BPF_PSEUDO_MAP_VALUE 2 362 363 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 364 * offset to another bpf function 365 */ 366 #define BPF_PSEUDO_CALL 1 367 368 /* flags for BPF_MAP_UPDATE_ELEM command */ 369 enum { 370 BPF_ANY = 0, /* create new element or update existing */ 371 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 372 BPF_EXIST = 2, /* update existing element */ 373 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 374 }; 375 376 /* flags for BPF_MAP_CREATE command */ 377 enum { 378 BPF_F_NO_PREALLOC = (1U << 0), 379 /* Instead of having one common LRU list in the 380 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 381 * which can scale and perform better. 382 * Note, the LRU nodes (including free nodes) cannot be moved 383 * across different LRU lists. 384 */ 385 BPF_F_NO_COMMON_LRU = (1U << 1), 386 /* Specify numa node during map creation */ 387 BPF_F_NUMA_NODE = (1U << 2), 388 389 /* Flags for accessing BPF object from syscall side. */ 390 BPF_F_RDONLY = (1U << 3), 391 BPF_F_WRONLY = (1U << 4), 392 393 /* Flag for stack_map, store build_id+offset instead of pointer */ 394 BPF_F_STACK_BUILD_ID = (1U << 5), 395 396 /* Zero-initialize hash function seed. This should only be used for testing. */ 397 BPF_F_ZERO_SEED = (1U << 6), 398 399 /* Flags for accessing BPF object from program side. */ 400 BPF_F_RDONLY_PROG = (1U << 7), 401 BPF_F_WRONLY_PROG = (1U << 8), 402 403 /* Clone map from listener for newly accepted socket */ 404 BPF_F_CLONE = (1U << 9), 405 406 /* Enable memory-mapping BPF map */ 407 BPF_F_MMAPABLE = (1U << 10), 408 }; 409 410 /* Flags for BPF_PROG_QUERY. */ 411 412 /* Query effective (directly attached + inherited from ancestor cgroups) 413 * programs that will be executed for events within a cgroup. 414 * attach_flags with this flag are returned only for directly attached programs. 415 */ 416 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 417 418 /* type for BPF_ENABLE_STATS */ 419 enum bpf_stats_type { 420 /* enabled run_time_ns and run_cnt */ 421 BPF_STATS_RUN_TIME = 0, 422 }; 423 424 enum bpf_stack_build_id_status { 425 /* user space need an empty entry to identify end of a trace */ 426 BPF_STACK_BUILD_ID_EMPTY = 0, 427 /* with valid build_id and offset */ 428 BPF_STACK_BUILD_ID_VALID = 1, 429 /* couldn't get build_id, fallback to ip */ 430 BPF_STACK_BUILD_ID_IP = 2, 431 }; 432 433 #define BPF_BUILD_ID_SIZE 20 434 struct bpf_stack_build_id { 435 __s32 status; 436 unsigned char build_id[BPF_BUILD_ID_SIZE]; 437 union { 438 __u64 offset; 439 __u64 ip; 440 }; 441 }; 442 443 #define BPF_OBJ_NAME_LEN 16U 444 445 union bpf_attr { 446 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 447 __u32 map_type; /* one of enum bpf_map_type */ 448 __u32 key_size; /* size of key in bytes */ 449 __u32 value_size; /* size of value in bytes */ 450 __u32 max_entries; /* max number of entries in a map */ 451 __u32 map_flags; /* BPF_MAP_CREATE related 452 * flags defined above. 453 */ 454 __u32 inner_map_fd; /* fd pointing to the inner map */ 455 __u32 numa_node; /* numa node (effective only if 456 * BPF_F_NUMA_NODE is set). 457 */ 458 char map_name[BPF_OBJ_NAME_LEN]; 459 __u32 map_ifindex; /* ifindex of netdev to create on */ 460 __u32 btf_fd; /* fd pointing to a BTF type data */ 461 __u32 btf_key_type_id; /* BTF type_id of the key */ 462 __u32 btf_value_type_id; /* BTF type_id of the value */ 463 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 464 * struct stored as the 465 * map value 466 */ 467 }; 468 469 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 470 __u32 map_fd; 471 __aligned_u64 key; 472 union { 473 __aligned_u64 value; 474 __aligned_u64 next_key; 475 }; 476 __u64 flags; 477 }; 478 479 struct { /* struct used by BPF_MAP_*_BATCH commands */ 480 __aligned_u64 in_batch; /* start batch, 481 * NULL to start from beginning 482 */ 483 __aligned_u64 out_batch; /* output: next start batch */ 484 __aligned_u64 keys; 485 __aligned_u64 values; 486 __u32 count; /* input/output: 487 * input: # of key/value 488 * elements 489 * output: # of filled elements 490 */ 491 __u32 map_fd; 492 __u64 elem_flags; 493 __u64 flags; 494 } batch; 495 496 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 497 __u32 prog_type; /* one of enum bpf_prog_type */ 498 __u32 insn_cnt; 499 __aligned_u64 insns; 500 __aligned_u64 license; 501 __u32 log_level; /* verbosity level of verifier */ 502 __u32 log_size; /* size of user buffer */ 503 __aligned_u64 log_buf; /* user supplied buffer */ 504 __u32 kern_version; /* not used */ 505 __u32 prog_flags; 506 char prog_name[BPF_OBJ_NAME_LEN]; 507 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 508 /* For some prog types expected attach type must be known at 509 * load time to verify attach type specific parts of prog 510 * (context accesses, allowed helpers, etc). 511 */ 512 __u32 expected_attach_type; 513 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 514 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 515 __aligned_u64 func_info; /* func info */ 516 __u32 func_info_cnt; /* number of bpf_func_info records */ 517 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 518 __aligned_u64 line_info; /* line info */ 519 __u32 line_info_cnt; /* number of bpf_line_info records */ 520 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 521 __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 522 }; 523 524 struct { /* anonymous struct used by BPF_OBJ_* commands */ 525 __aligned_u64 pathname; 526 __u32 bpf_fd; 527 __u32 file_flags; 528 }; 529 530 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 531 __u32 target_fd; /* container object to attach to */ 532 __u32 attach_bpf_fd; /* eBPF program to attach */ 533 __u32 attach_type; 534 __u32 attach_flags; 535 __u32 replace_bpf_fd; /* previously attached eBPF 536 * program to replace if 537 * BPF_F_REPLACE is used 538 */ 539 }; 540 541 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 542 __u32 prog_fd; 543 __u32 retval; 544 __u32 data_size_in; /* input: len of data_in */ 545 __u32 data_size_out; /* input/output: len of data_out 546 * returns ENOSPC if data_out 547 * is too small. 548 */ 549 __aligned_u64 data_in; 550 __aligned_u64 data_out; 551 __u32 repeat; 552 __u32 duration; 553 __u32 ctx_size_in; /* input: len of ctx_in */ 554 __u32 ctx_size_out; /* input/output: len of ctx_out 555 * returns ENOSPC if ctx_out 556 * is too small. 557 */ 558 __aligned_u64 ctx_in; 559 __aligned_u64 ctx_out; 560 } test; 561 562 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 563 union { 564 __u32 start_id; 565 __u32 prog_id; 566 __u32 map_id; 567 __u32 btf_id; 568 __u32 link_id; 569 }; 570 __u32 next_id; 571 __u32 open_flags; 572 }; 573 574 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 575 __u32 bpf_fd; 576 __u32 info_len; 577 __aligned_u64 info; 578 } info; 579 580 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 581 __u32 target_fd; /* container object to query */ 582 __u32 attach_type; 583 __u32 query_flags; 584 __u32 attach_flags; 585 __aligned_u64 prog_ids; 586 __u32 prog_cnt; 587 } query; 588 589 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 590 __u64 name; 591 __u32 prog_fd; 592 } raw_tracepoint; 593 594 struct { /* anonymous struct for BPF_BTF_LOAD */ 595 __aligned_u64 btf; 596 __aligned_u64 btf_log_buf; 597 __u32 btf_size; 598 __u32 btf_log_size; 599 __u32 btf_log_level; 600 }; 601 602 struct { 603 __u32 pid; /* input: pid */ 604 __u32 fd; /* input: fd */ 605 __u32 flags; /* input: flags */ 606 __u32 buf_len; /* input/output: buf len */ 607 __aligned_u64 buf; /* input/output: 608 * tp_name for tracepoint 609 * symbol for kprobe 610 * filename for uprobe 611 */ 612 __u32 prog_id; /* output: prod_id */ 613 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 614 __u64 probe_offset; /* output: probe_offset */ 615 __u64 probe_addr; /* output: probe_addr */ 616 } task_fd_query; 617 618 struct { /* struct used by BPF_LINK_CREATE command */ 619 __u32 prog_fd; /* eBPF program to attach */ 620 union { 621 __u32 target_fd; /* object to attach to */ 622 __u32 target_ifindex; /* target ifindex */ 623 }; 624 __u32 attach_type; /* attach type */ 625 __u32 flags; /* extra flags */ 626 } link_create; 627 628 struct { /* struct used by BPF_LINK_UPDATE command */ 629 __u32 link_fd; /* link fd */ 630 /* new program fd to update link with */ 631 __u32 new_prog_fd; 632 __u32 flags; /* extra flags */ 633 /* expected link's program fd; is specified only if 634 * BPF_F_REPLACE flag is set in flags */ 635 __u32 old_prog_fd; 636 } link_update; 637 638 struct { 639 __u32 link_fd; 640 } link_detach; 641 642 struct { /* struct used by BPF_ENABLE_STATS command */ 643 __u32 type; 644 } enable_stats; 645 646 struct { /* struct used by BPF_ITER_CREATE command */ 647 __u32 link_fd; 648 __u32 flags; 649 } iter_create; 650 651 } __attribute__((aligned(8))); 652 653 /* The description below is an attempt at providing documentation to eBPF 654 * developers about the multiple available eBPF helper functions. It can be 655 * parsed and used to produce a manual page. The workflow is the following, 656 * and requires the rst2man utility: 657 * 658 * $ ./scripts/bpf_helpers_doc.py \ 659 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 660 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 661 * $ man /tmp/bpf-helpers.7 662 * 663 * Note that in order to produce this external documentation, some RST 664 * formatting is used in the descriptions to get "bold" and "italics" in 665 * manual pages. Also note that the few trailing white spaces are 666 * intentional, removing them would break paragraphs for rst2man. 667 * 668 * Start of BPF helper function descriptions: 669 * 670 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 671 * Description 672 * Perform a lookup in *map* for an entry associated to *key*. 673 * Return 674 * Map value associated to *key*, or **NULL** if no entry was 675 * found. 676 * 677 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 678 * Description 679 * Add or update the value of the entry associated to *key* in 680 * *map* with *value*. *flags* is one of: 681 * 682 * **BPF_NOEXIST** 683 * The entry for *key* must not exist in the map. 684 * **BPF_EXIST** 685 * The entry for *key* must already exist in the map. 686 * **BPF_ANY** 687 * No condition on the existence of the entry for *key*. 688 * 689 * Flag value **BPF_NOEXIST** cannot be used for maps of types 690 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 691 * elements always exist), the helper would return an error. 692 * Return 693 * 0 on success, or a negative error in case of failure. 694 * 695 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 696 * Description 697 * Delete entry with *key* from *map*. 698 * Return 699 * 0 on success, or a negative error in case of failure. 700 * 701 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 702 * Description 703 * For tracing programs, safely attempt to read *size* bytes from 704 * kernel space address *unsafe_ptr* and store the data in *dst*. 705 * 706 * Generally, use **bpf_probe_read_user**\ () or 707 * **bpf_probe_read_kernel**\ () instead. 708 * Return 709 * 0 on success, or a negative error in case of failure. 710 * 711 * u64 bpf_ktime_get_ns(void) 712 * Description 713 * Return the time elapsed since system boot, in nanoseconds. 714 * Does not include time the system was suspended. 715 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 716 * Return 717 * Current *ktime*. 718 * 719 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 720 * Description 721 * This helper is a "printk()-like" facility for debugging. It 722 * prints a message defined by format *fmt* (of size *fmt_size*) 723 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 724 * available. It can take up to three additional **u64** 725 * arguments (as an eBPF helpers, the total number of arguments is 726 * limited to five). 727 * 728 * Each time the helper is called, it appends a line to the trace. 729 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 730 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 731 * The format of the trace is customizable, and the exact output 732 * one will get depends on the options set in 733 * *\/sys/kernel/debug/tracing/trace_options* (see also the 734 * *README* file under the same directory). However, it usually 735 * defaults to something like: 736 * 737 * :: 738 * 739 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 740 * 741 * In the above: 742 * 743 * * ``telnet`` is the name of the current task. 744 * * ``470`` is the PID of the current task. 745 * * ``001`` is the CPU number on which the task is 746 * running. 747 * * In ``.N..``, each character refers to a set of 748 * options (whether irqs are enabled, scheduling 749 * options, whether hard/softirqs are running, level of 750 * preempt_disabled respectively). **N** means that 751 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 752 * are set. 753 * * ``419421.045894`` is a timestamp. 754 * * ``0x00000001`` is a fake value used by BPF for the 755 * instruction pointer register. 756 * * ``<formatted msg>`` is the message formatted with 757 * *fmt*. 758 * 759 * The conversion specifiers supported by *fmt* are similar, but 760 * more limited than for printk(). They are **%d**, **%i**, 761 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 762 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 763 * of field, padding with zeroes, etc.) is available, and the 764 * helper will return **-EINVAL** (but print nothing) if it 765 * encounters an unknown specifier. 766 * 767 * Also, note that **bpf_trace_printk**\ () is slow, and should 768 * only be used for debugging purposes. For this reason, a notice 769 * bloc (spanning several lines) is printed to kernel logs and 770 * states that the helper should not be used "for production use" 771 * the first time this helper is used (or more precisely, when 772 * **trace_printk**\ () buffers are allocated). For passing values 773 * to user space, perf events should be preferred. 774 * Return 775 * The number of bytes written to the buffer, or a negative error 776 * in case of failure. 777 * 778 * u32 bpf_get_prandom_u32(void) 779 * Description 780 * Get a pseudo-random number. 781 * 782 * From a security point of view, this helper uses its own 783 * pseudo-random internal state, and cannot be used to infer the 784 * seed of other random functions in the kernel. However, it is 785 * essential to note that the generator used by the helper is not 786 * cryptographically secure. 787 * Return 788 * A random 32-bit unsigned value. 789 * 790 * u32 bpf_get_smp_processor_id(void) 791 * Description 792 * Get the SMP (symmetric multiprocessing) processor id. Note that 793 * all programs run with preemption disabled, which means that the 794 * SMP processor id is stable during all the execution of the 795 * program. 796 * Return 797 * The SMP id of the processor running the program. 798 * 799 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 800 * Description 801 * Store *len* bytes from address *from* into the packet 802 * associated to *skb*, at *offset*. *flags* are a combination of 803 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 804 * checksum for the packet after storing the bytes) and 805 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 806 * **->swhash** and *skb*\ **->l4hash** to 0). 807 * 808 * A call to this helper is susceptible to change the underlying 809 * packet buffer. Therefore, at load time, all checks on pointers 810 * previously done by the verifier are invalidated and must be 811 * performed again, if the helper is used in combination with 812 * direct packet access. 813 * Return 814 * 0 on success, or a negative error in case of failure. 815 * 816 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 817 * Description 818 * Recompute the layer 3 (e.g. IP) checksum for the packet 819 * associated to *skb*. Computation is incremental, so the helper 820 * must know the former value of the header field that was 821 * modified (*from*), the new value of this field (*to*), and the 822 * number of bytes (2 or 4) for this field, stored in *size*. 823 * Alternatively, it is possible to store the difference between 824 * the previous and the new values of the header field in *to*, by 825 * setting *from* and *size* to 0. For both methods, *offset* 826 * indicates the location of the IP checksum within the packet. 827 * 828 * This helper works in combination with **bpf_csum_diff**\ (), 829 * which does not update the checksum in-place, but offers more 830 * flexibility and can handle sizes larger than 2 or 4 for the 831 * checksum to update. 832 * 833 * A call to this helper is susceptible to change the underlying 834 * packet buffer. Therefore, at load time, all checks on pointers 835 * previously done by the verifier are invalidated and must be 836 * performed again, if the helper is used in combination with 837 * direct packet access. 838 * Return 839 * 0 on success, or a negative error in case of failure. 840 * 841 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 842 * Description 843 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 844 * packet associated to *skb*. Computation is incremental, so the 845 * helper must know the former value of the header field that was 846 * modified (*from*), the new value of this field (*to*), and the 847 * number of bytes (2 or 4) for this field, stored on the lowest 848 * four bits of *flags*. Alternatively, it is possible to store 849 * the difference between the previous and the new values of the 850 * header field in *to*, by setting *from* and the four lowest 851 * bits of *flags* to 0. For both methods, *offset* indicates the 852 * location of the IP checksum within the packet. In addition to 853 * the size of the field, *flags* can be added (bitwise OR) actual 854 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 855 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 856 * for updates resulting in a null checksum the value is set to 857 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 858 * the checksum is to be computed against a pseudo-header. 859 * 860 * This helper works in combination with **bpf_csum_diff**\ (), 861 * which does not update the checksum in-place, but offers more 862 * flexibility and can handle sizes larger than 2 or 4 for the 863 * checksum to update. 864 * 865 * A call to this helper is susceptible to change the underlying 866 * packet buffer. Therefore, at load time, all checks on pointers 867 * previously done by the verifier are invalidated and must be 868 * performed again, if the helper is used in combination with 869 * direct packet access. 870 * Return 871 * 0 on success, or a negative error in case of failure. 872 * 873 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 874 * Description 875 * This special helper is used to trigger a "tail call", or in 876 * other words, to jump into another eBPF program. The same stack 877 * frame is used (but values on stack and in registers for the 878 * caller are not accessible to the callee). This mechanism allows 879 * for program chaining, either for raising the maximum number of 880 * available eBPF instructions, or to execute given programs in 881 * conditional blocks. For security reasons, there is an upper 882 * limit to the number of successive tail calls that can be 883 * performed. 884 * 885 * Upon call of this helper, the program attempts to jump into a 886 * program referenced at index *index* in *prog_array_map*, a 887 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 888 * *ctx*, a pointer to the context. 889 * 890 * If the call succeeds, the kernel immediately runs the first 891 * instruction of the new program. This is not a function call, 892 * and it never returns to the previous program. If the call 893 * fails, then the helper has no effect, and the caller continues 894 * to run its subsequent instructions. A call can fail if the 895 * destination program for the jump does not exist (i.e. *index* 896 * is superior to the number of entries in *prog_array_map*), or 897 * if the maximum number of tail calls has been reached for this 898 * chain of programs. This limit is defined in the kernel by the 899 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 900 * which is currently set to 32. 901 * Return 902 * 0 on success, or a negative error in case of failure. 903 * 904 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 905 * Description 906 * Clone and redirect the packet associated to *skb* to another 907 * net device of index *ifindex*. Both ingress and egress 908 * interfaces can be used for redirection. The **BPF_F_INGRESS** 909 * value in *flags* is used to make the distinction (ingress path 910 * is selected if the flag is present, egress path otherwise). 911 * This is the only flag supported for now. 912 * 913 * In comparison with **bpf_redirect**\ () helper, 914 * **bpf_clone_redirect**\ () has the associated cost of 915 * duplicating the packet buffer, but this can be executed out of 916 * the eBPF program. Conversely, **bpf_redirect**\ () is more 917 * efficient, but it is handled through an action code where the 918 * redirection happens only after the eBPF program has returned. 919 * 920 * A call to this helper is susceptible to change the underlying 921 * packet buffer. Therefore, at load time, all checks on pointers 922 * previously done by the verifier are invalidated and must be 923 * performed again, if the helper is used in combination with 924 * direct packet access. 925 * Return 926 * 0 on success, or a negative error in case of failure. 927 * 928 * u64 bpf_get_current_pid_tgid(void) 929 * Return 930 * A 64-bit integer containing the current tgid and pid, and 931 * created as such: 932 * *current_task*\ **->tgid << 32 \|** 933 * *current_task*\ **->pid**. 934 * 935 * u64 bpf_get_current_uid_gid(void) 936 * Return 937 * A 64-bit integer containing the current GID and UID, and 938 * created as such: *current_gid* **<< 32 \|** *current_uid*. 939 * 940 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 941 * Description 942 * Copy the **comm** attribute of the current task into *buf* of 943 * *size_of_buf*. The **comm** attribute contains the name of 944 * the executable (excluding the path) for the current task. The 945 * *size_of_buf* must be strictly positive. On success, the 946 * helper makes sure that the *buf* is NUL-terminated. On failure, 947 * it is filled with zeroes. 948 * Return 949 * 0 on success, or a negative error in case of failure. 950 * 951 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 952 * Description 953 * Retrieve the classid for the current task, i.e. for the net_cls 954 * cgroup to which *skb* belongs. 955 * 956 * This helper can be used on TC egress path, but not on ingress. 957 * 958 * The net_cls cgroup provides an interface to tag network packets 959 * based on a user-provided identifier for all traffic coming from 960 * the tasks belonging to the related cgroup. See also the related 961 * kernel documentation, available from the Linux sources in file 962 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 963 * 964 * The Linux kernel has two versions for cgroups: there are 965 * cgroups v1 and cgroups v2. Both are available to users, who can 966 * use a mixture of them, but note that the net_cls cgroup is for 967 * cgroup v1 only. This makes it incompatible with BPF programs 968 * run on cgroups, which is a cgroup-v2-only feature (a socket can 969 * only hold data for one version of cgroups at a time). 970 * 971 * This helper is only available is the kernel was compiled with 972 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 973 * "**y**" or to "**m**". 974 * Return 975 * The classid, or 0 for the default unconfigured classid. 976 * 977 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 978 * Description 979 * Push a *vlan_tci* (VLAN tag control information) of protocol 980 * *vlan_proto* to the packet associated to *skb*, then update 981 * the checksum. Note that if *vlan_proto* is different from 982 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 983 * be **ETH_P_8021Q**. 984 * 985 * A call to this helper is susceptible to change the underlying 986 * packet buffer. Therefore, at load time, all checks on pointers 987 * previously done by the verifier are invalidated and must be 988 * performed again, if the helper is used in combination with 989 * direct packet access. 990 * Return 991 * 0 on success, or a negative error in case of failure. 992 * 993 * long bpf_skb_vlan_pop(struct sk_buff *skb) 994 * Description 995 * Pop a VLAN header from the packet associated to *skb*. 996 * 997 * A call to this helper is susceptible to change the underlying 998 * packet buffer. Therefore, at load time, all checks on pointers 999 * previously done by the verifier are invalidated and must be 1000 * performed again, if the helper is used in combination with 1001 * direct packet access. 1002 * Return 1003 * 0 on success, or a negative error in case of failure. 1004 * 1005 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1006 * Description 1007 * Get tunnel metadata. This helper takes a pointer *key* to an 1008 * empty **struct bpf_tunnel_key** of **size**, that will be 1009 * filled with tunnel metadata for the packet associated to *skb*. 1010 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1011 * indicates that the tunnel is based on IPv6 protocol instead of 1012 * IPv4. 1013 * 1014 * The **struct bpf_tunnel_key** is an object that generalizes the 1015 * principal parameters used by various tunneling protocols into a 1016 * single struct. This way, it can be used to easily make a 1017 * decision based on the contents of the encapsulation header, 1018 * "summarized" in this struct. In particular, it holds the IP 1019 * address of the remote end (IPv4 or IPv6, depending on the case) 1020 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1021 * this struct exposes the *key*\ **->tunnel_id**, which is 1022 * generally mapped to a VNI (Virtual Network Identifier), making 1023 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1024 * () helper. 1025 * 1026 * Let's imagine that the following code is part of a program 1027 * attached to the TC ingress interface, on one end of a GRE 1028 * tunnel, and is supposed to filter out all messages coming from 1029 * remote ends with IPv4 address other than 10.0.0.1: 1030 * 1031 * :: 1032 * 1033 * int ret; 1034 * struct bpf_tunnel_key key = {}; 1035 * 1036 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1037 * if (ret < 0) 1038 * return TC_ACT_SHOT; // drop packet 1039 * 1040 * if (key.remote_ipv4 != 0x0a000001) 1041 * return TC_ACT_SHOT; // drop packet 1042 * 1043 * return TC_ACT_OK; // accept packet 1044 * 1045 * This interface can also be used with all encapsulation devices 1046 * that can operate in "collect metadata" mode: instead of having 1047 * one network device per specific configuration, the "collect 1048 * metadata" mode only requires a single device where the 1049 * configuration can be extracted from this helper. 1050 * 1051 * This can be used together with various tunnels such as VXLan, 1052 * Geneve, GRE or IP in IP (IPIP). 1053 * Return 1054 * 0 on success, or a negative error in case of failure. 1055 * 1056 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1057 * Description 1058 * Populate tunnel metadata for packet associated to *skb.* The 1059 * tunnel metadata is set to the contents of *key*, of *size*. The 1060 * *flags* can be set to a combination of the following values: 1061 * 1062 * **BPF_F_TUNINFO_IPV6** 1063 * Indicate that the tunnel is based on IPv6 protocol 1064 * instead of IPv4. 1065 * **BPF_F_ZERO_CSUM_TX** 1066 * For IPv4 packets, add a flag to tunnel metadata 1067 * indicating that checksum computation should be skipped 1068 * and checksum set to zeroes. 1069 * **BPF_F_DONT_FRAGMENT** 1070 * Add a flag to tunnel metadata indicating that the 1071 * packet should not be fragmented. 1072 * **BPF_F_SEQ_NUMBER** 1073 * Add a flag to tunnel metadata indicating that a 1074 * sequence number should be added to tunnel header before 1075 * sending the packet. This flag was added for GRE 1076 * encapsulation, but might be used with other protocols 1077 * as well in the future. 1078 * 1079 * Here is a typical usage on the transmit path: 1080 * 1081 * :: 1082 * 1083 * struct bpf_tunnel_key key; 1084 * populate key ... 1085 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1086 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1087 * 1088 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1089 * helper for additional information. 1090 * Return 1091 * 0 on success, or a negative error in case of failure. 1092 * 1093 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1094 * Description 1095 * Read the value of a perf event counter. This helper relies on a 1096 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1097 * the perf event counter is selected when *map* is updated with 1098 * perf event file descriptors. The *map* is an array whose size 1099 * is the number of available CPUs, and each cell contains a value 1100 * relative to one CPU. The value to retrieve is indicated by 1101 * *flags*, that contains the index of the CPU to look up, masked 1102 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1103 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1104 * current CPU should be retrieved. 1105 * 1106 * Note that before Linux 4.13, only hardware perf event can be 1107 * retrieved. 1108 * 1109 * Also, be aware that the newer helper 1110 * **bpf_perf_event_read_value**\ () is recommended over 1111 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1112 * quirks where error and counter value are used as a return code 1113 * (which is wrong to do since ranges may overlap). This issue is 1114 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1115 * time provides more features over the **bpf_perf_event_read**\ 1116 * () interface. Please refer to the description of 1117 * **bpf_perf_event_read_value**\ () for details. 1118 * Return 1119 * The value of the perf event counter read from the map, or a 1120 * negative error code in case of failure. 1121 * 1122 * long bpf_redirect(u32 ifindex, u64 flags) 1123 * Description 1124 * Redirect the packet to another net device of index *ifindex*. 1125 * This helper is somewhat similar to **bpf_clone_redirect**\ 1126 * (), except that the packet is not cloned, which provides 1127 * increased performance. 1128 * 1129 * Except for XDP, both ingress and egress interfaces can be used 1130 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1131 * to make the distinction (ingress path is selected if the flag 1132 * is present, egress path otherwise). Currently, XDP only 1133 * supports redirection to the egress interface, and accepts no 1134 * flag at all. 1135 * 1136 * The same effect can also be attained with the more generic 1137 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1138 * redirect target instead of providing it directly to the helper. 1139 * Return 1140 * For XDP, the helper returns **XDP_REDIRECT** on success or 1141 * **XDP_ABORTED** on error. For other program types, the values 1142 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1143 * error. 1144 * 1145 * u32 bpf_get_route_realm(struct sk_buff *skb) 1146 * Description 1147 * Retrieve the realm or the route, that is to say the 1148 * **tclassid** field of the destination for the *skb*. The 1149 * indentifier retrieved is a user-provided tag, similar to the 1150 * one used with the net_cls cgroup (see description for 1151 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1152 * held by a route (a destination entry), not by a task. 1153 * 1154 * Retrieving this identifier works with the clsact TC egress hook 1155 * (see also **tc-bpf(8)**), or alternatively on conventional 1156 * classful egress qdiscs, but not on TC ingress path. In case of 1157 * clsact TC egress hook, this has the advantage that, internally, 1158 * the destination entry has not been dropped yet in the transmit 1159 * path. Therefore, the destination entry does not need to be 1160 * artificially held via **netif_keep_dst**\ () for a classful 1161 * qdisc until the *skb* is freed. 1162 * 1163 * This helper is available only if the kernel was compiled with 1164 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1165 * Return 1166 * The realm of the route for the packet associated to *skb*, or 0 1167 * if none was found. 1168 * 1169 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1170 * Description 1171 * Write raw *data* blob into a special BPF perf event held by 1172 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1173 * event must have the following attributes: **PERF_SAMPLE_RAW** 1174 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1175 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1176 * 1177 * The *flags* are used to indicate the index in *map* for which 1178 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1179 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1180 * to indicate that the index of the current CPU core should be 1181 * used. 1182 * 1183 * The value to write, of *size*, is passed through eBPF stack and 1184 * pointed by *data*. 1185 * 1186 * The context of the program *ctx* needs also be passed to the 1187 * helper. 1188 * 1189 * On user space, a program willing to read the values needs to 1190 * call **perf_event_open**\ () on the perf event (either for 1191 * one or for all CPUs) and to store the file descriptor into the 1192 * *map*. This must be done before the eBPF program can send data 1193 * into it. An example is available in file 1194 * *samples/bpf/trace_output_user.c* in the Linux kernel source 1195 * tree (the eBPF program counterpart is in 1196 * *samples/bpf/trace_output_kern.c*). 1197 * 1198 * **bpf_perf_event_output**\ () achieves better performance 1199 * than **bpf_trace_printk**\ () for sharing data with user 1200 * space, and is much better suitable for streaming data from eBPF 1201 * programs. 1202 * 1203 * Note that this helper is not restricted to tracing use cases 1204 * and can be used with programs attached to TC or XDP as well, 1205 * where it allows for passing data to user space listeners. Data 1206 * can be: 1207 * 1208 * * Only custom structs, 1209 * * Only the packet payload, or 1210 * * A combination of both. 1211 * Return 1212 * 0 on success, or a negative error in case of failure. 1213 * 1214 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 1215 * Description 1216 * This helper was provided as an easy way to load data from a 1217 * packet. It can be used to load *len* bytes from *offset* from 1218 * the packet associated to *skb*, into the buffer pointed by 1219 * *to*. 1220 * 1221 * Since Linux 4.7, usage of this helper has mostly been replaced 1222 * by "direct packet access", enabling packet data to be 1223 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 1224 * pointing respectively to the first byte of packet data and to 1225 * the byte after the last byte of packet data. However, it 1226 * remains useful if one wishes to read large quantities of data 1227 * at once from a packet into the eBPF stack. 1228 * Return 1229 * 0 on success, or a negative error in case of failure. 1230 * 1231 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1232 * Description 1233 * Walk a user or a kernel stack and return its id. To achieve 1234 * this, the helper needs *ctx*, which is a pointer to the context 1235 * on which the tracing program is executed, and a pointer to a 1236 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1237 * 1238 * The last argument, *flags*, holds the number of stack frames to 1239 * skip (from 0 to 255), masked with 1240 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1241 * a combination of the following flags: 1242 * 1243 * **BPF_F_USER_STACK** 1244 * Collect a user space stack instead of a kernel stack. 1245 * **BPF_F_FAST_STACK_CMP** 1246 * Compare stacks by hash only. 1247 * **BPF_F_REUSE_STACKID** 1248 * If two different stacks hash into the same *stackid*, 1249 * discard the old one. 1250 * 1251 * The stack id retrieved is a 32 bit long integer handle which 1252 * can be further combined with other data (including other stack 1253 * ids) and used as a key into maps. This can be useful for 1254 * generating a variety of graphs (such as flame graphs or off-cpu 1255 * graphs). 1256 * 1257 * For walking a stack, this helper is an improvement over 1258 * **bpf_probe_read**\ (), which can be used with unrolled loops 1259 * but is not efficient and consumes a lot of eBPF instructions. 1260 * Instead, **bpf_get_stackid**\ () can collect up to 1261 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1262 * this limit can be controlled with the **sysctl** program, and 1263 * that it should be manually increased in order to profile long 1264 * user stacks (such as stacks for Java programs). To do so, use: 1265 * 1266 * :: 1267 * 1268 * # sysctl kernel.perf_event_max_stack=<new value> 1269 * Return 1270 * The positive or null stack id on success, or a negative error 1271 * in case of failure. 1272 * 1273 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 1274 * Description 1275 * Compute a checksum difference, from the raw buffer pointed by 1276 * *from*, of length *from_size* (that must be a multiple of 4), 1277 * towards the raw buffer pointed by *to*, of size *to_size* 1278 * (same remark). An optional *seed* can be added to the value 1279 * (this can be cascaded, the seed may come from a previous call 1280 * to the helper). 1281 * 1282 * This is flexible enough to be used in several ways: 1283 * 1284 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 1285 * checksum, it can be used when pushing new data. 1286 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 1287 * checksum, it can be used when removing data from a packet. 1288 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 1289 * can be used to compute a diff. Note that *from_size* and 1290 * *to_size* do not need to be equal. 1291 * 1292 * This helper can be used in combination with 1293 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 1294 * which one can feed in the difference computed with 1295 * **bpf_csum_diff**\ (). 1296 * Return 1297 * The checksum result, or a negative error code in case of 1298 * failure. 1299 * 1300 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1301 * Description 1302 * Retrieve tunnel options metadata for the packet associated to 1303 * *skb*, and store the raw tunnel option data to the buffer *opt* 1304 * of *size*. 1305 * 1306 * This helper can be used with encapsulation devices that can 1307 * operate in "collect metadata" mode (please refer to the related 1308 * note in the description of **bpf_skb_get_tunnel_key**\ () for 1309 * more details). A particular example where this can be used is 1310 * in combination with the Geneve encapsulation protocol, where it 1311 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 1312 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 1313 * the eBPF program. This allows for full customization of these 1314 * headers. 1315 * Return 1316 * The size of the option data retrieved. 1317 * 1318 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 1319 * Description 1320 * Set tunnel options metadata for the packet associated to *skb* 1321 * to the option data contained in the raw buffer *opt* of *size*. 1322 * 1323 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 1324 * helper for additional information. 1325 * Return 1326 * 0 on success, or a negative error in case of failure. 1327 * 1328 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 1329 * Description 1330 * Change the protocol of the *skb* to *proto*. Currently 1331 * supported are transition from IPv4 to IPv6, and from IPv6 to 1332 * IPv4. The helper takes care of the groundwork for the 1333 * transition, including resizing the socket buffer. The eBPF 1334 * program is expected to fill the new headers, if any, via 1335 * **skb_store_bytes**\ () and to recompute the checksums with 1336 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 1337 * (). The main case for this helper is to perform NAT64 1338 * operations out of an eBPF program. 1339 * 1340 * Internally, the GSO type is marked as dodgy so that headers are 1341 * checked and segments are recalculated by the GSO/GRO engine. 1342 * The size for GSO target is adapted as well. 1343 * 1344 * All values for *flags* are reserved for future usage, and must 1345 * be left at zero. 1346 * 1347 * A call to this helper is susceptible to change the underlying 1348 * packet buffer. Therefore, at load time, all checks on pointers 1349 * previously done by the verifier are invalidated and must be 1350 * performed again, if the helper is used in combination with 1351 * direct packet access. 1352 * Return 1353 * 0 on success, or a negative error in case of failure. 1354 * 1355 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 1356 * Description 1357 * Change the packet type for the packet associated to *skb*. This 1358 * comes down to setting *skb*\ **->pkt_type** to *type*, except 1359 * the eBPF program does not have a write access to *skb*\ 1360 * **->pkt_type** beside this helper. Using a helper here allows 1361 * for graceful handling of errors. 1362 * 1363 * The major use case is to change incoming *skb*s to 1364 * **PACKET_HOST** in a programmatic way instead of having to 1365 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 1366 * example. 1367 * 1368 * Note that *type* only allows certain values. At this time, they 1369 * are: 1370 * 1371 * **PACKET_HOST** 1372 * Packet is for us. 1373 * **PACKET_BROADCAST** 1374 * Send packet to all. 1375 * **PACKET_MULTICAST** 1376 * Send packet to group. 1377 * **PACKET_OTHERHOST** 1378 * Send packet to someone else. 1379 * Return 1380 * 0 on success, or a negative error in case of failure. 1381 * 1382 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1383 * Description 1384 * Check whether *skb* is a descendant of the cgroup2 held by 1385 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1386 * Return 1387 * The return value depends on the result of the test, and can be: 1388 * 1389 * * 0, if the *skb* failed the cgroup2 descendant test. 1390 * * 1, if the *skb* succeeded the cgroup2 descendant test. 1391 * * A negative error code, if an error occurred. 1392 * 1393 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1394 * Description 1395 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1396 * not set, in particular if the hash was cleared due to mangling, 1397 * recompute this hash. Later accesses to the hash can be done 1398 * directly with *skb*\ **->hash**. 1399 * 1400 * Calling **bpf_set_hash_invalid**\ (), changing a packet 1401 * prototype with **bpf_skb_change_proto**\ (), or calling 1402 * **bpf_skb_store_bytes**\ () with the 1403 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1404 * the hash and to trigger a new computation for the next call to 1405 * **bpf_get_hash_recalc**\ (). 1406 * Return 1407 * The 32-bit hash. 1408 * 1409 * u64 bpf_get_current_task(void) 1410 * Return 1411 * A pointer to the current task struct. 1412 * 1413 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 1414 * Description 1415 * Attempt in a safe way to write *len* bytes from the buffer 1416 * *src* to *dst* in memory. It only works for threads that are in 1417 * user context, and *dst* must be a valid user space address. 1418 * 1419 * This helper should not be used to implement any kind of 1420 * security mechanism because of TOC-TOU attacks, but rather to 1421 * debug, divert, and manipulate execution of semi-cooperative 1422 * processes. 1423 * 1424 * Keep in mind that this feature is meant for experiments, and it 1425 * has a risk of crashing the system and running programs. 1426 * Therefore, when an eBPF program using this helper is attached, 1427 * a warning including PID and process name is printed to kernel 1428 * logs. 1429 * Return 1430 * 0 on success, or a negative error in case of failure. 1431 * 1432 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1433 * Description 1434 * Check whether the probe is being run is the context of a given 1435 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1436 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1437 * Return 1438 * The return value depends on the result of the test, and can be: 1439 * 1440 * * 0, if the *skb* task belongs to the cgroup2. 1441 * * 1, if the *skb* task does not belong to the cgroup2. 1442 * * A negative error code, if an error occurred. 1443 * 1444 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1445 * Description 1446 * Resize (trim or grow) the packet associated to *skb* to the 1447 * new *len*. The *flags* are reserved for future usage, and must 1448 * be left at zero. 1449 * 1450 * The basic idea is that the helper performs the needed work to 1451 * change the size of the packet, then the eBPF program rewrites 1452 * the rest via helpers like **bpf_skb_store_bytes**\ (), 1453 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1454 * and others. This helper is a slow path utility intended for 1455 * replies with control messages. And because it is targeted for 1456 * slow path, the helper itself can afford to be slow: it 1457 * implicitly linearizes, unclones and drops offloads from the 1458 * *skb*. 1459 * 1460 * A call to this helper is susceptible to change the underlying 1461 * packet buffer. Therefore, at load time, all checks on pointers 1462 * previously done by the verifier are invalidated and must be 1463 * performed again, if the helper is used in combination with 1464 * direct packet access. 1465 * Return 1466 * 0 on success, or a negative error in case of failure. 1467 * 1468 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1469 * Description 1470 * Pull in non-linear data in case the *skb* is non-linear and not 1471 * all of *len* are part of the linear section. Make *len* bytes 1472 * from *skb* readable and writable. If a zero value is passed for 1473 * *len*, then the whole length of the *skb* is pulled. 1474 * 1475 * This helper is only needed for reading and writing with direct 1476 * packet access. 1477 * 1478 * For direct packet access, testing that offsets to access 1479 * are within packet boundaries (test on *skb*\ **->data_end**) is 1480 * susceptible to fail if offsets are invalid, or if the requested 1481 * data is in non-linear parts of the *skb*. On failure the 1482 * program can just bail out, or in the case of a non-linear 1483 * buffer, use a helper to make the data available. The 1484 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1485 * the data. Another one consists in using **bpf_skb_pull_data** 1486 * to pull in once the non-linear parts, then retesting and 1487 * eventually access the data. 1488 * 1489 * At the same time, this also makes sure the *skb* is uncloned, 1490 * which is a necessary condition for direct write. As this needs 1491 * to be an invariant for the write part only, the verifier 1492 * detects writes and adds a prologue that is calling 1493 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1494 * the very beginning in case it is indeed cloned. 1495 * 1496 * A call to this helper is susceptible to change the underlying 1497 * packet buffer. Therefore, at load time, all checks on pointers 1498 * previously done by the verifier are invalidated and must be 1499 * performed again, if the helper is used in combination with 1500 * direct packet access. 1501 * Return 1502 * 0 on success, or a negative error in case of failure. 1503 * 1504 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1505 * Description 1506 * Add the checksum *csum* into *skb*\ **->csum** in case the 1507 * driver has supplied a checksum for the entire packet into that 1508 * field. Return an error otherwise. This helper is intended to be 1509 * used in combination with **bpf_csum_diff**\ (), in particular 1510 * when the checksum needs to be updated after data has been 1511 * written into the packet through direct packet access. 1512 * Return 1513 * The checksum on success, or a negative error code in case of 1514 * failure. 1515 * 1516 * void bpf_set_hash_invalid(struct sk_buff *skb) 1517 * Description 1518 * Invalidate the current *skb*\ **->hash**. It can be used after 1519 * mangling on headers through direct packet access, in order to 1520 * indicate that the hash is outdated and to trigger a 1521 * recalculation the next time the kernel tries to access this 1522 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1523 * 1524 * long bpf_get_numa_node_id(void) 1525 * Description 1526 * Return the id of the current NUMA node. The primary use case 1527 * for this helper is the selection of sockets for the local NUMA 1528 * node, when the program is attached to sockets using the 1529 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1530 * but the helper is also available to other eBPF program types, 1531 * similarly to **bpf_get_smp_processor_id**\ (). 1532 * Return 1533 * The id of current NUMA node. 1534 * 1535 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1536 * Description 1537 * Grows headroom of packet associated to *skb* and adjusts the 1538 * offset of the MAC header accordingly, adding *len* bytes of 1539 * space. It automatically extends and reallocates memory as 1540 * required. 1541 * 1542 * This helper can be used on a layer 3 *skb* to push a MAC header 1543 * for redirection into a layer 2 device. 1544 * 1545 * All values for *flags* are reserved for future usage, and must 1546 * be left at zero. 1547 * 1548 * A call to this helper is susceptible to change the underlying 1549 * packet buffer. Therefore, at load time, all checks on pointers 1550 * previously done by the verifier are invalidated and must be 1551 * performed again, if the helper is used in combination with 1552 * direct packet access. 1553 * Return 1554 * 0 on success, or a negative error in case of failure. 1555 * 1556 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1557 * Description 1558 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1559 * it is possible to use a negative value for *delta*. This helper 1560 * can be used to prepare the packet for pushing or popping 1561 * headers. 1562 * 1563 * A call to this helper is susceptible to change the underlying 1564 * packet buffer. Therefore, at load time, all checks on pointers 1565 * previously done by the verifier are invalidated and must be 1566 * performed again, if the helper is used in combination with 1567 * direct packet access. 1568 * Return 1569 * 0 on success, or a negative error in case of failure. 1570 * 1571 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1572 * Description 1573 * Copy a NUL terminated string from an unsafe kernel address 1574 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1575 * more details. 1576 * 1577 * Generally, use **bpf_probe_read_user_str**\ () or 1578 * **bpf_probe_read_kernel_str**\ () instead. 1579 * Return 1580 * On success, the strictly positive length of the string, 1581 * including the trailing NUL character. On error, a negative 1582 * value. 1583 * 1584 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1585 * Description 1586 * If the **struct sk_buff** pointed by *skb* has a known socket, 1587 * retrieve the cookie (generated by the kernel) of this socket. 1588 * If no cookie has been set yet, generate a new cookie. Once 1589 * generated, the socket cookie remains stable for the life of the 1590 * socket. This helper can be useful for monitoring per socket 1591 * networking traffic statistics as it provides a global socket 1592 * identifier that can be assumed unique. 1593 * Return 1594 * A 8-byte long non-decreasing number on success, or 0 if the 1595 * socket field is missing inside *skb*. 1596 * 1597 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1598 * Description 1599 * Equivalent to bpf_get_socket_cookie() helper that accepts 1600 * *skb*, but gets socket from **struct bpf_sock_addr** context. 1601 * Return 1602 * A 8-byte long non-decreasing number. 1603 * 1604 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1605 * Description 1606 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 1607 * *skb*, but gets socket from **struct bpf_sock_ops** context. 1608 * Return 1609 * A 8-byte long non-decreasing number. 1610 * 1611 * u32 bpf_get_socket_uid(struct sk_buff *skb) 1612 * Return 1613 * The owner UID of the socket associated to *skb*. If the socket 1614 * is **NULL**, or if it is not a full socket (i.e. if it is a 1615 * time-wait or a request socket instead), **overflowuid** value 1616 * is returned (note that **overflowuid** might also be the actual 1617 * UID value for the socket). 1618 * 1619 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 1620 * Description 1621 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1622 * to value *hash*. 1623 * Return 1624 * 0 1625 * 1626 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1627 * Description 1628 * Emulate a call to **setsockopt()** on the socket associated to 1629 * *bpf_socket*, which must be a full socket. The *level* at 1630 * which the option resides and the name *optname* of the option 1631 * must be specified, see **setsockopt(2)** for more information. 1632 * The option value of length *optlen* is pointed by *optval*. 1633 * 1634 * *bpf_socket* should be one of the following: 1635 * 1636 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1637 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1638 * and **BPF_CGROUP_INET6_CONNECT**. 1639 * 1640 * This helper actually implements a subset of **setsockopt()**. 1641 * It supports the following *level*\ s: 1642 * 1643 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1644 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1645 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1646 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 1647 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1648 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1649 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1650 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1651 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**. 1652 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1653 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1654 * Return 1655 * 0 on success, or a negative error in case of failure. 1656 * 1657 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1658 * Description 1659 * Grow or shrink the room for data in the packet associated to 1660 * *skb* by *len_diff*, and according to the selected *mode*. 1661 * 1662 * By default, the helper will reset any offloaded checksum 1663 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1664 * by the following flag: 1665 * 1666 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1667 * checksum data of the skb to CHECKSUM_NONE. 1668 * 1669 * There are two supported modes at this time: 1670 * 1671 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1672 * (room space is added or removed below the layer 2 header). 1673 * 1674 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1675 * (room space is added or removed below the layer 3 header). 1676 * 1677 * The following flags are supported at this time: 1678 * 1679 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1680 * Adjusting mss in this way is not allowed for datagrams. 1681 * 1682 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1683 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1684 * Any new space is reserved to hold a tunnel header. 1685 * Configure skb offsets and other fields accordingly. 1686 * 1687 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1688 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1689 * Use with ENCAP_L3 flags to further specify the tunnel type. 1690 * 1691 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1692 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1693 * type; *len* is the length of the inner MAC header. 1694 * 1695 * A call to this helper is susceptible to change the underlying 1696 * packet buffer. Therefore, at load time, all checks on pointers 1697 * previously done by the verifier are invalidated and must be 1698 * performed again, if the helper is used in combination with 1699 * direct packet access. 1700 * Return 1701 * 0 on success, or a negative error in case of failure. 1702 * 1703 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1704 * Description 1705 * Redirect the packet to the endpoint referenced by *map* at 1706 * index *key*. Depending on its type, this *map* can contain 1707 * references to net devices (for forwarding packets through other 1708 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1709 * but this is only implemented for native XDP (with driver 1710 * support) as of this writing). 1711 * 1712 * The lower two bits of *flags* are used as the return code if 1713 * the map lookup fails. This is so that the return value can be 1714 * one of the XDP program return codes up to **XDP_TX**, as chosen 1715 * by the caller. Any higher bits in the *flags* argument must be 1716 * unset. 1717 * 1718 * See also **bpf_redirect**\ (), which only supports redirecting 1719 * to an ifindex, but doesn't require a map to do so. 1720 * Return 1721 * **XDP_REDIRECT** on success, or the value of the two lower bits 1722 * of the *flags* argument on error. 1723 * 1724 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1725 * Description 1726 * Redirect the packet to the socket referenced by *map* (of type 1727 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1728 * egress interfaces can be used for redirection. The 1729 * **BPF_F_INGRESS** value in *flags* is used to make the 1730 * distinction (ingress path is selected if the flag is present, 1731 * egress path otherwise). This is the only flag supported for now. 1732 * Return 1733 * **SK_PASS** on success, or **SK_DROP** on error. 1734 * 1735 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1736 * Description 1737 * Add an entry to, or update a *map* referencing sockets. The 1738 * *skops* is used as a new value for the entry associated to 1739 * *key*. *flags* is one of: 1740 * 1741 * **BPF_NOEXIST** 1742 * The entry for *key* must not exist in the map. 1743 * **BPF_EXIST** 1744 * The entry for *key* must already exist in the map. 1745 * **BPF_ANY** 1746 * No condition on the existence of the entry for *key*. 1747 * 1748 * If the *map* has eBPF programs (parser and verdict), those will 1749 * be inherited by the socket being added. If the socket is 1750 * already attached to eBPF programs, this results in an error. 1751 * Return 1752 * 0 on success, or a negative error in case of failure. 1753 * 1754 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1755 * Description 1756 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1757 * *delta* (which can be positive or negative). Note that this 1758 * operation modifies the address stored in *xdp_md*\ **->data**, 1759 * so the latter must be loaded only after the helper has been 1760 * called. 1761 * 1762 * The use of *xdp_md*\ **->data_meta** is optional and programs 1763 * are not required to use it. The rationale is that when the 1764 * packet is processed with XDP (e.g. as DoS filter), it is 1765 * possible to push further meta data along with it before passing 1766 * to the stack, and to give the guarantee that an ingress eBPF 1767 * program attached as a TC classifier on the same device can pick 1768 * this up for further post-processing. Since TC works with socket 1769 * buffers, it remains possible to set from XDP the **mark** or 1770 * **priority** pointers, or other pointers for the socket buffer. 1771 * Having this scratch space generic and programmable allows for 1772 * more flexibility as the user is free to store whatever meta 1773 * data they need. 1774 * 1775 * A call to this helper is susceptible to change the underlying 1776 * packet buffer. Therefore, at load time, all checks on pointers 1777 * previously done by the verifier are invalidated and must be 1778 * performed again, if the helper is used in combination with 1779 * direct packet access. 1780 * Return 1781 * 0 on success, or a negative error in case of failure. 1782 * 1783 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 1784 * Description 1785 * Read the value of a perf event counter, and store it into *buf* 1786 * of size *buf_size*. This helper relies on a *map* of type 1787 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1788 * counter is selected when *map* is updated with perf event file 1789 * descriptors. The *map* is an array whose size is the number of 1790 * available CPUs, and each cell contains a value relative to one 1791 * CPU. The value to retrieve is indicated by *flags*, that 1792 * contains the index of the CPU to look up, masked with 1793 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1794 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1795 * current CPU should be retrieved. 1796 * 1797 * This helper behaves in a way close to 1798 * **bpf_perf_event_read**\ () helper, save that instead of 1799 * just returning the value observed, it fills the *buf* 1800 * structure. This allows for additional data to be retrieved: in 1801 * particular, the enabled and running times (in *buf*\ 1802 * **->enabled** and *buf*\ **->running**, respectively) are 1803 * copied. In general, **bpf_perf_event_read_value**\ () is 1804 * recommended over **bpf_perf_event_read**\ (), which has some 1805 * ABI issues and provides fewer functionalities. 1806 * 1807 * These values are interesting, because hardware PMU (Performance 1808 * Monitoring Unit) counters are limited resources. When there are 1809 * more PMU based perf events opened than available counters, 1810 * kernel will multiplex these events so each event gets certain 1811 * percentage (but not all) of the PMU time. In case that 1812 * multiplexing happens, the number of samples or counter value 1813 * will not reflect the case compared to when no multiplexing 1814 * occurs. This makes comparison between different runs difficult. 1815 * Typically, the counter value should be normalized before 1816 * comparing to other experiments. The usual normalization is done 1817 * as follows. 1818 * 1819 * :: 1820 * 1821 * normalized_counter = counter * t_enabled / t_running 1822 * 1823 * Where t_enabled is the time enabled for event and t_running is 1824 * the time running for event since last normalization. The 1825 * enabled and running times are accumulated since the perf event 1826 * open. To achieve scaling factor between two invocations of an 1827 * eBPF program, users can use CPU id as the key (which is 1828 * typical for perf array usage model) to remember the previous 1829 * value and do the calculation inside the eBPF program. 1830 * Return 1831 * 0 on success, or a negative error in case of failure. 1832 * 1833 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 1834 * Description 1835 * For en eBPF program attached to a perf event, retrieve the 1836 * value of the event counter associated to *ctx* and store it in 1837 * the structure pointed by *buf* and of size *buf_size*. Enabled 1838 * and running times are also stored in the structure (see 1839 * description of helper **bpf_perf_event_read_value**\ () for 1840 * more details). 1841 * Return 1842 * 0 on success, or a negative error in case of failure. 1843 * 1844 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 1845 * Description 1846 * Emulate a call to **getsockopt()** on the socket associated to 1847 * *bpf_socket*, which must be a full socket. The *level* at 1848 * which the option resides and the name *optname* of the option 1849 * must be specified, see **getsockopt(2)** for more information. 1850 * The retrieved value is stored in the structure pointed by 1851 * *opval* and of length *optlen*. 1852 * 1853 * *bpf_socket* should be one of the following: 1854 * 1855 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1856 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1857 * and **BPF_CGROUP_INET6_CONNECT**. 1858 * 1859 * This helper actually implements a subset of **getsockopt()**. 1860 * It supports the following *level*\ s: 1861 * 1862 * * **IPPROTO_TCP**, which supports *optname* 1863 * **TCP_CONGESTION**. 1864 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1865 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1866 * Return 1867 * 0 on success, or a negative error in case of failure. 1868 * 1869 * long bpf_override_return(struct pt_regs *regs, u64 rc) 1870 * Description 1871 * Used for error injection, this helper uses kprobes to override 1872 * the return value of the probed function, and to set it to *rc*. 1873 * The first argument is the context *regs* on which the kprobe 1874 * works. 1875 * 1876 * This helper works by setting the PC (program counter) 1877 * to an override function which is run in place of the original 1878 * probed function. This means the probed function is not run at 1879 * all. The replacement function just returns with the required 1880 * value. 1881 * 1882 * This helper has security implications, and thus is subject to 1883 * restrictions. It is only available if the kernel was compiled 1884 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1885 * option, and in this case it only works on functions tagged with 1886 * **ALLOW_ERROR_INJECTION** in the kernel code. 1887 * 1888 * Also, the helper is only available for the architectures having 1889 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1890 * x86 architecture is the only one to support this feature. 1891 * Return 1892 * 0 1893 * 1894 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 1895 * Description 1896 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1897 * for the full TCP socket associated to *bpf_sock_ops* to 1898 * *argval*. 1899 * 1900 * The primary use of this field is to determine if there should 1901 * be calls to eBPF programs of type 1902 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1903 * code. A program of the same type can change its value, per 1904 * connection and as necessary, when the connection is 1905 * established. This field is directly accessible for reading, but 1906 * this helper must be used for updates in order to return an 1907 * error if an eBPF program tries to set a callback that is not 1908 * supported in the current kernel. 1909 * 1910 * *argval* is a flag array which can combine these flags: 1911 * 1912 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1913 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1914 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1915 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1916 * 1917 * Therefore, this function can be used to clear a callback flag by 1918 * setting the appropriate bit to zero. e.g. to disable the RTO 1919 * callback: 1920 * 1921 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1922 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1923 * 1924 * Here are some examples of where one could call such eBPF 1925 * program: 1926 * 1927 * * When RTO fires. 1928 * * When a packet is retransmitted. 1929 * * When the connection terminates. 1930 * * When a packet is sent. 1931 * * When a packet is received. 1932 * Return 1933 * Code **-EINVAL** if the socket is not a full TCP socket; 1934 * otherwise, a positive number containing the bits that could not 1935 * be set is returned (which comes down to 0 if all bits were set 1936 * as required). 1937 * 1938 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1939 * Description 1940 * This helper is used in programs implementing policies at the 1941 * socket level. If the message *msg* is allowed to pass (i.e. if 1942 * the verdict eBPF program returns **SK_PASS**), redirect it to 1943 * the socket referenced by *map* (of type 1944 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1945 * egress interfaces can be used for redirection. The 1946 * **BPF_F_INGRESS** value in *flags* is used to make the 1947 * distinction (ingress path is selected if the flag is present, 1948 * egress path otherwise). This is the only flag supported for now. 1949 * Return 1950 * **SK_PASS** on success, or **SK_DROP** on error. 1951 * 1952 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1953 * Description 1954 * For socket policies, apply the verdict of the eBPF program to 1955 * the next *bytes* (number of bytes) of message *msg*. 1956 * 1957 * For example, this helper can be used in the following cases: 1958 * 1959 * * A single **sendmsg**\ () or **sendfile**\ () system call 1960 * contains multiple logical messages that the eBPF program is 1961 * supposed to read and for which it should apply a verdict. 1962 * * An eBPF program only cares to read the first *bytes* of a 1963 * *msg*. If the message has a large payload, then setting up 1964 * and calling the eBPF program repeatedly for all bytes, even 1965 * though the verdict is already known, would create unnecessary 1966 * overhead. 1967 * 1968 * When called from within an eBPF program, the helper sets a 1969 * counter internal to the BPF infrastructure, that is used to 1970 * apply the last verdict to the next *bytes*. If *bytes* is 1971 * smaller than the current data being processed from a 1972 * **sendmsg**\ () or **sendfile**\ () system call, the first 1973 * *bytes* will be sent and the eBPF program will be re-run with 1974 * the pointer for start of data pointing to byte number *bytes* 1975 * **+ 1**. If *bytes* is larger than the current data being 1976 * processed, then the eBPF verdict will be applied to multiple 1977 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1978 * consumed. 1979 * 1980 * Note that if a socket closes with the internal counter holding 1981 * a non-zero value, this is not a problem because data is not 1982 * being buffered for *bytes* and is sent as it is received. 1983 * Return 1984 * 0 1985 * 1986 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1987 * Description 1988 * For socket policies, prevent the execution of the verdict eBPF 1989 * program for message *msg* until *bytes* (byte number) have been 1990 * accumulated. 1991 * 1992 * This can be used when one needs a specific number of bytes 1993 * before a verdict can be assigned, even if the data spans 1994 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1995 * case would be a user calling **sendmsg**\ () repeatedly with 1996 * 1-byte long message segments. Obviously, this is bad for 1997 * performance, but it is still valid. If the eBPF program needs 1998 * *bytes* bytes to validate a header, this helper can be used to 1999 * prevent the eBPF program to be called again until *bytes* have 2000 * been accumulated. 2001 * Return 2002 * 0 2003 * 2004 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2005 * Description 2006 * For socket policies, pull in non-linear data from user space 2007 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2008 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2009 * respectively. 2010 * 2011 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2012 * *msg* it can only parse data that the (**data**, **data_end**) 2013 * pointers have already consumed. For **sendmsg**\ () hooks this 2014 * is likely the first scatterlist element. But for calls relying 2015 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2016 * be the range (**0**, **0**) because the data is shared with 2017 * user space and by default the objective is to avoid allowing 2018 * user space to modify data while (or after) eBPF verdict is 2019 * being decided. This helper can be used to pull in data and to 2020 * set the start and end pointer to given values. Data will be 2021 * copied if necessary (i.e. if data was not linear and if start 2022 * and end pointers do not point to the same chunk). 2023 * 2024 * A call to this helper is susceptible to change the underlying 2025 * packet buffer. Therefore, at load time, all checks on pointers 2026 * previously done by the verifier are invalidated and must be 2027 * performed again, if the helper is used in combination with 2028 * direct packet access. 2029 * 2030 * All values for *flags* are reserved for future usage, and must 2031 * be left at zero. 2032 * Return 2033 * 0 on success, or a negative error in case of failure. 2034 * 2035 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2036 * Description 2037 * Bind the socket associated to *ctx* to the address pointed by 2038 * *addr*, of length *addr_len*. This allows for making outgoing 2039 * connection from the desired IP address, which can be useful for 2040 * example when all processes inside a cgroup should use one 2041 * single IP address on a host that has multiple IP configured. 2042 * 2043 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2044 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2045 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2046 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2047 * behavior and lets the kernel efficiently pick up an unused 2048 * port as long as 4-tuple is unique. Passing non-zero port might 2049 * lead to degraded performance. 2050 * Return 2051 * 0 on success, or a negative error in case of failure. 2052 * 2053 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2054 * Description 2055 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2056 * possible to both shrink and grow the packet tail. 2057 * Shrink done via *delta* being a negative integer. 2058 * 2059 * A call to this helper is susceptible to change the underlying 2060 * packet buffer. Therefore, at load time, all checks on pointers 2061 * previously done by the verifier are invalidated and must be 2062 * performed again, if the helper is used in combination with 2063 * direct packet access. 2064 * Return 2065 * 0 on success, or a negative error in case of failure. 2066 * 2067 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2068 * Description 2069 * Retrieve the XFRM state (IP transform framework, see also 2070 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2071 * 2072 * The retrieved value is stored in the **struct bpf_xfrm_state** 2073 * pointed by *xfrm_state* and of length *size*. 2074 * 2075 * All values for *flags* are reserved for future usage, and must 2076 * be left at zero. 2077 * 2078 * This helper is available only if the kernel was compiled with 2079 * **CONFIG_XFRM** configuration option. 2080 * Return 2081 * 0 on success, or a negative error in case of failure. 2082 * 2083 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2084 * Description 2085 * Return a user or a kernel stack in bpf program provided buffer. 2086 * To achieve this, the helper needs *ctx*, which is a pointer 2087 * to the context on which the tracing program is executed. 2088 * To store the stacktrace, the bpf program provides *buf* with 2089 * a nonnegative *size*. 2090 * 2091 * The last argument, *flags*, holds the number of stack frames to 2092 * skip (from 0 to 255), masked with 2093 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2094 * the following flags: 2095 * 2096 * **BPF_F_USER_STACK** 2097 * Collect a user space stack instead of a kernel stack. 2098 * **BPF_F_USER_BUILD_ID** 2099 * Collect buildid+offset instead of ips for user stack, 2100 * only valid if **BPF_F_USER_STACK** is also specified. 2101 * 2102 * **bpf_get_stack**\ () can collect up to 2103 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2104 * to sufficient large buffer size. Note that 2105 * this limit can be controlled with the **sysctl** program, and 2106 * that it should be manually increased in order to profile long 2107 * user stacks (such as stacks for Java programs). To do so, use: 2108 * 2109 * :: 2110 * 2111 * # sysctl kernel.perf_event_max_stack=<new value> 2112 * Return 2113 * A non-negative value equal to or less than *size* on success, 2114 * or a negative error in case of failure. 2115 * 2116 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2117 * Description 2118 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2119 * it provides an easy way to load *len* bytes from *offset* 2120 * from the packet associated to *skb*, into the buffer pointed 2121 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2122 * a fifth argument *start_header* exists in order to select a 2123 * base offset to start from. *start_header* can be one of: 2124 * 2125 * **BPF_HDR_START_MAC** 2126 * Base offset to load data from is *skb*'s mac header. 2127 * **BPF_HDR_START_NET** 2128 * Base offset to load data from is *skb*'s network header. 2129 * 2130 * In general, "direct packet access" is the preferred method to 2131 * access packet data, however, this helper is in particular useful 2132 * in socket filters where *skb*\ **->data** does not always point 2133 * to the start of the mac header and where "direct packet access" 2134 * is not available. 2135 * Return 2136 * 0 on success, or a negative error in case of failure. 2137 * 2138 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2139 * Description 2140 * Do FIB lookup in kernel tables using parameters in *params*. 2141 * If lookup is successful and result shows packet is to be 2142 * forwarded, the neighbor tables are searched for the nexthop. 2143 * If successful (ie., FIB lookup shows forwarding and nexthop 2144 * is resolved), the nexthop address is returned in ipv4_dst 2145 * or ipv6_dst based on family, smac is set to mac address of 2146 * egress device, dmac is set to nexthop mac address, rt_metric 2147 * is set to metric from route (IPv4/IPv6 only), and ifindex 2148 * is set to the device index of the nexthop from the FIB lookup. 2149 * 2150 * *plen* argument is the size of the passed in struct. 2151 * *flags* argument can be a combination of one or more of the 2152 * following values: 2153 * 2154 * **BPF_FIB_LOOKUP_DIRECT** 2155 * Do a direct table lookup vs full lookup using FIB 2156 * rules. 2157 * **BPF_FIB_LOOKUP_OUTPUT** 2158 * Perform lookup from an egress perspective (default is 2159 * ingress). 2160 * 2161 * *ctx* is either **struct xdp_md** for XDP programs or 2162 * **struct sk_buff** tc cls_act programs. 2163 * Return 2164 * * < 0 if any input argument is invalid 2165 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2166 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2167 * packet is not forwarded or needs assist from full stack 2168 * 2169 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2170 * Description 2171 * Add an entry to, or update a sockhash *map* referencing sockets. 2172 * The *skops* is used as a new value for the entry associated to 2173 * *key*. *flags* is one of: 2174 * 2175 * **BPF_NOEXIST** 2176 * The entry for *key* must not exist in the map. 2177 * **BPF_EXIST** 2178 * The entry for *key* must already exist in the map. 2179 * **BPF_ANY** 2180 * No condition on the existence of the entry for *key*. 2181 * 2182 * If the *map* has eBPF programs (parser and verdict), those will 2183 * be inherited by the socket being added. If the socket is 2184 * already attached to eBPF programs, this results in an error. 2185 * Return 2186 * 0 on success, or a negative error in case of failure. 2187 * 2188 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 2189 * Description 2190 * This helper is used in programs implementing policies at the 2191 * socket level. If the message *msg* is allowed to pass (i.e. if 2192 * the verdict eBPF program returns **SK_PASS**), redirect it to 2193 * the socket referenced by *map* (of type 2194 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2195 * egress interfaces can be used for redirection. The 2196 * **BPF_F_INGRESS** value in *flags* is used to make the 2197 * distinction (ingress path is selected if the flag is present, 2198 * egress path otherwise). This is the only flag supported for now. 2199 * Return 2200 * **SK_PASS** on success, or **SK_DROP** on error. 2201 * 2202 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 2203 * Description 2204 * This helper is used in programs implementing policies at the 2205 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 2206 * if the verdeict eBPF program returns **SK_PASS**), redirect it 2207 * to the socket referenced by *map* (of type 2208 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 2209 * egress interfaces can be used for redirection. The 2210 * **BPF_F_INGRESS** value in *flags* is used to make the 2211 * distinction (ingress path is selected if the flag is present, 2212 * egress otherwise). This is the only flag supported for now. 2213 * Return 2214 * **SK_PASS** on success, or **SK_DROP** on error. 2215 * 2216 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2217 * Description 2218 * Encapsulate the packet associated to *skb* within a Layer 3 2219 * protocol header. This header is provided in the buffer at 2220 * address *hdr*, with *len* its size in bytes. *type* indicates 2221 * the protocol of the header and can be one of: 2222 * 2223 * **BPF_LWT_ENCAP_SEG6** 2224 * IPv6 encapsulation with Segment Routing Header 2225 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2226 * the IPv6 header is computed by the kernel. 2227 * **BPF_LWT_ENCAP_SEG6_INLINE** 2228 * Only works if *skb* contains an IPv6 packet. Insert a 2229 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2230 * the IPv6 header. 2231 * **BPF_LWT_ENCAP_IP** 2232 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 2233 * must be IPv4 or IPv6, followed by zero or more 2234 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 2235 * total bytes in all prepended headers. Please note that 2236 * if **skb_is_gso**\ (*skb*) is true, no more than two 2237 * headers can be prepended, and the inner header, if 2238 * present, should be either GRE or UDP/GUE. 2239 * 2240 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 2241 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 2242 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 2243 * **BPF_PROG_TYPE_LWT_XMIT**. 2244 * 2245 * A call to this helper is susceptible to change the underlying 2246 * packet buffer. Therefore, at load time, all checks on pointers 2247 * previously done by the verifier are invalidated and must be 2248 * performed again, if the helper is used in combination with 2249 * direct packet access. 2250 * Return 2251 * 0 on success, or a negative error in case of failure. 2252 * 2253 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2254 * Description 2255 * Store *len* bytes from address *from* into the packet 2256 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2257 * inside the outermost IPv6 Segment Routing Header can be 2258 * modified through this helper. 2259 * 2260 * A call to this helper is susceptible to change the underlying 2261 * packet buffer. Therefore, at load time, all checks on pointers 2262 * previously done by the verifier are invalidated and must be 2263 * performed again, if the helper is used in combination with 2264 * direct packet access. 2265 * Return 2266 * 0 on success, or a negative error in case of failure. 2267 * 2268 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2269 * Description 2270 * Adjust the size allocated to TLVs in the outermost IPv6 2271 * Segment Routing Header contained in the packet associated to 2272 * *skb*, at position *offset* by *delta* bytes. Only offsets 2273 * after the segments are accepted. *delta* can be as well 2274 * positive (growing) as negative (shrinking). 2275 * 2276 * A call to this helper is susceptible to change the underlying 2277 * packet buffer. Therefore, at load time, all checks on pointers 2278 * previously done by the verifier are invalidated and must be 2279 * performed again, if the helper is used in combination with 2280 * direct packet access. 2281 * Return 2282 * 0 on success, or a negative error in case of failure. 2283 * 2284 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2285 * Description 2286 * Apply an IPv6 Segment Routing action of type *action* to the 2287 * packet associated to *skb*. Each action takes a parameter 2288 * contained at address *param*, and of length *param_len* bytes. 2289 * *action* can be one of: 2290 * 2291 * **SEG6_LOCAL_ACTION_END_X** 2292 * End.X action: Endpoint with Layer-3 cross-connect. 2293 * Type of *param*: **struct in6_addr**. 2294 * **SEG6_LOCAL_ACTION_END_T** 2295 * End.T action: Endpoint with specific IPv6 table lookup. 2296 * Type of *param*: **int**. 2297 * **SEG6_LOCAL_ACTION_END_B6** 2298 * End.B6 action: Endpoint bound to an SRv6 policy. 2299 * Type of *param*: **struct ipv6_sr_hdr**. 2300 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2301 * End.B6.Encap action: Endpoint bound to an SRv6 2302 * encapsulation policy. 2303 * Type of *param*: **struct ipv6_sr_hdr**. 2304 * 2305 * A call to this helper is susceptible to change the underlying 2306 * packet buffer. Therefore, at load time, all checks on pointers 2307 * previously done by the verifier are invalidated and must be 2308 * performed again, if the helper is used in combination with 2309 * direct packet access. 2310 * Return 2311 * 0 on success, or a negative error in case of failure. 2312 * 2313 * long bpf_rc_repeat(void *ctx) 2314 * Description 2315 * This helper is used in programs implementing IR decoding, to 2316 * report a successfully decoded repeat key message. This delays 2317 * the generation of a key up event for previously generated 2318 * key down event. 2319 * 2320 * Some IR protocols like NEC have a special IR message for 2321 * repeating last button, for when a button is held down. 2322 * 2323 * The *ctx* should point to the lirc sample as passed into 2324 * the program. 2325 * 2326 * This helper is only available is the kernel was compiled with 2327 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2328 * "**y**". 2329 * Return 2330 * 0 2331 * 2332 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2333 * Description 2334 * This helper is used in programs implementing IR decoding, to 2335 * report a successfully decoded key press with *scancode*, 2336 * *toggle* value in the given *protocol*. The scancode will be 2337 * translated to a keycode using the rc keymap, and reported as 2338 * an input key down event. After a period a key up event is 2339 * generated. This period can be extended by calling either 2340 * **bpf_rc_keydown**\ () again with the same values, or calling 2341 * **bpf_rc_repeat**\ (). 2342 * 2343 * Some protocols include a toggle bit, in case the button was 2344 * released and pressed again between consecutive scancodes. 2345 * 2346 * The *ctx* should point to the lirc sample as passed into 2347 * the program. 2348 * 2349 * The *protocol* is the decoded protocol number (see 2350 * **enum rc_proto** for some predefined values). 2351 * 2352 * This helper is only available is the kernel was compiled with 2353 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2354 * "**y**". 2355 * Return 2356 * 0 2357 * 2358 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2359 * Description 2360 * Return the cgroup v2 id of the socket associated with the *skb*. 2361 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2362 * helper for cgroup v1 by providing a tag resp. identifier that 2363 * can be matched on or used for map lookups e.g. to implement 2364 * policy. The cgroup v2 id of a given path in the hierarchy is 2365 * exposed in user space through the f_handle API in order to get 2366 * to the same 64-bit id. 2367 * 2368 * This helper can be used on TC egress path, but not on ingress, 2369 * and is available only if the kernel was compiled with the 2370 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2371 * Return 2372 * The id is returned or 0 in case the id could not be retrieved. 2373 * 2374 * u64 bpf_get_current_cgroup_id(void) 2375 * Return 2376 * A 64-bit integer containing the current cgroup id based 2377 * on the cgroup within which the current task is running. 2378 * 2379 * void *bpf_get_local_storage(void *map, u64 flags) 2380 * Description 2381 * Get the pointer to the local storage area. 2382 * The type and the size of the local storage is defined 2383 * by the *map* argument. 2384 * The *flags* meaning is specific for each map type, 2385 * and has to be 0 for cgroup local storage. 2386 * 2387 * Depending on the BPF program type, a local storage area 2388 * can be shared between multiple instances of the BPF program, 2389 * running simultaneously. 2390 * 2391 * A user should care about the synchronization by himself. 2392 * For example, by using the **BPF_STX_XADD** instruction to alter 2393 * the shared data. 2394 * Return 2395 * A pointer to the local storage area. 2396 * 2397 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 2398 * Description 2399 * Select a **SO_REUSEPORT** socket from a 2400 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 2401 * It checks the selected socket is matching the incoming 2402 * request in the socket buffer. 2403 * Return 2404 * 0 on success, or a negative error in case of failure. 2405 * 2406 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 2407 * Description 2408 * Return id of cgroup v2 that is ancestor of cgroup associated 2409 * with the *skb* at the *ancestor_level*. The root cgroup is at 2410 * *ancestor_level* zero and each step down the hierarchy 2411 * increments the level. If *ancestor_level* == level of cgroup 2412 * associated with *skb*, then return value will be same as that 2413 * of **bpf_skb_cgroup_id**\ (). 2414 * 2415 * The helper is useful to implement policies based on cgroups 2416 * that are upper in hierarchy than immediate cgroup associated 2417 * with *skb*. 2418 * 2419 * The format of returned id and helper limitations are same as in 2420 * **bpf_skb_cgroup_id**\ (). 2421 * Return 2422 * The id is returned or 0 in case the id could not be retrieved. 2423 * 2424 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2425 * Description 2426 * Look for TCP socket matching *tuple*, optionally in a child 2427 * network namespace *netns*. The return value must be checked, 2428 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2429 * 2430 * The *ctx* should point to the context of the program, such as 2431 * the skb or socket (depending on the hook in use). This is used 2432 * to determine the base network namespace for the lookup. 2433 * 2434 * *tuple_size* must be one of: 2435 * 2436 * **sizeof**\ (*tuple*\ **->ipv4**) 2437 * Look for an IPv4 socket. 2438 * **sizeof**\ (*tuple*\ **->ipv6**) 2439 * Look for an IPv6 socket. 2440 * 2441 * If the *netns* is a negative signed 32-bit integer, then the 2442 * socket lookup table in the netns associated with the *ctx* 2443 * will be used. For the TC hooks, this is the netns of the device 2444 * in the skb. For socket hooks, this is the netns of the socket. 2445 * If *netns* is any other signed 32-bit value greater than or 2446 * equal to zero then it specifies the ID of the netns relative to 2447 * the netns associated with the *ctx*. *netns* values beyond the 2448 * range of 32-bit integers are reserved for future use. 2449 * 2450 * All values for *flags* are reserved for future usage, and must 2451 * be left at zero. 2452 * 2453 * This helper is available only if the kernel was compiled with 2454 * **CONFIG_NET** configuration option. 2455 * Return 2456 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2457 * For sockets with reuseport option, the **struct bpf_sock** 2458 * result is from *reuse*\ **->socks**\ [] using the hash of the 2459 * tuple. 2460 * 2461 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2462 * Description 2463 * Look for UDP socket matching *tuple*, optionally in a child 2464 * network namespace *netns*. The return value must be checked, 2465 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2466 * 2467 * The *ctx* should point to the context of the program, such as 2468 * the skb or socket (depending on the hook in use). This is used 2469 * to determine the base network namespace for the lookup. 2470 * 2471 * *tuple_size* must be one of: 2472 * 2473 * **sizeof**\ (*tuple*\ **->ipv4**) 2474 * Look for an IPv4 socket. 2475 * **sizeof**\ (*tuple*\ **->ipv6**) 2476 * Look for an IPv6 socket. 2477 * 2478 * If the *netns* is a negative signed 32-bit integer, then the 2479 * socket lookup table in the netns associated with the *ctx* 2480 * will be used. For the TC hooks, this is the netns of the device 2481 * in the skb. For socket hooks, this is the netns of the socket. 2482 * If *netns* is any other signed 32-bit value greater than or 2483 * equal to zero then it specifies the ID of the netns relative to 2484 * the netns associated with the *ctx*. *netns* values beyond the 2485 * range of 32-bit integers are reserved for future use. 2486 * 2487 * All values for *flags* are reserved for future usage, and must 2488 * be left at zero. 2489 * 2490 * This helper is available only if the kernel was compiled with 2491 * **CONFIG_NET** configuration option. 2492 * Return 2493 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2494 * For sockets with reuseport option, the **struct bpf_sock** 2495 * result is from *reuse*\ **->socks**\ [] using the hash of the 2496 * tuple. 2497 * 2498 * long bpf_sk_release(struct bpf_sock *sock) 2499 * Description 2500 * Release the reference held by *sock*. *sock* must be a 2501 * non-**NULL** pointer that was returned from 2502 * **bpf_sk_lookup_xxx**\ (). 2503 * Return 2504 * 0 on success, or a negative error in case of failure. 2505 * 2506 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 2507 * Description 2508 * Push an element *value* in *map*. *flags* is one of: 2509 * 2510 * **BPF_EXIST** 2511 * If the queue/stack is full, the oldest element is 2512 * removed to make room for this. 2513 * Return 2514 * 0 on success, or a negative error in case of failure. 2515 * 2516 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 2517 * Description 2518 * Pop an element from *map*. 2519 * Return 2520 * 0 on success, or a negative error in case of failure. 2521 * 2522 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 2523 * Description 2524 * Get an element from *map* without removing it. 2525 * Return 2526 * 0 on success, or a negative error in case of failure. 2527 * 2528 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2529 * Description 2530 * For socket policies, insert *len* bytes into *msg* at offset 2531 * *start*. 2532 * 2533 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2534 * *msg* it may want to insert metadata or options into the *msg*. 2535 * This can later be read and used by any of the lower layer BPF 2536 * hooks. 2537 * 2538 * This helper may fail if under memory pressure (a malloc 2539 * fails) in these cases BPF programs will get an appropriate 2540 * error and BPF programs will need to handle them. 2541 * Return 2542 * 0 on success, or a negative error in case of failure. 2543 * 2544 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 2545 * Description 2546 * Will remove *len* bytes from a *msg* starting at byte *start*. 2547 * This may result in **ENOMEM** errors under certain situations if 2548 * an allocation and copy are required due to a full ring buffer. 2549 * However, the helper will try to avoid doing the allocation 2550 * if possible. Other errors can occur if input parameters are 2551 * invalid either due to *start* byte not being valid part of *msg* 2552 * payload and/or *pop* value being to large. 2553 * Return 2554 * 0 on success, or a negative error in case of failure. 2555 * 2556 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 2557 * Description 2558 * This helper is used in programs implementing IR decoding, to 2559 * report a successfully decoded pointer movement. 2560 * 2561 * The *ctx* should point to the lirc sample as passed into 2562 * the program. 2563 * 2564 * This helper is only available is the kernel was compiled with 2565 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2566 * "**y**". 2567 * Return 2568 * 0 2569 * 2570 * long bpf_spin_lock(struct bpf_spin_lock *lock) 2571 * Description 2572 * Acquire a spinlock represented by the pointer *lock*, which is 2573 * stored as part of a value of a map. Taking the lock allows to 2574 * safely update the rest of the fields in that value. The 2575 * spinlock can (and must) later be released with a call to 2576 * **bpf_spin_unlock**\ (\ *lock*\ ). 2577 * 2578 * Spinlocks in BPF programs come with a number of restrictions 2579 * and constraints: 2580 * 2581 * * **bpf_spin_lock** objects are only allowed inside maps of 2582 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2583 * list could be extended in the future). 2584 * * BTF description of the map is mandatory. 2585 * * The BPF program can take ONE lock at a time, since taking two 2586 * or more could cause dead locks. 2587 * * Only one **struct bpf_spin_lock** is allowed per map element. 2588 * * When the lock is taken, calls (either BPF to BPF or helpers) 2589 * are not allowed. 2590 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2591 * allowed inside a spinlock-ed region. 2592 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2593 * the lock, on all execution paths, before it returns. 2594 * * The BPF program can access **struct bpf_spin_lock** only via 2595 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2596 * helpers. Loading or storing data into the **struct 2597 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2598 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2599 * of the map value must be a struct and have **struct 2600 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2601 * Nested lock inside another struct is not allowed. 2602 * * The **struct bpf_spin_lock** *lock* field in a map value must 2603 * be aligned on a multiple of 4 bytes in that value. 2604 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2605 * the **bpf_spin_lock** field to user space. 2606 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2607 * a BPF program, do not update the **bpf_spin_lock** field. 2608 * * **bpf_spin_lock** cannot be on the stack or inside a 2609 * networking packet (it can only be inside of a map values). 2610 * * **bpf_spin_lock** is available to root only. 2611 * * Tracing programs and socket filter programs cannot use 2612 * **bpf_spin_lock**\ () due to insufficient preemption checks 2613 * (but this may change in the future). 2614 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2615 * Return 2616 * 0 2617 * 2618 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 2619 * Description 2620 * Release the *lock* previously locked by a call to 2621 * **bpf_spin_lock**\ (\ *lock*\ ). 2622 * Return 2623 * 0 2624 * 2625 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 2626 * Description 2627 * This helper gets a **struct bpf_sock** pointer such 2628 * that all the fields in this **bpf_sock** can be accessed. 2629 * Return 2630 * A **struct bpf_sock** pointer on success, or **NULL** in 2631 * case of failure. 2632 * 2633 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2634 * Description 2635 * This helper gets a **struct bpf_tcp_sock** pointer from a 2636 * **struct bpf_sock** pointer. 2637 * Return 2638 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2639 * case of failure. 2640 * 2641 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 2642 * Description 2643 * Set ECN (Explicit Congestion Notification) field of IP header 2644 * to **CE** (Congestion Encountered) if current value is **ECT** 2645 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2646 * and IPv4. 2647 * Return 2648 * 1 if the **CE** flag is set (either by the current helper call 2649 * or because it was already present), 0 if it is not set. 2650 * 2651 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2652 * Description 2653 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2654 * **bpf_sk_release**\ () is unnecessary and not allowed. 2655 * Return 2656 * A **struct bpf_sock** pointer on success, or **NULL** in 2657 * case of failure. 2658 * 2659 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2660 * Description 2661 * Look for TCP socket matching *tuple*, optionally in a child 2662 * network namespace *netns*. The return value must be checked, 2663 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2664 * 2665 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2666 * that it also returns timewait or request sockets. Use 2667 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2668 * full structure. 2669 * 2670 * This helper is available only if the kernel was compiled with 2671 * **CONFIG_NET** configuration option. 2672 * Return 2673 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2674 * For sockets with reuseport option, the **struct bpf_sock** 2675 * result is from *reuse*\ **->socks**\ [] using the hash of the 2676 * tuple. 2677 * 2678 * long bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2679 * Description 2680 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2681 * the listening socket in *sk*. 2682 * 2683 * *iph* points to the start of the IPv4 or IPv6 header, while 2684 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2685 * **sizeof**\ (**struct ip6hdr**). 2686 * 2687 * *th* points to the start of the TCP header, while *th_len* 2688 * contains **sizeof**\ (**struct tcphdr**). 2689 * Return 2690 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2691 * error otherwise. 2692 * 2693 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2694 * Description 2695 * Get name of sysctl in /proc/sys/ and copy it into provided by 2696 * program buffer *buf* of size *buf_len*. 2697 * 2698 * The buffer is always NUL terminated, unless it's zero-sized. 2699 * 2700 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2701 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2702 * only (e.g. "tcp_mem"). 2703 * Return 2704 * Number of character copied (not including the trailing NUL). 2705 * 2706 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2707 * truncated name in this case). 2708 * 2709 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2710 * Description 2711 * Get current value of sysctl as it is presented in /proc/sys 2712 * (incl. newline, etc), and copy it as a string into provided 2713 * by program buffer *buf* of size *buf_len*. 2714 * 2715 * The whole value is copied, no matter what file position user 2716 * space issued e.g. sys_read at. 2717 * 2718 * The buffer is always NUL terminated, unless it's zero-sized. 2719 * Return 2720 * Number of character copied (not including the trailing NUL). 2721 * 2722 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2723 * truncated name in this case). 2724 * 2725 * **-EINVAL** if current value was unavailable, e.g. because 2726 * sysctl is uninitialized and read returns -EIO for it. 2727 * 2728 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 2729 * Description 2730 * Get new value being written by user space to sysctl (before 2731 * the actual write happens) and copy it as a string into 2732 * provided by program buffer *buf* of size *buf_len*. 2733 * 2734 * User space may write new value at file position > 0. 2735 * 2736 * The buffer is always NUL terminated, unless it's zero-sized. 2737 * Return 2738 * Number of character copied (not including the trailing NUL). 2739 * 2740 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2741 * truncated name in this case). 2742 * 2743 * **-EINVAL** if sysctl is being read. 2744 * 2745 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 2746 * Description 2747 * Override new value being written by user space to sysctl with 2748 * value provided by program in buffer *buf* of size *buf_len*. 2749 * 2750 * *buf* should contain a string in same form as provided by user 2751 * space on sysctl write. 2752 * 2753 * User space may write new value at file position > 0. To override 2754 * the whole sysctl value file position should be set to zero. 2755 * Return 2756 * 0 on success. 2757 * 2758 * **-E2BIG** if the *buf_len* is too big. 2759 * 2760 * **-EINVAL** if sysctl is being read. 2761 * 2762 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2763 * Description 2764 * Convert the initial part of the string from buffer *buf* of 2765 * size *buf_len* to a long integer according to the given base 2766 * and save the result in *res*. 2767 * 2768 * The string may begin with an arbitrary amount of white space 2769 * (as determined by **isspace**\ (3)) followed by a single 2770 * optional '**-**' sign. 2771 * 2772 * Five least significant bits of *flags* encode base, other bits 2773 * are currently unused. 2774 * 2775 * Base must be either 8, 10, 16 or 0 to detect it automatically 2776 * similar to user space **strtol**\ (3). 2777 * Return 2778 * Number of characters consumed on success. Must be positive but 2779 * no more than *buf_len*. 2780 * 2781 * **-EINVAL** if no valid digits were found or unsupported base 2782 * was provided. 2783 * 2784 * **-ERANGE** if resulting value was out of range. 2785 * 2786 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2787 * Description 2788 * Convert the initial part of the string from buffer *buf* of 2789 * size *buf_len* to an unsigned long integer according to the 2790 * given base and save the result in *res*. 2791 * 2792 * The string may begin with an arbitrary amount of white space 2793 * (as determined by **isspace**\ (3)). 2794 * 2795 * Five least significant bits of *flags* encode base, other bits 2796 * are currently unused. 2797 * 2798 * Base must be either 8, 10, 16 or 0 to detect it automatically 2799 * similar to user space **strtoul**\ (3). 2800 * Return 2801 * Number of characters consumed on success. Must be positive but 2802 * no more than *buf_len*. 2803 * 2804 * **-EINVAL** if no valid digits were found or unsupported base 2805 * was provided. 2806 * 2807 * **-ERANGE** if resulting value was out of range. 2808 * 2809 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags) 2810 * Description 2811 * Get a bpf-local-storage from a *sk*. 2812 * 2813 * Logically, it could be thought of getting the value from 2814 * a *map* with *sk* as the **key**. From this 2815 * perspective, the usage is not much different from 2816 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2817 * helper enforces the key must be a full socket and the map must 2818 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2819 * 2820 * Underneath, the value is stored locally at *sk* instead of 2821 * the *map*. The *map* is used as the bpf-local-storage 2822 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2823 * searched against all bpf-local-storages residing at *sk*. 2824 * 2825 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2826 * used such that a new bpf-local-storage will be 2827 * created if one does not exist. *value* can be used 2828 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2829 * the initial value of a bpf-local-storage. If *value* is 2830 * **NULL**, the new bpf-local-storage will be zero initialized. 2831 * Return 2832 * A bpf-local-storage pointer is returned on success. 2833 * 2834 * **NULL** if not found or there was an error in adding 2835 * a new bpf-local-storage. 2836 * 2837 * long bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk) 2838 * Description 2839 * Delete a bpf-local-storage from a *sk*. 2840 * Return 2841 * 0 on success. 2842 * 2843 * **-ENOENT** if the bpf-local-storage cannot be found. 2844 * 2845 * long bpf_send_signal(u32 sig) 2846 * Description 2847 * Send signal *sig* to the process of the current task. 2848 * The signal may be delivered to any of this process's threads. 2849 * Return 2850 * 0 on success or successfully queued. 2851 * 2852 * **-EBUSY** if work queue under nmi is full. 2853 * 2854 * **-EINVAL** if *sig* is invalid. 2855 * 2856 * **-EPERM** if no permission to send the *sig*. 2857 * 2858 * **-EAGAIN** if bpf program can try again. 2859 * 2860 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 2861 * Description 2862 * Try to issue a SYN cookie for the packet with corresponding 2863 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2864 * 2865 * *iph* points to the start of the IPv4 or IPv6 header, while 2866 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2867 * **sizeof**\ (**struct ip6hdr**). 2868 * 2869 * *th* points to the start of the TCP header, while *th_len* 2870 * contains the length of the TCP header. 2871 * Return 2872 * On success, lower 32 bits hold the generated SYN cookie in 2873 * followed by 16 bits which hold the MSS value for that cookie, 2874 * and the top 16 bits are unused. 2875 * 2876 * On failure, the returned value is one of the following: 2877 * 2878 * **-EINVAL** SYN cookie cannot be issued due to error 2879 * 2880 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2881 * 2882 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2883 * 2884 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2885 * 2886 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2887 * Description 2888 * Write raw *data* blob into a special BPF perf event held by 2889 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2890 * event must have the following attributes: **PERF_SAMPLE_RAW** 2891 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2892 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2893 * 2894 * The *flags* are used to indicate the index in *map* for which 2895 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2896 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2897 * to indicate that the index of the current CPU core should be 2898 * used. 2899 * 2900 * The value to write, of *size*, is passed through eBPF stack and 2901 * pointed by *data*. 2902 * 2903 * *ctx* is a pointer to in-kernel struct sk_buff. 2904 * 2905 * This helper is similar to **bpf_perf_event_output**\ () but 2906 * restricted to raw_tracepoint bpf programs. 2907 * Return 2908 * 0 on success, or a negative error in case of failure. 2909 * 2910 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 2911 * Description 2912 * Safely attempt to read *size* bytes from user space address 2913 * *unsafe_ptr* and store the data in *dst*. 2914 * Return 2915 * 0 on success, or a negative error in case of failure. 2916 * 2917 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 2918 * Description 2919 * Safely attempt to read *size* bytes from kernel space address 2920 * *unsafe_ptr* and store the data in *dst*. 2921 * Return 2922 * 0 on success, or a negative error in case of failure. 2923 * 2924 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 2925 * Description 2926 * Copy a NUL terminated string from an unsafe user address 2927 * *unsafe_ptr* to *dst*. The *size* should include the 2928 * terminating NUL byte. In case the string length is smaller than 2929 * *size*, the target is not padded with further NUL bytes. If the 2930 * string length is larger than *size*, just *size*-1 bytes are 2931 * copied and the last byte is set to NUL. 2932 * 2933 * On success, the length of the copied string is returned. This 2934 * makes this helper useful in tracing programs for reading 2935 * strings, and more importantly to get its length at runtime. See 2936 * the following snippet: 2937 * 2938 * :: 2939 * 2940 * SEC("kprobe/sys_open") 2941 * void bpf_sys_open(struct pt_regs *ctx) 2942 * { 2943 * char buf[PATHLEN]; // PATHLEN is defined to 256 2944 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2945 * ctx->di); 2946 * 2947 * // Consume buf, for example push it to 2948 * // userspace via bpf_perf_event_output(); we 2949 * // can use res (the string length) as event 2950 * // size, after checking its boundaries. 2951 * } 2952 * 2953 * In comparison, using **bpf_probe_read_user**\ () helper here 2954 * instead to read the string would require to estimate the length 2955 * at compile time, and would often result in copying more memory 2956 * than necessary. 2957 * 2958 * Another useful use case is when parsing individual process 2959 * arguments or individual environment variables navigating 2960 * *current*\ **->mm->arg_start** and *current*\ 2961 * **->mm->env_start**: using this helper and the return value, 2962 * one can quickly iterate at the right offset of the memory area. 2963 * Return 2964 * On success, the strictly positive length of the string, 2965 * including the trailing NUL character. On error, a negative 2966 * value. 2967 * 2968 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 2969 * Description 2970 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2971 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 2972 * Return 2973 * On success, the strictly positive length of the string, including 2974 * the trailing NUL character. On error, a negative value. 2975 * 2976 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 2977 * Description 2978 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 2979 * *rcv_nxt* is the ack_seq to be sent out. 2980 * Return 2981 * 0 on success, or a negative error in case of failure. 2982 * 2983 * long bpf_send_signal_thread(u32 sig) 2984 * Description 2985 * Send signal *sig* to the thread corresponding to the current task. 2986 * Return 2987 * 0 on success or successfully queued. 2988 * 2989 * **-EBUSY** if work queue under nmi is full. 2990 * 2991 * **-EINVAL** if *sig* is invalid. 2992 * 2993 * **-EPERM** if no permission to send the *sig*. 2994 * 2995 * **-EAGAIN** if bpf program can try again. 2996 * 2997 * u64 bpf_jiffies64(void) 2998 * Description 2999 * Obtain the 64bit jiffies 3000 * Return 3001 * The 64 bit jiffies 3002 * 3003 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3004 * Description 3005 * For an eBPF program attached to a perf event, retrieve the 3006 * branch records (**struct perf_branch_entry**) associated to *ctx* 3007 * and store it in the buffer pointed by *buf* up to size 3008 * *size* bytes. 3009 * Return 3010 * On success, number of bytes written to *buf*. On error, a 3011 * negative value. 3012 * 3013 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3014 * instead return the number of bytes required to store all the 3015 * branch entries. If this flag is set, *buf* may be NULL. 3016 * 3017 * **-EINVAL** if arguments invalid or **size** not a multiple 3018 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3019 * 3020 * **-ENOENT** if architecture does not support branch records. 3021 * 3022 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3023 * Description 3024 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3025 * *namespace* will be returned in *nsdata*. 3026 * Return 3027 * 0 on success, or one of the following in case of failure: 3028 * 3029 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3030 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3031 * 3032 * **-ENOENT** if pidns does not exists for the current task. 3033 * 3034 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3035 * Description 3036 * Write raw *data* blob into a special BPF perf event held by 3037 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3038 * event must have the following attributes: **PERF_SAMPLE_RAW** 3039 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3040 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3041 * 3042 * The *flags* are used to indicate the index in *map* for which 3043 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3044 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3045 * to indicate that the index of the current CPU core should be 3046 * used. 3047 * 3048 * The value to write, of *size*, is passed through eBPF stack and 3049 * pointed by *data*. 3050 * 3051 * *ctx* is a pointer to in-kernel struct xdp_buff. 3052 * 3053 * This helper is similar to **bpf_perf_eventoutput**\ () but 3054 * restricted to raw_tracepoint bpf programs. 3055 * Return 3056 * 0 on success, or a negative error in case of failure. 3057 * 3058 * u64 bpf_get_netns_cookie(void *ctx) 3059 * Description 3060 * Retrieve the cookie (generated by the kernel) of the network 3061 * namespace the input *ctx* is associated with. The network 3062 * namespace cookie remains stable for its lifetime and provides 3063 * a global identifier that can be assumed unique. If *ctx* is 3064 * NULL, then the helper returns the cookie for the initial 3065 * network namespace. The cookie itself is very similar to that 3066 * of **bpf_get_socket_cookie**\ () helper, but for network 3067 * namespaces instead of sockets. 3068 * Return 3069 * A 8-byte long opaque number. 3070 * 3071 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3072 * Description 3073 * Return id of cgroup v2 that is ancestor of the cgroup associated 3074 * with the current task at the *ancestor_level*. The root cgroup 3075 * is at *ancestor_level* zero and each step down the hierarchy 3076 * increments the level. If *ancestor_level* == level of cgroup 3077 * associated with the current task, then return value will be the 3078 * same as that of **bpf_get_current_cgroup_id**\ (). 3079 * 3080 * The helper is useful to implement policies based on cgroups 3081 * that are upper in hierarchy than immediate cgroup associated 3082 * with the current task. 3083 * 3084 * The format of returned id and helper limitations are same as in 3085 * **bpf_get_current_cgroup_id**\ (). 3086 * Return 3087 * The id is returned or 0 in case the id could not be retrieved. 3088 * 3089 * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags) 3090 * Description 3091 * Helper is overloaded depending on BPF program type. This 3092 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3093 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3094 * 3095 * Assign the *sk* to the *skb*. When combined with appropriate 3096 * routing configuration to receive the packet towards the socket, 3097 * will cause *skb* to be delivered to the specified socket. 3098 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3099 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3100 * interfere with successful delivery to the socket. 3101 * 3102 * This operation is only valid from TC ingress path. 3103 * 3104 * The *flags* argument must be zero. 3105 * Return 3106 * 0 on success, or a negative error in case of failure: 3107 * 3108 * **-EINVAL** if specified *flags* are not supported. 3109 * 3110 * **-ENOENT** if the socket is unavailable for assignment. 3111 * 3112 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3113 * 3114 * **-EOPNOTSUPP** if the operation is not supported, for example 3115 * a call from outside of TC ingress. 3116 * 3117 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3118 * (reuseport). 3119 * 3120 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3121 * Description 3122 * Helper is overloaded depending on BPF program type. This 3123 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3124 * 3125 * Select the *sk* as a result of a socket lookup. 3126 * 3127 * For the operation to succeed passed socket must be compatible 3128 * with the packet description provided by the *ctx* object. 3129 * 3130 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3131 * be an exact match. While IP family (**AF_INET** or 3132 * **AF_INET6**) must be compatible, that is IPv6 sockets 3133 * that are not v6-only can be selected for IPv4 packets. 3134 * 3135 * Only TCP listeners and UDP unconnected sockets can be 3136 * selected. *sk* can also be NULL to reset any previous 3137 * selection. 3138 * 3139 * *flags* argument can combination of following values: 3140 * 3141 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3142 * socket selection, potentially done by a BPF program 3143 * that ran before us. 3144 * 3145 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3146 * load-balancing within reuseport group for the socket 3147 * being selected. 3148 * 3149 * On success *ctx->sk* will point to the selected socket. 3150 * 3151 * Return 3152 * 0 on success, or a negative errno in case of failure. 3153 * 3154 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3155 * not compatible with packet family (*ctx->family*). 3156 * 3157 * * **-EEXIST** if socket has been already selected, 3158 * potentially by another program, and 3159 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3160 * 3161 * * **-EINVAL** if unsupported flags were specified. 3162 * 3163 * * **-EPROTOTYPE** if socket L4 protocol 3164 * (*sk->protocol*) doesn't match packet protocol 3165 * (*ctx->protocol*). 3166 * 3167 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3168 * state (TCP listening or UDP unconnected). 3169 * 3170 * u64 bpf_ktime_get_boot_ns(void) 3171 * Description 3172 * Return the time elapsed since system boot, in nanoseconds. 3173 * Does include the time the system was suspended. 3174 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3175 * Return 3176 * Current *ktime*. 3177 * 3178 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3179 * Description 3180 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3181 * out the format string. 3182 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3183 * the format string itself. The *data* and *data_len* are format string 3184 * arguments. The *data* are a **u64** array and corresponding format string 3185 * values are stored in the array. For strings and pointers where pointees 3186 * are accessed, only the pointer values are stored in the *data* array. 3187 * The *data_len* is the size of *data* in bytes. 3188 * 3189 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3190 * Reading kernel memory may fail due to either invalid address or 3191 * valid address but requiring a major memory fault. If reading kernel memory 3192 * fails, the string for **%s** will be an empty string, and the ip 3193 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3194 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3195 * Return 3196 * 0 on success, or a negative error in case of failure: 3197 * 3198 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3199 * by returning 1 from bpf program. 3200 * 3201 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3202 * 3203 * **-E2BIG** if *fmt* contains too many format specifiers. 3204 * 3205 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3206 * 3207 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 3208 * Description 3209 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3210 * The *m* represents the seq_file. The *data* and *len* represent the 3211 * data to write in bytes. 3212 * Return 3213 * 0 on success, or a negative error in case of failure: 3214 * 3215 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3216 * 3217 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk) 3218 * Description 3219 * Return the cgroup v2 id of the socket *sk*. 3220 * 3221 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one 3222 * returned from **bpf_sk_lookup_xxx**\ (), 3223 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3224 * same as in **bpf_skb_cgroup_id**\ (). 3225 * 3226 * This helper is available only if the kernel was compiled with 3227 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3228 * Return 3229 * The id is returned or 0 in case the id could not be retrieved. 3230 * 3231 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level) 3232 * Description 3233 * Return id of cgroup v2 that is ancestor of cgroup associated 3234 * with the *sk* at the *ancestor_level*. The root cgroup is at 3235 * *ancestor_level* zero and each step down the hierarchy 3236 * increments the level. If *ancestor_level* == level of cgroup 3237 * associated with *sk*, then return value will be same as that 3238 * of **bpf_sk_cgroup_id**\ (). 3239 * 3240 * The helper is useful to implement policies based on cgroups 3241 * that are upper in hierarchy than immediate cgroup associated 3242 * with *sk*. 3243 * 3244 * The format of returned id and helper limitations are same as in 3245 * **bpf_sk_cgroup_id**\ (). 3246 * Return 3247 * The id is returned or 0 in case the id could not be retrieved. 3248 * 3249 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 3250 * Description 3251 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 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 * 0 on success, or a negative error in case of failure. 3258 * 3259 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 3260 * Description 3261 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3262 * Return 3263 * Valid pointer with *size* bytes of memory available; NULL, 3264 * otherwise. 3265 * 3266 * void bpf_ringbuf_submit(void *data, u64 flags) 3267 * Description 3268 * Submit reserved ring buffer sample, pointed to by *data*. 3269 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3270 * of new data availability is sent. 3271 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3272 * of new data availability is sent unconditionally. 3273 * Return 3274 * Nothing. Always succeeds. 3275 * 3276 * void bpf_ringbuf_discard(void *data, u64 flags) 3277 * Description 3278 * Discard reserved ring buffer sample, pointed to by *data*. 3279 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3280 * of new data availability is sent. 3281 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3282 * of new data availability is sent unconditionally. 3283 * Return 3284 * Nothing. Always succeeds. 3285 * 3286 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 3287 * Description 3288 * Query various characteristics of provided ring buffer. What 3289 * exactly is queries is determined by *flags*: 3290 * 3291 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3292 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3293 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3294 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3295 * 3296 * Data returned is just a momentary snapshot of actual values 3297 * and could be inaccurate, so this facility should be used to 3298 * power heuristics and for reporting, not to make 100% correct 3299 * calculation. 3300 * Return 3301 * Requested value, or 0, if *flags* are not recognized. 3302 * 3303 * long bpf_csum_level(struct sk_buff *skb, u64 level) 3304 * Description 3305 * Change the skbs checksum level by one layer up or down, or 3306 * reset it entirely to none in order to have the stack perform 3307 * checksum validation. The level is applicable to the following 3308 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3309 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3310 * through **bpf_skb_adjust_room**\ () helper with passing in 3311 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3312 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3313 * the UDP header is removed. Similarly, an encap of the latter 3314 * into the former could be accompanied by a helper call to 3315 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3316 * skb is still intended to be processed in higher layers of the 3317 * stack instead of just egressing at tc. 3318 * 3319 * There are three supported level settings at this time: 3320 * 3321 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3322 * with CHECKSUM_UNNECESSARY. 3323 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3324 * with CHECKSUM_UNNECESSARY. 3325 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3326 * sets CHECKSUM_NONE to force checksum validation by the stack. 3327 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3328 * skb->csum_level. 3329 * Return 3330 * 0 on success, or a negative error in case of failure. In the 3331 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3332 * is returned or the error code -EACCES in case the skb is not 3333 * subject to CHECKSUM_UNNECESSARY. 3334 * 3335 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 3336 * Description 3337 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3338 * Return 3339 * *sk* if casting is valid, or NULL otherwise. 3340 * 3341 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 3342 * Description 3343 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3344 * Return 3345 * *sk* if casting is valid, or NULL otherwise. 3346 * 3347 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 3348 * Description 3349 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3350 * Return 3351 * *sk* if casting is valid, or NULL otherwise. 3352 * 3353 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 3354 * Description 3355 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3356 * Return 3357 * *sk* if casting is valid, or NULL otherwise. 3358 * 3359 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 3360 * Description 3361 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3362 * Return 3363 * *sk* if casting is valid, or NULL otherwise. 3364 * 3365 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 3366 * Description 3367 * Return a user or a kernel stack in bpf program provided buffer. 3368 * To achieve this, the helper needs *task*, which is a valid 3369 * pointer to struct task_struct. To store the stacktrace, the 3370 * bpf program provides *buf* with a nonnegative *size*. 3371 * 3372 * The last argument, *flags*, holds the number of stack frames to 3373 * skip (from 0 to 255), masked with 3374 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3375 * the following flags: 3376 * 3377 * **BPF_F_USER_STACK** 3378 * Collect a user space stack instead of a kernel stack. 3379 * **BPF_F_USER_BUILD_ID** 3380 * Collect buildid+offset instead of ips for user stack, 3381 * only valid if **BPF_F_USER_STACK** is also specified. 3382 * 3383 * **bpf_get_task_stack**\ () can collect up to 3384 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3385 * to sufficient large buffer size. Note that 3386 * this limit can be controlled with the **sysctl** program, and 3387 * that it should be manually increased in order to profile long 3388 * user stacks (such as stacks for Java programs). To do so, use: 3389 * 3390 * :: 3391 * 3392 * # sysctl kernel.perf_event_max_stack=<new value> 3393 * Return 3394 * A non-negative value equal to or less than *size* on success, 3395 * or a negative error in case of failure. 3396 * 3397 */ 3398 #define __BPF_FUNC_MAPPER(FN) \ 3399 FN(unspec), \ 3400 FN(map_lookup_elem), \ 3401 FN(map_update_elem), \ 3402 FN(map_delete_elem), \ 3403 FN(probe_read), \ 3404 FN(ktime_get_ns), \ 3405 FN(trace_printk), \ 3406 FN(get_prandom_u32), \ 3407 FN(get_smp_processor_id), \ 3408 FN(skb_store_bytes), \ 3409 FN(l3_csum_replace), \ 3410 FN(l4_csum_replace), \ 3411 FN(tail_call), \ 3412 FN(clone_redirect), \ 3413 FN(get_current_pid_tgid), \ 3414 FN(get_current_uid_gid), \ 3415 FN(get_current_comm), \ 3416 FN(get_cgroup_classid), \ 3417 FN(skb_vlan_push), \ 3418 FN(skb_vlan_pop), \ 3419 FN(skb_get_tunnel_key), \ 3420 FN(skb_set_tunnel_key), \ 3421 FN(perf_event_read), \ 3422 FN(redirect), \ 3423 FN(get_route_realm), \ 3424 FN(perf_event_output), \ 3425 FN(skb_load_bytes), \ 3426 FN(get_stackid), \ 3427 FN(csum_diff), \ 3428 FN(skb_get_tunnel_opt), \ 3429 FN(skb_set_tunnel_opt), \ 3430 FN(skb_change_proto), \ 3431 FN(skb_change_type), \ 3432 FN(skb_under_cgroup), \ 3433 FN(get_hash_recalc), \ 3434 FN(get_current_task), \ 3435 FN(probe_write_user), \ 3436 FN(current_task_under_cgroup), \ 3437 FN(skb_change_tail), \ 3438 FN(skb_pull_data), \ 3439 FN(csum_update), \ 3440 FN(set_hash_invalid), \ 3441 FN(get_numa_node_id), \ 3442 FN(skb_change_head), \ 3443 FN(xdp_adjust_head), \ 3444 FN(probe_read_str), \ 3445 FN(get_socket_cookie), \ 3446 FN(get_socket_uid), \ 3447 FN(set_hash), \ 3448 FN(setsockopt), \ 3449 FN(skb_adjust_room), \ 3450 FN(redirect_map), \ 3451 FN(sk_redirect_map), \ 3452 FN(sock_map_update), \ 3453 FN(xdp_adjust_meta), \ 3454 FN(perf_event_read_value), \ 3455 FN(perf_prog_read_value), \ 3456 FN(getsockopt), \ 3457 FN(override_return), \ 3458 FN(sock_ops_cb_flags_set), \ 3459 FN(msg_redirect_map), \ 3460 FN(msg_apply_bytes), \ 3461 FN(msg_cork_bytes), \ 3462 FN(msg_pull_data), \ 3463 FN(bind), \ 3464 FN(xdp_adjust_tail), \ 3465 FN(skb_get_xfrm_state), \ 3466 FN(get_stack), \ 3467 FN(skb_load_bytes_relative), \ 3468 FN(fib_lookup), \ 3469 FN(sock_hash_update), \ 3470 FN(msg_redirect_hash), \ 3471 FN(sk_redirect_hash), \ 3472 FN(lwt_push_encap), \ 3473 FN(lwt_seg6_store_bytes), \ 3474 FN(lwt_seg6_adjust_srh), \ 3475 FN(lwt_seg6_action), \ 3476 FN(rc_repeat), \ 3477 FN(rc_keydown), \ 3478 FN(skb_cgroup_id), \ 3479 FN(get_current_cgroup_id), \ 3480 FN(get_local_storage), \ 3481 FN(sk_select_reuseport), \ 3482 FN(skb_ancestor_cgroup_id), \ 3483 FN(sk_lookup_tcp), \ 3484 FN(sk_lookup_udp), \ 3485 FN(sk_release), \ 3486 FN(map_push_elem), \ 3487 FN(map_pop_elem), \ 3488 FN(map_peek_elem), \ 3489 FN(msg_push_data), \ 3490 FN(msg_pop_data), \ 3491 FN(rc_pointer_rel), \ 3492 FN(spin_lock), \ 3493 FN(spin_unlock), \ 3494 FN(sk_fullsock), \ 3495 FN(tcp_sock), \ 3496 FN(skb_ecn_set_ce), \ 3497 FN(get_listener_sock), \ 3498 FN(skc_lookup_tcp), \ 3499 FN(tcp_check_syncookie), \ 3500 FN(sysctl_get_name), \ 3501 FN(sysctl_get_current_value), \ 3502 FN(sysctl_get_new_value), \ 3503 FN(sysctl_set_new_value), \ 3504 FN(strtol), \ 3505 FN(strtoul), \ 3506 FN(sk_storage_get), \ 3507 FN(sk_storage_delete), \ 3508 FN(send_signal), \ 3509 FN(tcp_gen_syncookie), \ 3510 FN(skb_output), \ 3511 FN(probe_read_user), \ 3512 FN(probe_read_kernel), \ 3513 FN(probe_read_user_str), \ 3514 FN(probe_read_kernel_str), \ 3515 FN(tcp_send_ack), \ 3516 FN(send_signal_thread), \ 3517 FN(jiffies64), \ 3518 FN(read_branch_records), \ 3519 FN(get_ns_current_pid_tgid), \ 3520 FN(xdp_output), \ 3521 FN(get_netns_cookie), \ 3522 FN(get_current_ancestor_cgroup_id), \ 3523 FN(sk_assign), \ 3524 FN(ktime_get_boot_ns), \ 3525 FN(seq_printf), \ 3526 FN(seq_write), \ 3527 FN(sk_cgroup_id), \ 3528 FN(sk_ancestor_cgroup_id), \ 3529 FN(ringbuf_output), \ 3530 FN(ringbuf_reserve), \ 3531 FN(ringbuf_submit), \ 3532 FN(ringbuf_discard), \ 3533 FN(ringbuf_query), \ 3534 FN(csum_level), \ 3535 FN(skc_to_tcp6_sock), \ 3536 FN(skc_to_tcp_sock), \ 3537 FN(skc_to_tcp_timewait_sock), \ 3538 FN(skc_to_tcp_request_sock), \ 3539 FN(skc_to_udp6_sock), \ 3540 FN(get_task_stack), \ 3541 /* */ 3542 3543 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3544 * function eBPF program intends to call 3545 */ 3546 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3547 enum bpf_func_id { 3548 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 3549 __BPF_FUNC_MAX_ID, 3550 }; 3551 #undef __BPF_ENUM_FN 3552 3553 /* All flags used by eBPF helper functions, placed here. */ 3554 3555 /* BPF_FUNC_skb_store_bytes flags. */ 3556 enum { 3557 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 3558 BPF_F_INVALIDATE_HASH = (1ULL << 1), 3559 }; 3560 3561 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3562 * First 4 bits are for passing the header field size. 3563 */ 3564 enum { 3565 BPF_F_HDR_FIELD_MASK = 0xfULL, 3566 }; 3567 3568 /* BPF_FUNC_l4_csum_replace flags. */ 3569 enum { 3570 BPF_F_PSEUDO_HDR = (1ULL << 4), 3571 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 3572 BPF_F_MARK_ENFORCE = (1ULL << 6), 3573 }; 3574 3575 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 3576 enum { 3577 BPF_F_INGRESS = (1ULL << 0), 3578 }; 3579 3580 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 3581 enum { 3582 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 3583 }; 3584 3585 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 3586 enum { 3587 BPF_F_SKIP_FIELD_MASK = 0xffULL, 3588 BPF_F_USER_STACK = (1ULL << 8), 3589 /* flags used by BPF_FUNC_get_stackid only. */ 3590 BPF_F_FAST_STACK_CMP = (1ULL << 9), 3591 BPF_F_REUSE_STACKID = (1ULL << 10), 3592 /* flags used by BPF_FUNC_get_stack only. */ 3593 BPF_F_USER_BUILD_ID = (1ULL << 11), 3594 }; 3595 3596 /* BPF_FUNC_skb_set_tunnel_key flags. */ 3597 enum { 3598 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 3599 BPF_F_DONT_FRAGMENT = (1ULL << 2), 3600 BPF_F_SEQ_NUMBER = (1ULL << 3), 3601 }; 3602 3603 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3604 * BPF_FUNC_perf_event_read_value flags. 3605 */ 3606 enum { 3607 BPF_F_INDEX_MASK = 0xffffffffULL, 3608 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 3609 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 3610 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 3611 }; 3612 3613 /* Current network namespace */ 3614 enum { 3615 BPF_F_CURRENT_NETNS = (-1L), 3616 }; 3617 3618 /* BPF_FUNC_csum_level level values. */ 3619 enum { 3620 BPF_CSUM_LEVEL_QUERY, 3621 BPF_CSUM_LEVEL_INC, 3622 BPF_CSUM_LEVEL_DEC, 3623 BPF_CSUM_LEVEL_RESET, 3624 }; 3625 3626 /* BPF_FUNC_skb_adjust_room flags. */ 3627 enum { 3628 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 3629 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 3630 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 3631 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 3632 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 3633 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 3634 }; 3635 3636 enum { 3637 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 3638 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 3639 }; 3640 3641 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 3642 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 3643 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3644 3645 /* BPF_FUNC_sysctl_get_name flags. */ 3646 enum { 3647 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 3648 }; 3649 3650 /* BPF_FUNC_sk_storage_get flags */ 3651 enum { 3652 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0), 3653 }; 3654 3655 /* BPF_FUNC_read_branch_records flags. */ 3656 enum { 3657 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 3658 }; 3659 3660 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 3661 * BPF_FUNC_bpf_ringbuf_output flags. 3662 */ 3663 enum { 3664 BPF_RB_NO_WAKEUP = (1ULL << 0), 3665 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 3666 }; 3667 3668 /* BPF_FUNC_bpf_ringbuf_query flags */ 3669 enum { 3670 BPF_RB_AVAIL_DATA = 0, 3671 BPF_RB_RING_SIZE = 1, 3672 BPF_RB_CONS_POS = 2, 3673 BPF_RB_PROD_POS = 3, 3674 }; 3675 3676 /* BPF ring buffer constants */ 3677 enum { 3678 BPF_RINGBUF_BUSY_BIT = (1U << 31), 3679 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 3680 BPF_RINGBUF_HDR_SZ = 8, 3681 }; 3682 3683 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 3684 enum { 3685 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 3686 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 3687 }; 3688 3689 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 3690 enum bpf_adj_room_mode { 3691 BPF_ADJ_ROOM_NET, 3692 BPF_ADJ_ROOM_MAC, 3693 }; 3694 3695 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 3696 enum bpf_hdr_start_off { 3697 BPF_HDR_START_MAC, 3698 BPF_HDR_START_NET, 3699 }; 3700 3701 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3702 enum bpf_lwt_encap_mode { 3703 BPF_LWT_ENCAP_SEG6, 3704 BPF_LWT_ENCAP_SEG6_INLINE, 3705 BPF_LWT_ENCAP_IP, 3706 }; 3707 3708 #define __bpf_md_ptr(type, name) \ 3709 union { \ 3710 type name; \ 3711 __u64 :64; \ 3712 } __attribute__((aligned(8))) 3713 3714 /* user accessible mirror of in-kernel sk_buff. 3715 * new fields can only be added to the end of this structure 3716 */ 3717 struct __sk_buff { 3718 __u32 len; 3719 __u32 pkt_type; 3720 __u32 mark; 3721 __u32 queue_mapping; 3722 __u32 protocol; 3723 __u32 vlan_present; 3724 __u32 vlan_tci; 3725 __u32 vlan_proto; 3726 __u32 priority; 3727 __u32 ingress_ifindex; 3728 __u32 ifindex; 3729 __u32 tc_index; 3730 __u32 cb[5]; 3731 __u32 hash; 3732 __u32 tc_classid; 3733 __u32 data; 3734 __u32 data_end; 3735 __u32 napi_id; 3736 3737 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 3738 __u32 family; 3739 __u32 remote_ip4; /* Stored in network byte order */ 3740 __u32 local_ip4; /* Stored in network byte order */ 3741 __u32 remote_ip6[4]; /* Stored in network byte order */ 3742 __u32 local_ip6[4]; /* Stored in network byte order */ 3743 __u32 remote_port; /* Stored in network byte order */ 3744 __u32 local_port; /* stored in host byte order */ 3745 /* ... here. */ 3746 3747 __u32 data_meta; 3748 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 3749 __u64 tstamp; 3750 __u32 wire_len; 3751 __u32 gso_segs; 3752 __bpf_md_ptr(struct bpf_sock *, sk); 3753 __u32 gso_size; 3754 }; 3755 3756 struct bpf_tunnel_key { 3757 __u32 tunnel_id; 3758 union { 3759 __u32 remote_ipv4; 3760 __u32 remote_ipv6[4]; 3761 }; 3762 __u8 tunnel_tos; 3763 __u8 tunnel_ttl; 3764 __u16 tunnel_ext; /* Padding, future use. */ 3765 __u32 tunnel_label; 3766 }; 3767 3768 /* user accessible mirror of in-kernel xfrm_state. 3769 * new fields can only be added to the end of this structure 3770 */ 3771 struct bpf_xfrm_state { 3772 __u32 reqid; 3773 __u32 spi; /* Stored in network byte order */ 3774 __u16 family; 3775 __u16 ext; /* Padding, future use. */ 3776 union { 3777 __u32 remote_ipv4; /* Stored in network byte order */ 3778 __u32 remote_ipv6[4]; /* Stored in network byte order */ 3779 }; 3780 }; 3781 3782 /* Generic BPF return codes which all BPF program types may support. 3783 * The values are binary compatible with their TC_ACT_* counter-part to 3784 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 3785 * programs. 3786 * 3787 * XDP is handled seprately, see XDP_*. 3788 */ 3789 enum bpf_ret_code { 3790 BPF_OK = 0, 3791 /* 1 reserved */ 3792 BPF_DROP = 2, 3793 /* 3-6 reserved */ 3794 BPF_REDIRECT = 7, 3795 /* >127 are reserved for prog type specific return codes. 3796 * 3797 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 3798 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 3799 * changed and should be routed based on its new L3 header. 3800 * (This is an L3 redirect, as opposed to L2 redirect 3801 * represented by BPF_REDIRECT above). 3802 */ 3803 BPF_LWT_REROUTE = 128, 3804 }; 3805 3806 struct bpf_sock { 3807 __u32 bound_dev_if; 3808 __u32 family; 3809 __u32 type; 3810 __u32 protocol; 3811 __u32 mark; 3812 __u32 priority; 3813 /* IP address also allows 1 and 2 bytes access */ 3814 __u32 src_ip4; 3815 __u32 src_ip6[4]; 3816 __u32 src_port; /* host byte order */ 3817 __u32 dst_port; /* network byte order */ 3818 __u32 dst_ip4; 3819 __u32 dst_ip6[4]; 3820 __u32 state; 3821 __s32 rx_queue_mapping; 3822 }; 3823 3824 struct bpf_tcp_sock { 3825 __u32 snd_cwnd; /* Sending congestion window */ 3826 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 3827 __u32 rtt_min; 3828 __u32 snd_ssthresh; /* Slow start size threshold */ 3829 __u32 rcv_nxt; /* What we want to receive next */ 3830 __u32 snd_nxt; /* Next sequence we send */ 3831 __u32 snd_una; /* First byte we want an ack for */ 3832 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 3833 __u32 ecn_flags; /* ECN status bits. */ 3834 __u32 rate_delivered; /* saved rate sample: packets delivered */ 3835 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 3836 __u32 packets_out; /* Packets which are "in flight" */ 3837 __u32 retrans_out; /* Retransmitted packets out */ 3838 __u32 total_retrans; /* Total retransmits for entire connection */ 3839 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 3840 * total number of segments in. 3841 */ 3842 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 3843 * total number of data segments in. 3844 */ 3845 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 3846 * The total number of segments sent. 3847 */ 3848 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 3849 * total number of data segments sent. 3850 */ 3851 __u32 lost_out; /* Lost packets */ 3852 __u32 sacked_out; /* SACK'd packets */ 3853 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 3854 * sum(delta(rcv_nxt)), or how many bytes 3855 * were acked. 3856 */ 3857 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 3858 * sum(delta(snd_una)), or how many bytes 3859 * were acked. 3860 */ 3861 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 3862 * total number of DSACK blocks received 3863 */ 3864 __u32 delivered; /* Total data packets delivered incl. rexmits */ 3865 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 3866 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 3867 }; 3868 3869 struct bpf_sock_tuple { 3870 union { 3871 struct { 3872 __be32 saddr; 3873 __be32 daddr; 3874 __be16 sport; 3875 __be16 dport; 3876 } ipv4; 3877 struct { 3878 __be32 saddr[4]; 3879 __be32 daddr[4]; 3880 __be16 sport; 3881 __be16 dport; 3882 } ipv6; 3883 }; 3884 }; 3885 3886 struct bpf_xdp_sock { 3887 __u32 queue_id; 3888 }; 3889 3890 #define XDP_PACKET_HEADROOM 256 3891 3892 /* User return codes for XDP prog type. 3893 * A valid XDP program must return one of these defined values. All other 3894 * return codes are reserved for future use. Unknown return codes will 3895 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 3896 */ 3897 enum xdp_action { 3898 XDP_ABORTED = 0, 3899 XDP_DROP, 3900 XDP_PASS, 3901 XDP_TX, 3902 XDP_REDIRECT, 3903 }; 3904 3905 /* user accessible metadata for XDP packet hook 3906 * new fields must be added to the end of this structure 3907 */ 3908 struct xdp_md { 3909 __u32 data; 3910 __u32 data_end; 3911 __u32 data_meta; 3912 /* Below access go through struct xdp_rxq_info */ 3913 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 3914 __u32 rx_queue_index; /* rxq->queue_index */ 3915 3916 __u32 egress_ifindex; /* txq->dev->ifindex */ 3917 }; 3918 3919 /* DEVMAP map-value layout 3920 * 3921 * The struct data-layout of map-value is a configuration interface. 3922 * New members can only be added to the end of this structure. 3923 */ 3924 struct bpf_devmap_val { 3925 __u32 ifindex; /* device index */ 3926 union { 3927 int fd; /* prog fd on map write */ 3928 __u32 id; /* prog id on map read */ 3929 } bpf_prog; 3930 }; 3931 3932 /* CPUMAP map-value layout 3933 * 3934 * The struct data-layout of map-value is a configuration interface. 3935 * New members can only be added to the end of this structure. 3936 */ 3937 struct bpf_cpumap_val { 3938 __u32 qsize; /* queue size to remote target CPU */ 3939 union { 3940 int fd; /* prog fd on map write */ 3941 __u32 id; /* prog id on map read */ 3942 } bpf_prog; 3943 }; 3944 3945 enum sk_action { 3946 SK_DROP = 0, 3947 SK_PASS, 3948 }; 3949 3950 /* user accessible metadata for SK_MSG packet hook, new fields must 3951 * be added to the end of this structure 3952 */ 3953 struct sk_msg_md { 3954 __bpf_md_ptr(void *, data); 3955 __bpf_md_ptr(void *, data_end); 3956 3957 __u32 family; 3958 __u32 remote_ip4; /* Stored in network byte order */ 3959 __u32 local_ip4; /* Stored in network byte order */ 3960 __u32 remote_ip6[4]; /* Stored in network byte order */ 3961 __u32 local_ip6[4]; /* Stored in network byte order */ 3962 __u32 remote_port; /* Stored in network byte order */ 3963 __u32 local_port; /* stored in host byte order */ 3964 __u32 size; /* Total size of sk_msg */ 3965 3966 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 3967 }; 3968 3969 struct sk_reuseport_md { 3970 /* 3971 * Start of directly accessible data. It begins from 3972 * the tcp/udp header. 3973 */ 3974 __bpf_md_ptr(void *, data); 3975 /* End of directly accessible data */ 3976 __bpf_md_ptr(void *, data_end); 3977 /* 3978 * Total length of packet (starting from the tcp/udp header). 3979 * Note that the directly accessible bytes (data_end - data) 3980 * could be less than this "len". Those bytes could be 3981 * indirectly read by a helper "bpf_skb_load_bytes()". 3982 */ 3983 __u32 len; 3984 /* 3985 * Eth protocol in the mac header (network byte order). e.g. 3986 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 3987 */ 3988 __u32 eth_protocol; 3989 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 3990 __u32 bind_inany; /* Is sock bound to an INANY address? */ 3991 __u32 hash; /* A hash of the packet 4 tuples */ 3992 }; 3993 3994 #define BPF_TAG_SIZE 8 3995 3996 struct bpf_prog_info { 3997 __u32 type; 3998 __u32 id; 3999 __u8 tag[BPF_TAG_SIZE]; 4000 __u32 jited_prog_len; 4001 __u32 xlated_prog_len; 4002 __aligned_u64 jited_prog_insns; 4003 __aligned_u64 xlated_prog_insns; 4004 __u64 load_time; /* ns since boottime */ 4005 __u32 created_by_uid; 4006 __u32 nr_map_ids; 4007 __aligned_u64 map_ids; 4008 char name[BPF_OBJ_NAME_LEN]; 4009 __u32 ifindex; 4010 __u32 gpl_compatible:1; 4011 __u32 :31; /* alignment pad */ 4012 __u64 netns_dev; 4013 __u64 netns_ino; 4014 __u32 nr_jited_ksyms; 4015 __u32 nr_jited_func_lens; 4016 __aligned_u64 jited_ksyms; 4017 __aligned_u64 jited_func_lens; 4018 __u32 btf_id; 4019 __u32 func_info_rec_size; 4020 __aligned_u64 func_info; 4021 __u32 nr_func_info; 4022 __u32 nr_line_info; 4023 __aligned_u64 line_info; 4024 __aligned_u64 jited_line_info; 4025 __u32 nr_jited_line_info; 4026 __u32 line_info_rec_size; 4027 __u32 jited_line_info_rec_size; 4028 __u32 nr_prog_tags; 4029 __aligned_u64 prog_tags; 4030 __u64 run_time_ns; 4031 __u64 run_cnt; 4032 } __attribute__((aligned(8))); 4033 4034 struct bpf_map_info { 4035 __u32 type; 4036 __u32 id; 4037 __u32 key_size; 4038 __u32 value_size; 4039 __u32 max_entries; 4040 __u32 map_flags; 4041 char name[BPF_OBJ_NAME_LEN]; 4042 __u32 ifindex; 4043 __u32 btf_vmlinux_value_type_id; 4044 __u64 netns_dev; 4045 __u64 netns_ino; 4046 __u32 btf_id; 4047 __u32 btf_key_type_id; 4048 __u32 btf_value_type_id; 4049 } __attribute__((aligned(8))); 4050 4051 struct bpf_btf_info { 4052 __aligned_u64 btf; 4053 __u32 btf_size; 4054 __u32 id; 4055 } __attribute__((aligned(8))); 4056 4057 struct bpf_link_info { 4058 __u32 type; 4059 __u32 id; 4060 __u32 prog_id; 4061 union { 4062 struct { 4063 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 4064 __u32 tp_name_len; /* in/out: tp_name buffer len */ 4065 } raw_tracepoint; 4066 struct { 4067 __u32 attach_type; 4068 } tracing; 4069 struct { 4070 __u64 cgroup_id; 4071 __u32 attach_type; 4072 } cgroup; 4073 struct { 4074 __u32 netns_ino; 4075 __u32 attach_type; 4076 } netns; 4077 struct { 4078 __u32 ifindex; 4079 } xdp; 4080 }; 4081 } __attribute__((aligned(8))); 4082 4083 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 4084 * by user and intended to be used by socket (e.g. to bind to, depends on 4085 * attach type). 4086 */ 4087 struct bpf_sock_addr { 4088 __u32 user_family; /* Allows 4-byte read, but no write. */ 4089 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4090 * Stored in network byte order. 4091 */ 4092 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4093 * Stored in network byte order. 4094 */ 4095 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 4096 * Stored in network byte order 4097 */ 4098 __u32 family; /* Allows 4-byte read, but no write */ 4099 __u32 type; /* Allows 4-byte read, but no write */ 4100 __u32 protocol; /* Allows 4-byte read, but no write */ 4101 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 4102 * Stored in network byte order. 4103 */ 4104 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 4105 * Stored in network byte order. 4106 */ 4107 __bpf_md_ptr(struct bpf_sock *, sk); 4108 }; 4109 4110 /* User bpf_sock_ops struct to access socket values and specify request ops 4111 * and their replies. 4112 * Some of this fields are in network (bigendian) byte order and may need 4113 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 4114 * New fields can only be added at the end of this structure 4115 */ 4116 struct bpf_sock_ops { 4117 __u32 op; 4118 union { 4119 __u32 args[4]; /* Optionally passed to bpf program */ 4120 __u32 reply; /* Returned by bpf program */ 4121 __u32 replylong[4]; /* Optionally returned by bpf prog */ 4122 }; 4123 __u32 family; 4124 __u32 remote_ip4; /* Stored in network byte order */ 4125 __u32 local_ip4; /* Stored in network byte order */ 4126 __u32 remote_ip6[4]; /* Stored in network byte order */ 4127 __u32 local_ip6[4]; /* Stored in network byte order */ 4128 __u32 remote_port; /* Stored in network byte order */ 4129 __u32 local_port; /* stored in host byte order */ 4130 __u32 is_fullsock; /* Some TCP fields are only valid if 4131 * there is a full socket. If not, the 4132 * fields read as zero. 4133 */ 4134 __u32 snd_cwnd; 4135 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 4136 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 4137 __u32 state; 4138 __u32 rtt_min; 4139 __u32 snd_ssthresh; 4140 __u32 rcv_nxt; 4141 __u32 snd_nxt; 4142 __u32 snd_una; 4143 __u32 mss_cache; 4144 __u32 ecn_flags; 4145 __u32 rate_delivered; 4146 __u32 rate_interval_us; 4147 __u32 packets_out; 4148 __u32 retrans_out; 4149 __u32 total_retrans; 4150 __u32 segs_in; 4151 __u32 data_segs_in; 4152 __u32 segs_out; 4153 __u32 data_segs_out; 4154 __u32 lost_out; 4155 __u32 sacked_out; 4156 __u32 sk_txhash; 4157 __u64 bytes_received; 4158 __u64 bytes_acked; 4159 __bpf_md_ptr(struct bpf_sock *, sk); 4160 }; 4161 4162 /* Definitions for bpf_sock_ops_cb_flags */ 4163 enum { 4164 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 4165 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 4166 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 4167 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 4168 /* Mask of all currently supported cb flags */ 4169 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF, 4170 }; 4171 4172 /* List of known BPF sock_ops operators. 4173 * New entries can only be added at the end 4174 */ 4175 enum { 4176 BPF_SOCK_OPS_VOID, 4177 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 4178 * -1 if default value should be used 4179 */ 4180 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 4181 * window (in packets) or -1 if default 4182 * value should be used 4183 */ 4184 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 4185 * active connection is initialized 4186 */ 4187 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 4188 * active connection is 4189 * established 4190 */ 4191 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 4192 * passive connection is 4193 * established 4194 */ 4195 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 4196 * needs ECN 4197 */ 4198 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 4199 * based on the path and may be 4200 * dependent on the congestion control 4201 * algorithm. In general it indicates 4202 * a congestion threshold. RTTs above 4203 * this indicate congestion 4204 */ 4205 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 4206 * Arg1: value of icsk_retransmits 4207 * Arg2: value of icsk_rto 4208 * Arg3: whether RTO has expired 4209 */ 4210 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 4211 * Arg1: sequence number of 1st byte 4212 * Arg2: # segments 4213 * Arg3: return value of 4214 * tcp_transmit_skb (0 => success) 4215 */ 4216 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 4217 * Arg1: old_state 4218 * Arg2: new_state 4219 */ 4220 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 4221 * socket transition to LISTEN state. 4222 */ 4223 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 4224 */ 4225 }; 4226 4227 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 4228 * changes between the TCP and BPF versions. Ideally this should never happen. 4229 * If it does, we need to add code to convert them before calling 4230 * the BPF sock_ops function. 4231 */ 4232 enum { 4233 BPF_TCP_ESTABLISHED = 1, 4234 BPF_TCP_SYN_SENT, 4235 BPF_TCP_SYN_RECV, 4236 BPF_TCP_FIN_WAIT1, 4237 BPF_TCP_FIN_WAIT2, 4238 BPF_TCP_TIME_WAIT, 4239 BPF_TCP_CLOSE, 4240 BPF_TCP_CLOSE_WAIT, 4241 BPF_TCP_LAST_ACK, 4242 BPF_TCP_LISTEN, 4243 BPF_TCP_CLOSING, /* Now a valid state */ 4244 BPF_TCP_NEW_SYN_RECV, 4245 4246 BPF_TCP_MAX_STATES /* Leave at the end! */ 4247 }; 4248 4249 enum { 4250 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 4251 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 4252 }; 4253 4254 struct bpf_perf_event_value { 4255 __u64 counter; 4256 __u64 enabled; 4257 __u64 running; 4258 }; 4259 4260 enum { 4261 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 4262 BPF_DEVCG_ACC_READ = (1ULL << 1), 4263 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 4264 }; 4265 4266 enum { 4267 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 4268 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 4269 }; 4270 4271 struct bpf_cgroup_dev_ctx { 4272 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 4273 __u32 access_type; 4274 __u32 major; 4275 __u32 minor; 4276 }; 4277 4278 struct bpf_raw_tracepoint_args { 4279 __u64 args[0]; 4280 }; 4281 4282 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 4283 * OUTPUT: Do lookup from egress perspective; default is ingress 4284 */ 4285 enum { 4286 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 4287 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 4288 }; 4289 4290 enum { 4291 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 4292 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 4293 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 4294 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 4295 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 4296 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 4297 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 4298 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 4299 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 4300 }; 4301 4302 struct bpf_fib_lookup { 4303 /* input: network family for lookup (AF_INET, AF_INET6) 4304 * output: network family of egress nexthop 4305 */ 4306 __u8 family; 4307 4308 /* set if lookup is to consider L4 data - e.g., FIB rules */ 4309 __u8 l4_protocol; 4310 __be16 sport; 4311 __be16 dport; 4312 4313 /* total length of packet from network header - used for MTU check */ 4314 __u16 tot_len; 4315 4316 /* input: L3 device index for lookup 4317 * output: device index from FIB lookup 4318 */ 4319 __u32 ifindex; 4320 4321 union { 4322 /* inputs to lookup */ 4323 __u8 tos; /* AF_INET */ 4324 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 4325 4326 /* output: metric of fib result (IPv4/IPv6 only) */ 4327 __u32 rt_metric; 4328 }; 4329 4330 union { 4331 __be32 ipv4_src; 4332 __u32 ipv6_src[4]; /* in6_addr; network order */ 4333 }; 4334 4335 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 4336 * network header. output: bpf_fib_lookup sets to gateway address 4337 * if FIB lookup returns gateway route 4338 */ 4339 union { 4340 __be32 ipv4_dst; 4341 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4342 }; 4343 4344 /* output */ 4345 __be16 h_vlan_proto; 4346 __be16 h_vlan_TCI; 4347 __u8 smac[6]; /* ETH_ALEN */ 4348 __u8 dmac[6]; /* ETH_ALEN */ 4349 }; 4350 4351 enum bpf_task_fd_type { 4352 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 4353 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 4354 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 4355 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 4356 BPF_FD_TYPE_UPROBE, /* filename + offset */ 4357 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 4358 }; 4359 4360 enum { 4361 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 4362 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 4363 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 4364 }; 4365 4366 struct bpf_flow_keys { 4367 __u16 nhoff; 4368 __u16 thoff; 4369 __u16 addr_proto; /* ETH_P_* of valid addrs */ 4370 __u8 is_frag; 4371 __u8 is_first_frag; 4372 __u8 is_encap; 4373 __u8 ip_proto; 4374 __be16 n_proto; 4375 __be16 sport; 4376 __be16 dport; 4377 union { 4378 struct { 4379 __be32 ipv4_src; 4380 __be32 ipv4_dst; 4381 }; 4382 struct { 4383 __u32 ipv6_src[4]; /* in6_addr; network order */ 4384 __u32 ipv6_dst[4]; /* in6_addr; network order */ 4385 }; 4386 }; 4387 __u32 flags; 4388 __be32 flow_label; 4389 }; 4390 4391 struct bpf_func_info { 4392 __u32 insn_off; 4393 __u32 type_id; 4394 }; 4395 4396 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 4397 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 4398 4399 struct bpf_line_info { 4400 __u32 insn_off; 4401 __u32 file_name_off; 4402 __u32 line_off; 4403 __u32 line_col; 4404 }; 4405 4406 struct bpf_spin_lock { 4407 __u32 val; 4408 }; 4409 4410 struct bpf_sysctl { 4411 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 4412 * Allows 1,2,4-byte read, but no write. 4413 */ 4414 __u32 file_pos; /* Sysctl file position to read from, write to. 4415 * Allows 1,2,4-byte read an 4-byte write. 4416 */ 4417 }; 4418 4419 struct bpf_sockopt { 4420 __bpf_md_ptr(struct bpf_sock *, sk); 4421 __bpf_md_ptr(void *, optval); 4422 __bpf_md_ptr(void *, optval_end); 4423 4424 __s32 level; 4425 __s32 optname; 4426 __s32 optlen; 4427 __s32 retval; 4428 }; 4429 4430 struct bpf_pidns_info { 4431 __u32 pid; 4432 __u32 tgid; 4433 }; 4434 4435 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 4436 struct bpf_sk_lookup { 4437 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 4438 4439 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 4440 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 4441 __u32 remote_ip4; /* Network byte order */ 4442 __u32 remote_ip6[4]; /* Network byte order */ 4443 __u32 remote_port; /* Network byte order */ 4444 __u32 local_ip4; /* Network byte order */ 4445 __u32 local_ip6[4]; /* Network byte order */ 4446 __u32 local_port; /* Host byte order */ 4447 }; 4448 4449 #endif /* _UAPI__LINUX_BPF_H__ */ 4450