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