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