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_MEMSX 0x80 /* load with sign extension */ 23#define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 24#define BPF_XADD 0xc0 /* exclusive add - legacy name */ 25 26/* alu/jmp fields */ 27#define BPF_MOV 0xb0 /* mov reg to reg */ 28#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 29 30/* change endianness of a register */ 31#define BPF_END 0xd0 /* flags for endianness conversion: */ 32#define BPF_TO_LE 0x00 /* convert to little-endian */ 33#define BPF_TO_BE 0x08 /* convert to big-endian */ 34#define BPF_FROM_LE BPF_TO_LE 35#define BPF_FROM_BE BPF_TO_BE 36 37/* jmp encodings */ 38#define BPF_JNE 0x50 /* jump != */ 39#define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 40#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 41#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 42#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 43#define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 44#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 45#define BPF_JCOND 0xe0 /* conditional pseudo jumps: may_goto, goto_or_nop */ 46#define BPF_CALL 0x80 /* function call */ 47#define BPF_EXIT 0x90 /* function return */ 48 49/* atomic op type fields (stored in immediate) */ 50#define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 51#define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 52#define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 53 54enum bpf_cond_pseudo_jmp { 55 BPF_MAY_GOTO = 0, 56}; 57 58/* Register numbers */ 59enum { 60 BPF_REG_0 = 0, 61 BPF_REG_1, 62 BPF_REG_2, 63 BPF_REG_3, 64 BPF_REG_4, 65 BPF_REG_5, 66 BPF_REG_6, 67 BPF_REG_7, 68 BPF_REG_8, 69 BPF_REG_9, 70 BPF_REG_10, 71 __MAX_BPF_REG, 72}; 73 74/* BPF has 10 general purpose 64-bit registers and stack frame. */ 75#define MAX_BPF_REG __MAX_BPF_REG 76 77struct bpf_insn { 78 __u8 code; /* opcode */ 79 __u8 dst_reg:4; /* dest register */ 80 __u8 src_reg:4; /* source register */ 81 __s16 off; /* signed offset */ 82 __s32 imm; /* signed immediate constant */ 83}; 84 85/* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 86 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 87 * the trailing flexible array member) instead. 88 */ 89struct bpf_lpm_trie_key { 90 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 91 __u8 data[0]; /* Arbitrary size */ 92}; 93 94/* Header for bpf_lpm_trie_key structs */ 95struct bpf_lpm_trie_key_hdr { 96 __u32 prefixlen; 97}; 98 99/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 100struct bpf_lpm_trie_key_u8 { 101 union { 102 struct bpf_lpm_trie_key_hdr hdr; 103 __u32 prefixlen; 104 }; 105 __u8 data[]; /* Arbitrary size */ 106}; 107 108struct bpf_cgroup_storage_key { 109 __u64 cgroup_inode_id; /* cgroup inode id */ 110 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 111}; 112 113enum bpf_cgroup_iter_order { 114 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 115 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 116 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 117 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 118 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 119}; 120 121union bpf_iter_link_info { 122 struct { 123 __u32 map_fd; 124 } map; 125 struct { 126 enum bpf_cgroup_iter_order order; 127 128 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 129 * both are zero, the walk starts from the default cgroup v2 130 * root. For walking v1 hierarchy, one should always explicitly 131 * specify cgroup_fd. 132 */ 133 __u32 cgroup_fd; 134 __u64 cgroup_id; 135 } cgroup; 136 /* Parameters of task iterators. */ 137 struct { 138 __u32 tid; 139 __u32 pid; 140 __u32 pid_fd; 141 } task; 142}; 143 144/* BPF syscall commands, see bpf(2) man-page for more details. */ 145/** 146 * DOC: eBPF Syscall Preamble 147 * 148 * The operation to be performed by the **bpf**\ () system call is determined 149 * by the *cmd* argument. Each operation takes an accompanying argument, 150 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 151 * below). The size argument is the size of the union pointed to by *attr*. 152 */ 153/** 154 * DOC: eBPF Syscall Commands 155 * 156 * BPF_MAP_CREATE 157 * Description 158 * Create a map and return a file descriptor that refers to the 159 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 160 * is automatically enabled for the new file descriptor. 161 * 162 * Applying **close**\ (2) to the file descriptor returned by 163 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 164 * 165 * Return 166 * A new file descriptor (a nonnegative integer), or -1 if an 167 * error occurred (in which case, *errno* is set appropriately). 168 * 169 * BPF_MAP_LOOKUP_ELEM 170 * Description 171 * Look up an element with a given *key* in the map referred to 172 * by the file descriptor *map_fd*. 173 * 174 * The *flags* argument may be specified as one of the 175 * following: 176 * 177 * **BPF_F_LOCK** 178 * Look up the value of a spin-locked map without 179 * returning the lock. This must be specified if the 180 * elements contain a spinlock. 181 * 182 * Return 183 * Returns zero on success. On error, -1 is returned and *errno* 184 * is set appropriately. 185 * 186 * BPF_MAP_UPDATE_ELEM 187 * Description 188 * Create or update an element (key/value pair) in a specified map. 189 * 190 * The *flags* argument should be specified as one of the 191 * following: 192 * 193 * **BPF_ANY** 194 * Create a new element or update an existing element. 195 * **BPF_NOEXIST** 196 * Create a new element only if it did not exist. 197 * **BPF_EXIST** 198 * Update an existing element. 199 * **BPF_F_LOCK** 200 * Update a spin_lock-ed map element. 201 * 202 * Return 203 * Returns zero on success. On error, -1 is returned and *errno* 204 * is set appropriately. 205 * 206 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 207 * **E2BIG**, **EEXIST**, or **ENOENT**. 208 * 209 * **E2BIG** 210 * The number of elements in the map reached the 211 * *max_entries* limit specified at map creation time. 212 * **EEXIST** 213 * If *flags* specifies **BPF_NOEXIST** and the element 214 * with *key* already exists in the map. 215 * **ENOENT** 216 * If *flags* specifies **BPF_EXIST** and the element with 217 * *key* does not exist in the map. 218 * 219 * BPF_MAP_DELETE_ELEM 220 * Description 221 * Look up and delete an element by key in a specified map. 222 * 223 * Return 224 * Returns zero on success. On error, -1 is returned and *errno* 225 * is set appropriately. 226 * 227 * BPF_MAP_GET_NEXT_KEY 228 * Description 229 * Look up an element by key in a specified map and return the key 230 * of the next element. Can be used to iterate over all elements 231 * in the map. 232 * 233 * Return 234 * Returns zero on success. On error, -1 is returned and *errno* 235 * is set appropriately. 236 * 237 * The following cases can be used to iterate over all elements of 238 * the map: 239 * 240 * * If *key* is not found, the operation returns zero and sets 241 * the *next_key* pointer to the key of the first element. 242 * * If *key* is found, the operation returns zero and sets the 243 * *next_key* pointer to the key of the next element. 244 * * If *key* is the last element, returns -1 and *errno* is set 245 * to **ENOENT**. 246 * 247 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 248 * **EINVAL** on error. 249 * 250 * BPF_PROG_LOAD 251 * Description 252 * Verify and load an eBPF program, returning a new file 253 * descriptor associated with the program. 254 * 255 * Applying **close**\ (2) to the file descriptor returned by 256 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 257 * 258 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 259 * automatically enabled for the new file descriptor. 260 * 261 * Return 262 * A new file descriptor (a nonnegative integer), or -1 if an 263 * error occurred (in which case, *errno* is set appropriately). 264 * 265 * BPF_OBJ_PIN 266 * Description 267 * Pin an eBPF program or map referred by the specified *bpf_fd* 268 * to the provided *pathname* on the filesystem. 269 * 270 * The *pathname* argument must not contain a dot ("."). 271 * 272 * On success, *pathname* retains a reference to the eBPF object, 273 * preventing deallocation of the object when the original 274 * *bpf_fd* is closed. This allow the eBPF object to live beyond 275 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 276 * process. 277 * 278 * Applying **unlink**\ (2) or similar calls to the *pathname* 279 * unpins the object from the filesystem, removing the reference. 280 * If no other file descriptors or filesystem nodes refer to the 281 * same object, it will be deallocated (see NOTES). 282 * 283 * The filesystem type for the parent directory of *pathname* must 284 * be **BPF_FS_MAGIC**. 285 * 286 * Return 287 * Returns zero on success. On error, -1 is returned and *errno* 288 * is set appropriately. 289 * 290 * BPF_OBJ_GET 291 * Description 292 * Open a file descriptor for the eBPF object pinned to the 293 * specified *pathname*. 294 * 295 * Return 296 * A new file descriptor (a nonnegative integer), or -1 if an 297 * error occurred (in which case, *errno* is set appropriately). 298 * 299 * BPF_PROG_ATTACH 300 * Description 301 * Attach an eBPF program to a *target_fd* at the specified 302 * *attach_type* hook. 303 * 304 * The *attach_type* specifies the eBPF attachment point to 305 * attach the program to, and must be one of *bpf_attach_type* 306 * (see below). 307 * 308 * The *attach_bpf_fd* must be a valid file descriptor for a 309 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 310 * or sock_ops type corresponding to the specified *attach_type*. 311 * 312 * The *target_fd* must be a valid file descriptor for a kernel 313 * object which depends on the attach type of *attach_bpf_fd*: 314 * 315 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 316 * **BPF_PROG_TYPE_CGROUP_SKB**, 317 * **BPF_PROG_TYPE_CGROUP_SOCK**, 318 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 319 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 320 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 321 * **BPF_PROG_TYPE_SOCK_OPS** 322 * 323 * Control Group v2 hierarchy with the eBPF controller 324 * enabled. Requires the kernel to be compiled with 325 * **CONFIG_CGROUP_BPF**. 326 * 327 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 328 * 329 * Network namespace (eg /proc/self/ns/net). 330 * 331 * **BPF_PROG_TYPE_LIRC_MODE2** 332 * 333 * LIRC device path (eg /dev/lircN). Requires the kernel 334 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 335 * 336 * **BPF_PROG_TYPE_SK_SKB**, 337 * **BPF_PROG_TYPE_SK_MSG** 338 * 339 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 340 * 341 * Return 342 * Returns zero on success. On error, -1 is returned and *errno* 343 * is set appropriately. 344 * 345 * BPF_PROG_DETACH 346 * Description 347 * Detach the eBPF program associated with the *target_fd* at the 348 * hook specified by *attach_type*. The program must have been 349 * previously attached using **BPF_PROG_ATTACH**. 350 * 351 * Return 352 * Returns zero on success. On error, -1 is returned and *errno* 353 * is set appropriately. 354 * 355 * BPF_PROG_TEST_RUN 356 * Description 357 * Run the eBPF program associated with the *prog_fd* a *repeat* 358 * number of times against a provided program context *ctx_in* and 359 * data *data_in*, and return the modified program context 360 * *ctx_out*, *data_out* (for example, packet data), result of the 361 * execution *retval*, and *duration* of the test run. 362 * 363 * The sizes of the buffers provided as input and output 364 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 365 * be provided in the corresponding variables *ctx_size_in*, 366 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 367 * of these parameters are not provided (ie set to NULL), the 368 * corresponding size field must be zero. 369 * 370 * Some program types have particular requirements: 371 * 372 * **BPF_PROG_TYPE_SK_LOOKUP** 373 * *data_in* and *data_out* must be NULL. 374 * 375 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 376 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 377 * 378 * *ctx_out*, *data_in* and *data_out* must be NULL. 379 * *repeat* must be zero. 380 * 381 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 382 * 383 * Return 384 * Returns zero on success. On error, -1 is returned and *errno* 385 * is set appropriately. 386 * 387 * **ENOSPC** 388 * Either *data_size_out* or *ctx_size_out* is too small. 389 * **ENOTSUPP** 390 * This command is not supported by the program type of 391 * the program referred to by *prog_fd*. 392 * 393 * BPF_PROG_GET_NEXT_ID 394 * Description 395 * Fetch the next eBPF program currently loaded into the kernel. 396 * 397 * Looks for the eBPF program with an id greater than *start_id* 398 * and updates *next_id* on success. If no other eBPF programs 399 * remain with ids higher than *start_id*, returns -1 and sets 400 * *errno* to **ENOENT**. 401 * 402 * Return 403 * Returns zero on success. On error, or when no id remains, -1 404 * is returned and *errno* is set appropriately. 405 * 406 * BPF_MAP_GET_NEXT_ID 407 * Description 408 * Fetch the next eBPF map currently loaded into the kernel. 409 * 410 * Looks for the eBPF map with an id greater than *start_id* 411 * and updates *next_id* on success. If no other eBPF maps 412 * remain with ids higher than *start_id*, returns -1 and sets 413 * *errno* to **ENOENT**. 414 * 415 * Return 416 * Returns zero on success. On error, or when no id remains, -1 417 * is returned and *errno* is set appropriately. 418 * 419 * BPF_PROG_GET_FD_BY_ID 420 * Description 421 * Open a file descriptor for the eBPF program corresponding to 422 * *prog_id*. 423 * 424 * Return 425 * A new file descriptor (a nonnegative integer), or -1 if an 426 * error occurred (in which case, *errno* is set appropriately). 427 * 428 * BPF_MAP_GET_FD_BY_ID 429 * Description 430 * Open a file descriptor for the eBPF map corresponding to 431 * *map_id*. 432 * 433 * Return 434 * A new file descriptor (a nonnegative integer), or -1 if an 435 * error occurred (in which case, *errno* is set appropriately). 436 * 437 * BPF_OBJ_GET_INFO_BY_FD 438 * Description 439 * Obtain information about the eBPF object corresponding to 440 * *bpf_fd*. 441 * 442 * Populates up to *info_len* bytes of *info*, which will be in 443 * one of the following formats depending on the eBPF object type 444 * of *bpf_fd*: 445 * 446 * * **struct bpf_prog_info** 447 * * **struct bpf_map_info** 448 * * **struct bpf_btf_info** 449 * * **struct bpf_link_info** 450 * 451 * Return 452 * Returns zero on success. On error, -1 is returned and *errno* 453 * is set appropriately. 454 * 455 * BPF_PROG_QUERY 456 * Description 457 * Obtain information about eBPF programs associated with the 458 * specified *attach_type* hook. 459 * 460 * The *target_fd* must be a valid file descriptor for a kernel 461 * object which depends on the attach type of *attach_bpf_fd*: 462 * 463 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 464 * **BPF_PROG_TYPE_CGROUP_SKB**, 465 * **BPF_PROG_TYPE_CGROUP_SOCK**, 466 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 467 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 468 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 469 * **BPF_PROG_TYPE_SOCK_OPS** 470 * 471 * Control Group v2 hierarchy with the eBPF controller 472 * enabled. Requires the kernel to be compiled with 473 * **CONFIG_CGROUP_BPF**. 474 * 475 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 476 * 477 * Network namespace (eg /proc/self/ns/net). 478 * 479 * **BPF_PROG_TYPE_LIRC_MODE2** 480 * 481 * LIRC device path (eg /dev/lircN). Requires the kernel 482 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 483 * 484 * **BPF_PROG_QUERY** always fetches the number of programs 485 * attached and the *attach_flags* which were used to attach those 486 * programs. Additionally, if *prog_ids* is nonzero and the number 487 * of attached programs is less than *prog_cnt*, populates 488 * *prog_ids* with the eBPF program ids of the programs attached 489 * at *target_fd*. 490 * 491 * The following flags may alter the result: 492 * 493 * **BPF_F_QUERY_EFFECTIVE** 494 * Only return information regarding programs which are 495 * currently effective at the specified *target_fd*. 496 * 497 * Return 498 * Returns zero on success. On error, -1 is returned and *errno* 499 * is set appropriately. 500 * 501 * BPF_RAW_TRACEPOINT_OPEN 502 * Description 503 * Attach an eBPF program to a tracepoint *name* to access kernel 504 * internal arguments of the tracepoint in their raw form. 505 * 506 * The *prog_fd* must be a valid file descriptor associated with 507 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 508 * 509 * No ABI guarantees are made about the content of tracepoint 510 * arguments exposed to the corresponding eBPF program. 511 * 512 * Applying **close**\ (2) to the file descriptor returned by 513 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 514 * 515 * Return 516 * A new file descriptor (a nonnegative integer), or -1 if an 517 * error occurred (in which case, *errno* is set appropriately). 518 * 519 * BPF_BTF_LOAD 520 * Description 521 * Verify and load BPF Type Format (BTF) metadata into the kernel, 522 * returning a new file descriptor associated with the metadata. 523 * BTF is described in more detail at 524 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 525 * 526 * The *btf* parameter must point to valid memory providing 527 * *btf_size* bytes of BTF binary metadata. 528 * 529 * The returned file descriptor can be passed to other **bpf**\ () 530 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 531 * associate the BTF with those objects. 532 * 533 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 534 * parameters to specify a *btf_log_buf*, *btf_log_size* and 535 * *btf_log_level* which allow the kernel to return freeform log 536 * output regarding the BTF verification process. 537 * 538 * Return 539 * A new file descriptor (a nonnegative integer), or -1 if an 540 * error occurred (in which case, *errno* is set appropriately). 541 * 542 * BPF_BTF_GET_FD_BY_ID 543 * Description 544 * Open a file descriptor for the BPF Type Format (BTF) 545 * corresponding to *btf_id*. 546 * 547 * Return 548 * A new file descriptor (a nonnegative integer), or -1 if an 549 * error occurred (in which case, *errno* is set appropriately). 550 * 551 * BPF_TASK_FD_QUERY 552 * Description 553 * Obtain information about eBPF programs associated with the 554 * target process identified by *pid* and *fd*. 555 * 556 * If the *pid* and *fd* are associated with a tracepoint, kprobe 557 * or uprobe perf event, then the *prog_id* and *fd_type* will 558 * be populated with the eBPF program id and file descriptor type 559 * of type **bpf_task_fd_type**. If associated with a kprobe or 560 * uprobe, the *probe_offset* and *probe_addr* will also be 561 * populated. Optionally, if *buf* is provided, then up to 562 * *buf_len* bytes of *buf* will be populated with the name of 563 * the tracepoint, kprobe or uprobe. 564 * 565 * The resulting *prog_id* may be introspected in deeper detail 566 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 567 * 568 * Return 569 * Returns zero on success. On error, -1 is returned and *errno* 570 * is set appropriately. 571 * 572 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 573 * Description 574 * Look up an element with the given *key* in the map referred to 575 * by the file descriptor *fd*, and if found, delete the element. 576 * 577 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 578 * types, the *flags* argument needs to be set to 0, but for other 579 * map types, it may be specified as: 580 * 581 * **BPF_F_LOCK** 582 * Look up and delete the value of a spin-locked map 583 * without returning the lock. This must be specified if 584 * the elements contain a spinlock. 585 * 586 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 587 * implement this command as a "pop" operation, deleting the top 588 * element rather than one corresponding to *key*. 589 * The *key* and *key_len* parameters should be zeroed when 590 * issuing this operation for these map types. 591 * 592 * This command is only valid for the following map types: 593 * * **BPF_MAP_TYPE_QUEUE** 594 * * **BPF_MAP_TYPE_STACK** 595 * * **BPF_MAP_TYPE_HASH** 596 * * **BPF_MAP_TYPE_PERCPU_HASH** 597 * * **BPF_MAP_TYPE_LRU_HASH** 598 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 599 * 600 * Return 601 * Returns zero on success. On error, -1 is returned and *errno* 602 * is set appropriately. 603 * 604 * BPF_MAP_FREEZE 605 * Description 606 * Freeze the permissions of the specified map. 607 * 608 * Write permissions may be frozen by passing zero *flags*. 609 * Upon success, no future syscall invocations may alter the 610 * map state of *map_fd*. Write operations from eBPF programs 611 * are still possible for a frozen map. 612 * 613 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 614 * 615 * Return 616 * Returns zero on success. On error, -1 is returned and *errno* 617 * is set appropriately. 618 * 619 * BPF_BTF_GET_NEXT_ID 620 * Description 621 * Fetch the next BPF Type Format (BTF) object currently loaded 622 * into the kernel. 623 * 624 * Looks for the BTF object with an id greater than *start_id* 625 * and updates *next_id* on success. If no other BTF objects 626 * remain with ids higher than *start_id*, returns -1 and sets 627 * *errno* to **ENOENT**. 628 * 629 * Return 630 * Returns zero on success. On error, or when no id remains, -1 631 * is returned and *errno* is set appropriately. 632 * 633 * BPF_MAP_LOOKUP_BATCH 634 * Description 635 * Iterate and fetch multiple elements in a map. 636 * 637 * Two opaque values are used to manage batch operations, 638 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 639 * to NULL to begin the batched operation. After each subsequent 640 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 641 * *out_batch* as the *in_batch* for the next operation to 642 * continue iteration from the current point. Both *in_batch* and 643 * *out_batch* must point to memory large enough to hold a key, 644 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH, 645 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters 646 * must be at least 4 bytes wide regardless of key size. 647 * 648 * The *keys* and *values* are output parameters which must point 649 * to memory large enough to hold *count* items based on the key 650 * and value size of the map *map_fd*. The *keys* buffer must be 651 * of *key_size* * *count*. The *values* buffer must be of 652 * *value_size* * *count*. 653 * 654 * The *elem_flags* argument may be specified as one of the 655 * following: 656 * 657 * **BPF_F_LOCK** 658 * Look up the value of a spin-locked map without 659 * returning the lock. This must be specified if the 660 * elements contain a spinlock. 661 * 662 * On success, *count* elements from the map are copied into the 663 * user buffer, with the keys copied into *keys* and the values 664 * copied into the corresponding indices in *values*. 665 * 666 * If an error is returned and *errno* is not **EFAULT**, *count* 667 * is set to the number of successfully processed elements. 668 * 669 * Return 670 * Returns zero on success. On error, -1 is returned and *errno* 671 * is set appropriately. 672 * 673 * May set *errno* to **ENOSPC** to indicate that *keys* or 674 * *values* is too small to dump an entire bucket during 675 * iteration of a hash-based map type. 676 * 677 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 678 * Description 679 * Iterate and delete all elements in a map. 680 * 681 * This operation has the same behavior as 682 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 683 * 684 * * Every element that is successfully returned is also deleted 685 * from the map. This is at least *count* elements. Note that 686 * *count* is both an input and an output parameter. 687 * * Upon returning with *errno* set to **EFAULT**, up to 688 * *count* elements may be deleted without returning the keys 689 * and values of the deleted elements. 690 * 691 * Return 692 * Returns zero on success. On error, -1 is returned and *errno* 693 * is set appropriately. 694 * 695 * BPF_MAP_UPDATE_BATCH 696 * Description 697 * Update multiple elements in a map by *key*. 698 * 699 * The *keys* and *values* are input parameters which must point 700 * to memory large enough to hold *count* items based on the key 701 * and value size of the map *map_fd*. The *keys* buffer must be 702 * of *key_size* * *count*. The *values* buffer must be of 703 * *value_size* * *count*. 704 * 705 * Each element specified in *keys* is sequentially updated to the 706 * value in the corresponding index in *values*. The *in_batch* 707 * and *out_batch* parameters are ignored and should be zeroed. 708 * 709 * The *elem_flags* argument should be specified as one of the 710 * following: 711 * 712 * **BPF_ANY** 713 * Create new elements or update a existing elements. 714 * **BPF_NOEXIST** 715 * Create new elements only if they do not exist. 716 * **BPF_EXIST** 717 * Update existing elements. 718 * **BPF_F_LOCK** 719 * Update spin_lock-ed map elements. This must be 720 * specified if the map value contains a spinlock. 721 * 722 * On success, *count* elements from the map are updated. 723 * 724 * If an error is returned and *errno* is not **EFAULT**, *count* 725 * is set to the number of successfully processed elements. 726 * 727 * Return 728 * Returns zero on success. On error, -1 is returned and *errno* 729 * is set appropriately. 730 * 731 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 732 * **E2BIG**. **E2BIG** indicates that the number of elements in 733 * the map reached the *max_entries* limit specified at map 734 * creation time. 735 * 736 * May set *errno* to one of the following error codes under 737 * specific circumstances: 738 * 739 * **EEXIST** 740 * If *flags* specifies **BPF_NOEXIST** and the element 741 * with *key* already exists in the map. 742 * **ENOENT** 743 * If *flags* specifies **BPF_EXIST** and the element with 744 * *key* does not exist in the map. 745 * 746 * BPF_MAP_DELETE_BATCH 747 * Description 748 * Delete multiple elements in a map by *key*. 749 * 750 * The *keys* parameter is an input parameter which must point 751 * to memory large enough to hold *count* items based on the key 752 * size of the map *map_fd*, that is, *key_size* * *count*. 753 * 754 * Each element specified in *keys* is sequentially deleted. The 755 * *in_batch*, *out_batch*, and *values* parameters are ignored 756 * and should be zeroed. 757 * 758 * The *elem_flags* argument may be specified as one of the 759 * following: 760 * 761 * **BPF_F_LOCK** 762 * Look up the value of a spin-locked map without 763 * returning the lock. This must be specified if the 764 * elements contain a spinlock. 765 * 766 * On success, *count* elements from the map are updated. 767 * 768 * If an error is returned and *errno* is not **EFAULT**, *count* 769 * is set to the number of successfully processed elements. If 770 * *errno* is **EFAULT**, up to *count* elements may be been 771 * deleted. 772 * 773 * Return 774 * Returns zero on success. On error, -1 is returned and *errno* 775 * is set appropriately. 776 * 777 * BPF_LINK_CREATE 778 * Description 779 * Attach an eBPF program to a *target_fd* at the specified 780 * *attach_type* hook and return a file descriptor handle for 781 * managing the link. 782 * 783 * Return 784 * A new file descriptor (a nonnegative integer), or -1 if an 785 * error occurred (in which case, *errno* is set appropriately). 786 * 787 * BPF_LINK_UPDATE 788 * Description 789 * Update the eBPF program in the specified *link_fd* to 790 * *new_prog_fd*. 791 * 792 * Return 793 * Returns zero on success. On error, -1 is returned and *errno* 794 * is set appropriately. 795 * 796 * BPF_LINK_GET_FD_BY_ID 797 * Description 798 * Open a file descriptor for the eBPF Link corresponding to 799 * *link_id*. 800 * 801 * Return 802 * A new file descriptor (a nonnegative integer), or -1 if an 803 * error occurred (in which case, *errno* is set appropriately). 804 * 805 * BPF_LINK_GET_NEXT_ID 806 * Description 807 * Fetch the next eBPF link currently loaded into the kernel. 808 * 809 * Looks for the eBPF link with an id greater than *start_id* 810 * and updates *next_id* on success. If no other eBPF links 811 * remain with ids higher than *start_id*, returns -1 and sets 812 * *errno* to **ENOENT**. 813 * 814 * Return 815 * Returns zero on success. On error, or when no id remains, -1 816 * is returned and *errno* is set appropriately. 817 * 818 * BPF_ENABLE_STATS 819 * Description 820 * Enable eBPF runtime statistics gathering. 821 * 822 * Runtime statistics gathering for the eBPF runtime is disabled 823 * by default to minimize the corresponding performance overhead. 824 * This command enables statistics globally. 825 * 826 * Multiple programs may independently enable statistics. 827 * After gathering the desired statistics, eBPF runtime statistics 828 * may be disabled again by calling **close**\ (2) for the file 829 * descriptor returned by this function. Statistics will only be 830 * disabled system-wide when all outstanding file descriptors 831 * returned by prior calls for this subcommand are closed. 832 * 833 * Return 834 * A new file descriptor (a nonnegative integer), or -1 if an 835 * error occurred (in which case, *errno* is set appropriately). 836 * 837 * BPF_ITER_CREATE 838 * Description 839 * Create an iterator on top of the specified *link_fd* (as 840 * previously created using **BPF_LINK_CREATE**) and return a 841 * file descriptor that can be used to trigger the iteration. 842 * 843 * If the resulting file descriptor is pinned to the filesystem 844 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 845 * for that path will trigger the iterator to read kernel state 846 * using the eBPF program attached to *link_fd*. 847 * 848 * Return 849 * A new file descriptor (a nonnegative integer), or -1 if an 850 * error occurred (in which case, *errno* is set appropriately). 851 * 852 * BPF_LINK_DETACH 853 * Description 854 * Forcefully detach the specified *link_fd* from its 855 * corresponding attachment point. 856 * 857 * Return 858 * Returns zero on success. On error, -1 is returned and *errno* 859 * is set appropriately. 860 * 861 * BPF_PROG_BIND_MAP 862 * Description 863 * Bind a map to the lifetime of an eBPF program. 864 * 865 * The map identified by *map_fd* is bound to the program 866 * identified by *prog_fd* and only released when *prog_fd* is 867 * released. This may be used in cases where metadata should be 868 * associated with a program which otherwise does not contain any 869 * references to the map (for example, embedded in the eBPF 870 * program instructions). 871 * 872 * Return 873 * Returns zero on success. On error, -1 is returned and *errno* 874 * is set appropriately. 875 * 876 * BPF_TOKEN_CREATE 877 * Description 878 * Create BPF token with embedded information about what 879 * BPF-related functionality it allows: 880 * - a set of allowed bpf() syscall commands; 881 * - a set of allowed BPF map types to be created with 882 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 883 * - a set of allowed BPF program types and BPF program attach 884 * types to be loaded with BPF_PROG_LOAD command, if 885 * BPF_PROG_LOAD itself is allowed. 886 * 887 * BPF token is created (derived) from an instance of BPF FS, 888 * assuming it has necessary delegation mount options specified. 889 * This BPF token can be passed as an extra parameter to various 890 * bpf() syscall commands to grant BPF subsystem functionality to 891 * unprivileged processes. 892 * 893 * When created, BPF token is "associated" with the owning 894 * user namespace of BPF FS instance (super block) that it was 895 * derived from, and subsequent BPF operations performed with 896 * BPF token would be performing capabilities checks (i.e., 897 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 898 * that user namespace. Without BPF token, such capabilities 899 * have to be granted in init user namespace, making bpf() 900 * syscall incompatible with user namespace, for the most part. 901 * 902 * Return 903 * A new file descriptor (a nonnegative integer), or -1 if an 904 * error occurred (in which case, *errno* is set appropriately). 905 * 906 * NOTES 907 * eBPF objects (maps and programs) can be shared between processes. 908 * 909 * * After **fork**\ (2), the child inherits file descriptors 910 * referring to the same eBPF objects. 911 * * File descriptors referring to eBPF objects can be transferred over 912 * **unix**\ (7) domain sockets. 913 * * File descriptors referring to eBPF objects can be duplicated in the 914 * usual way, using **dup**\ (2) and similar calls. 915 * * File descriptors referring to eBPF objects can be pinned to the 916 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 917 * 918 * An eBPF object is deallocated only after all file descriptors referring 919 * to the object have been closed and no references remain pinned to the 920 * filesystem or attached (for example, bound to a program or device). 921 */ 922enum bpf_cmd { 923 BPF_MAP_CREATE, 924 BPF_MAP_LOOKUP_ELEM, 925 BPF_MAP_UPDATE_ELEM, 926 BPF_MAP_DELETE_ELEM, 927 BPF_MAP_GET_NEXT_KEY, 928 BPF_PROG_LOAD, 929 BPF_OBJ_PIN, 930 BPF_OBJ_GET, 931 BPF_PROG_ATTACH, 932 BPF_PROG_DETACH, 933 BPF_PROG_TEST_RUN, 934 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 935 BPF_PROG_GET_NEXT_ID, 936 BPF_MAP_GET_NEXT_ID, 937 BPF_PROG_GET_FD_BY_ID, 938 BPF_MAP_GET_FD_BY_ID, 939 BPF_OBJ_GET_INFO_BY_FD, 940 BPF_PROG_QUERY, 941 BPF_RAW_TRACEPOINT_OPEN, 942 BPF_BTF_LOAD, 943 BPF_BTF_GET_FD_BY_ID, 944 BPF_TASK_FD_QUERY, 945 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 946 BPF_MAP_FREEZE, 947 BPF_BTF_GET_NEXT_ID, 948 BPF_MAP_LOOKUP_BATCH, 949 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 950 BPF_MAP_UPDATE_BATCH, 951 BPF_MAP_DELETE_BATCH, 952 BPF_LINK_CREATE, 953 BPF_LINK_UPDATE, 954 BPF_LINK_GET_FD_BY_ID, 955 BPF_LINK_GET_NEXT_ID, 956 BPF_ENABLE_STATS, 957 BPF_ITER_CREATE, 958 BPF_LINK_DETACH, 959 BPF_PROG_BIND_MAP, 960 BPF_TOKEN_CREATE, 961 __MAX_BPF_CMD, 962}; 963 964enum bpf_map_type { 965 BPF_MAP_TYPE_UNSPEC, 966 BPF_MAP_TYPE_HASH, 967 BPF_MAP_TYPE_ARRAY, 968 BPF_MAP_TYPE_PROG_ARRAY, 969 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 970 BPF_MAP_TYPE_PERCPU_HASH, 971 BPF_MAP_TYPE_PERCPU_ARRAY, 972 BPF_MAP_TYPE_STACK_TRACE, 973 BPF_MAP_TYPE_CGROUP_ARRAY, 974 BPF_MAP_TYPE_LRU_HASH, 975 BPF_MAP_TYPE_LRU_PERCPU_HASH, 976 BPF_MAP_TYPE_LPM_TRIE, 977 BPF_MAP_TYPE_ARRAY_OF_MAPS, 978 BPF_MAP_TYPE_HASH_OF_MAPS, 979 BPF_MAP_TYPE_DEVMAP, 980 BPF_MAP_TYPE_SOCKMAP, 981 BPF_MAP_TYPE_CPUMAP, 982 BPF_MAP_TYPE_XSKMAP, 983 BPF_MAP_TYPE_SOCKHASH, 984 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 985 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 986 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 987 * both cgroup-attached and other progs and supports all functionality 988 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 989 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 990 */ 991 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 992 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 993 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 994 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 995 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 996 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 997 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 998 * deprecated. 999 */ 1000 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 1001 BPF_MAP_TYPE_QUEUE, 1002 BPF_MAP_TYPE_STACK, 1003 BPF_MAP_TYPE_SK_STORAGE, 1004 BPF_MAP_TYPE_DEVMAP_HASH, 1005 BPF_MAP_TYPE_STRUCT_OPS, 1006 BPF_MAP_TYPE_RINGBUF, 1007 BPF_MAP_TYPE_INODE_STORAGE, 1008 BPF_MAP_TYPE_TASK_STORAGE, 1009 BPF_MAP_TYPE_BLOOM_FILTER, 1010 BPF_MAP_TYPE_USER_RINGBUF, 1011 BPF_MAP_TYPE_CGRP_STORAGE, 1012 BPF_MAP_TYPE_ARENA, 1013 __MAX_BPF_MAP_TYPE 1014}; 1015 1016/* Note that tracing related programs such as 1017 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1018 * are not subject to a stable API since kernel internal data 1019 * structures can change from release to release and may 1020 * therefore break existing tracing BPF programs. Tracing BPF 1021 * programs correspond to /a/ specific kernel which is to be 1022 * analyzed, and not /a/ specific kernel /and/ all future ones. 1023 */ 1024enum bpf_prog_type { 1025 BPF_PROG_TYPE_UNSPEC, 1026 BPF_PROG_TYPE_SOCKET_FILTER, 1027 BPF_PROG_TYPE_KPROBE, 1028 BPF_PROG_TYPE_SCHED_CLS, 1029 BPF_PROG_TYPE_SCHED_ACT, 1030 BPF_PROG_TYPE_TRACEPOINT, 1031 BPF_PROG_TYPE_XDP, 1032 BPF_PROG_TYPE_PERF_EVENT, 1033 BPF_PROG_TYPE_CGROUP_SKB, 1034 BPF_PROG_TYPE_CGROUP_SOCK, 1035 BPF_PROG_TYPE_LWT_IN, 1036 BPF_PROG_TYPE_LWT_OUT, 1037 BPF_PROG_TYPE_LWT_XMIT, 1038 BPF_PROG_TYPE_SOCK_OPS, 1039 BPF_PROG_TYPE_SK_SKB, 1040 BPF_PROG_TYPE_CGROUP_DEVICE, 1041 BPF_PROG_TYPE_SK_MSG, 1042 BPF_PROG_TYPE_RAW_TRACEPOINT, 1043 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1044 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1045 BPF_PROG_TYPE_LIRC_MODE2, 1046 BPF_PROG_TYPE_SK_REUSEPORT, 1047 BPF_PROG_TYPE_FLOW_DISSECTOR, 1048 BPF_PROG_TYPE_CGROUP_SYSCTL, 1049 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1050 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1051 BPF_PROG_TYPE_TRACING, 1052 BPF_PROG_TYPE_STRUCT_OPS, 1053 BPF_PROG_TYPE_EXT, 1054 BPF_PROG_TYPE_LSM, 1055 BPF_PROG_TYPE_SK_LOOKUP, 1056 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1057 BPF_PROG_TYPE_NETFILTER, 1058 __MAX_BPF_PROG_TYPE 1059}; 1060 1061enum bpf_attach_type { 1062 BPF_CGROUP_INET_INGRESS, 1063 BPF_CGROUP_INET_EGRESS, 1064 BPF_CGROUP_INET_SOCK_CREATE, 1065 BPF_CGROUP_SOCK_OPS, 1066 BPF_SK_SKB_STREAM_PARSER, 1067 BPF_SK_SKB_STREAM_VERDICT, 1068 BPF_CGROUP_DEVICE, 1069 BPF_SK_MSG_VERDICT, 1070 BPF_CGROUP_INET4_BIND, 1071 BPF_CGROUP_INET6_BIND, 1072 BPF_CGROUP_INET4_CONNECT, 1073 BPF_CGROUP_INET6_CONNECT, 1074 BPF_CGROUP_INET4_POST_BIND, 1075 BPF_CGROUP_INET6_POST_BIND, 1076 BPF_CGROUP_UDP4_SENDMSG, 1077 BPF_CGROUP_UDP6_SENDMSG, 1078 BPF_LIRC_MODE2, 1079 BPF_FLOW_DISSECTOR, 1080 BPF_CGROUP_SYSCTL, 1081 BPF_CGROUP_UDP4_RECVMSG, 1082 BPF_CGROUP_UDP6_RECVMSG, 1083 BPF_CGROUP_GETSOCKOPT, 1084 BPF_CGROUP_SETSOCKOPT, 1085 BPF_TRACE_RAW_TP, 1086 BPF_TRACE_FENTRY, 1087 BPF_TRACE_FEXIT, 1088 BPF_MODIFY_RETURN, 1089 BPF_LSM_MAC, 1090 BPF_TRACE_ITER, 1091 BPF_CGROUP_INET4_GETPEERNAME, 1092 BPF_CGROUP_INET6_GETPEERNAME, 1093 BPF_CGROUP_INET4_GETSOCKNAME, 1094 BPF_CGROUP_INET6_GETSOCKNAME, 1095 BPF_XDP_DEVMAP, 1096 BPF_CGROUP_INET_SOCK_RELEASE, 1097 BPF_XDP_CPUMAP, 1098 BPF_SK_LOOKUP, 1099 BPF_XDP, 1100 BPF_SK_SKB_VERDICT, 1101 BPF_SK_REUSEPORT_SELECT, 1102 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1103 BPF_PERF_EVENT, 1104 BPF_TRACE_KPROBE_MULTI, 1105 BPF_LSM_CGROUP, 1106 BPF_STRUCT_OPS, 1107 BPF_NETFILTER, 1108 BPF_TCX_INGRESS, 1109 BPF_TCX_EGRESS, 1110 BPF_TRACE_UPROBE_MULTI, 1111 BPF_CGROUP_UNIX_CONNECT, 1112 BPF_CGROUP_UNIX_SENDMSG, 1113 BPF_CGROUP_UNIX_RECVMSG, 1114 BPF_CGROUP_UNIX_GETPEERNAME, 1115 BPF_CGROUP_UNIX_GETSOCKNAME, 1116 BPF_NETKIT_PRIMARY, 1117 BPF_NETKIT_PEER, 1118 __MAX_BPF_ATTACH_TYPE 1119}; 1120 1121#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1122 1123enum bpf_link_type { 1124 BPF_LINK_TYPE_UNSPEC = 0, 1125 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1126 BPF_LINK_TYPE_TRACING = 2, 1127 BPF_LINK_TYPE_CGROUP = 3, 1128 BPF_LINK_TYPE_ITER = 4, 1129 BPF_LINK_TYPE_NETNS = 5, 1130 BPF_LINK_TYPE_XDP = 6, 1131 BPF_LINK_TYPE_PERF_EVENT = 7, 1132 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1133 BPF_LINK_TYPE_STRUCT_OPS = 9, 1134 BPF_LINK_TYPE_NETFILTER = 10, 1135 BPF_LINK_TYPE_TCX = 11, 1136 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1137 BPF_LINK_TYPE_NETKIT = 13, 1138 __MAX_BPF_LINK_TYPE, 1139}; 1140 1141#define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1142 1143enum bpf_perf_event_type { 1144 BPF_PERF_EVENT_UNSPEC = 0, 1145 BPF_PERF_EVENT_UPROBE = 1, 1146 BPF_PERF_EVENT_URETPROBE = 2, 1147 BPF_PERF_EVENT_KPROBE = 3, 1148 BPF_PERF_EVENT_KRETPROBE = 4, 1149 BPF_PERF_EVENT_TRACEPOINT = 5, 1150 BPF_PERF_EVENT_EVENT = 6, 1151}; 1152 1153/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1154 * 1155 * NONE(default): No further bpf programs allowed in the subtree. 1156 * 1157 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1158 * the program in this cgroup yields to sub-cgroup program. 1159 * 1160 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1161 * that cgroup program gets run in addition to the program in this cgroup. 1162 * 1163 * Only one program is allowed to be attached to a cgroup with 1164 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1165 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1166 * release old program and attach the new one. Attach flags has to match. 1167 * 1168 * Multiple programs are allowed to be attached to a cgroup with 1169 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1170 * (those that were attached first, run first) 1171 * The programs of sub-cgroup are executed first, then programs of 1172 * this cgroup and then programs of parent cgroup. 1173 * When children program makes decision (like picking TCP CA or sock bind) 1174 * parent program has a chance to override it. 1175 * 1176 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1177 * programs for a cgroup. Though it's possible to replace an old program at 1178 * any position by also specifying BPF_F_REPLACE flag and position itself in 1179 * replace_bpf_fd attribute. Old program at this position will be released. 1180 * 1181 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1182 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1183 * Ex1: 1184 * cgrp1 (MULTI progs A, B) -> 1185 * cgrp2 (OVERRIDE prog C) -> 1186 * cgrp3 (MULTI prog D) -> 1187 * cgrp4 (OVERRIDE prog E) -> 1188 * cgrp5 (NONE prog F) 1189 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1190 * if prog F is detached, the execution is E,D,A,B 1191 * if prog F and D are detached, the execution is E,A,B 1192 * if prog F, E and D are detached, the execution is C,A,B 1193 * 1194 * All eligible programs are executed regardless of return code from 1195 * earlier programs. 1196 */ 1197#define BPF_F_ALLOW_OVERRIDE (1U << 0) 1198#define BPF_F_ALLOW_MULTI (1U << 1) 1199/* Generic attachment flags. */ 1200#define BPF_F_REPLACE (1U << 2) 1201#define BPF_F_BEFORE (1U << 3) 1202#define BPF_F_AFTER (1U << 4) 1203#define BPF_F_ID (1U << 5) 1204#define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1205 1206/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1207 * verifier will perform strict alignment checking as if the kernel 1208 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1209 * and NET_IP_ALIGN defined to 2. 1210 */ 1211#define BPF_F_STRICT_ALIGNMENT (1U << 0) 1212 1213/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1214 * verifier will allow any alignment whatsoever. On platforms 1215 * with strict alignment requirements for loads ands stores (such 1216 * as sparc and mips) the verifier validates that all loads and 1217 * stores provably follow this requirement. This flag turns that 1218 * checking and enforcement off. 1219 * 1220 * It is mostly used for testing when we want to validate the 1221 * context and memory access aspects of the verifier, but because 1222 * of an unaligned access the alignment check would trigger before 1223 * the one we are interested in. 1224 */ 1225#define BPF_F_ANY_ALIGNMENT (1U << 1) 1226 1227/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1228 * Verifier does sub-register def/use analysis and identifies instructions whose 1229 * def only matters for low 32-bit, high 32-bit is never referenced later 1230 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1231 * that it is safe to ignore clearing high 32-bit for these instructions. This 1232 * saves some back-ends a lot of code-gen. However such optimization is not 1233 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1234 * hence hasn't used verifier's analysis result. But, we really want to have a 1235 * way to be able to verify the correctness of the described optimization on 1236 * x86_64 on which testsuites are frequently exercised. 1237 * 1238 * So, this flag is introduced. Once it is set, verifier will randomize high 1239 * 32-bit for those instructions who has been identified as safe to ignore them. 1240 * Then, if verifier is not doing correct analysis, such randomization will 1241 * regress tests to expose bugs. 1242 */ 1243#define BPF_F_TEST_RND_HI32 (1U << 2) 1244 1245/* The verifier internal test flag. Behavior is undefined */ 1246#define BPF_F_TEST_STATE_FREQ (1U << 3) 1247 1248/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1249 * restrict map and helper usage for such programs. Sleepable BPF programs can 1250 * only be attached to hooks where kernel execution context allows sleeping. 1251 * Such programs are allowed to use helpers that may sleep like 1252 * bpf_copy_from_user(). 1253 */ 1254#define BPF_F_SLEEPABLE (1U << 4) 1255 1256/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1257 * fully support xdp frags. 1258 */ 1259#define BPF_F_XDP_HAS_FRAGS (1U << 5) 1260 1261/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1262 * program becomes device-bound but can access XDP metadata. 1263 */ 1264#define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1265 1266/* The verifier internal test flag. Behavior is undefined */ 1267#define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1268 1269/* link_create.kprobe_multi.flags used in LINK_CREATE command for 1270 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1271 */ 1272enum { 1273 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1274}; 1275 1276/* link_create.uprobe_multi.flags used in LINK_CREATE command for 1277 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1278 */ 1279enum { 1280 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1281}; 1282 1283/* link_create.netfilter.flags used in LINK_CREATE command for 1284 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1285 */ 1286#define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1287 1288/* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1289 * the following extensions: 1290 * 1291 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1292 * insn[0].imm: map fd or fd_idx 1293 * insn[1].imm: 0 1294 * insn[0].off: 0 1295 * insn[1].off: 0 1296 * ldimm64 rewrite: address of map 1297 * verifier type: CONST_PTR_TO_MAP 1298 */ 1299#define BPF_PSEUDO_MAP_FD 1 1300#define BPF_PSEUDO_MAP_IDX 5 1301 1302/* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1303 * insn[0].imm: map fd or fd_idx 1304 * insn[1].imm: offset into value 1305 * insn[0].off: 0 1306 * insn[1].off: 0 1307 * ldimm64 rewrite: address of map[0]+offset 1308 * verifier type: PTR_TO_MAP_VALUE 1309 */ 1310#define BPF_PSEUDO_MAP_VALUE 2 1311#define BPF_PSEUDO_MAP_IDX_VALUE 6 1312 1313/* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1314 * insn[0].imm: kernel btd id of VAR 1315 * insn[1].imm: 0 1316 * insn[0].off: 0 1317 * insn[1].off: 0 1318 * ldimm64 rewrite: address of the kernel variable 1319 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1320 * is struct/union. 1321 */ 1322#define BPF_PSEUDO_BTF_ID 3 1323/* insn[0].src_reg: BPF_PSEUDO_FUNC 1324 * insn[0].imm: insn offset to the func 1325 * insn[1].imm: 0 1326 * insn[0].off: 0 1327 * insn[1].off: 0 1328 * ldimm64 rewrite: address of the function 1329 * verifier type: PTR_TO_FUNC. 1330 */ 1331#define BPF_PSEUDO_FUNC 4 1332 1333/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1334 * offset to another bpf function 1335 */ 1336#define BPF_PSEUDO_CALL 1 1337/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1338 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1339 */ 1340#define BPF_PSEUDO_KFUNC_CALL 2 1341 1342enum bpf_addr_space_cast { 1343 BPF_ADDR_SPACE_CAST = 1, 1344}; 1345 1346/* flags for BPF_MAP_UPDATE_ELEM command */ 1347enum { 1348 BPF_ANY = 0, /* create new element or update existing */ 1349 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1350 BPF_EXIST = 2, /* update existing element */ 1351 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1352}; 1353 1354/* flags for BPF_MAP_CREATE command */ 1355enum { 1356 BPF_F_NO_PREALLOC = (1U << 0), 1357/* Instead of having one common LRU list in the 1358 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1359 * which can scale and perform better. 1360 * Note, the LRU nodes (including free nodes) cannot be moved 1361 * across different LRU lists. 1362 */ 1363 BPF_F_NO_COMMON_LRU = (1U << 1), 1364/* Specify numa node during map creation */ 1365 BPF_F_NUMA_NODE = (1U << 2), 1366 1367/* Flags for accessing BPF object from syscall side. */ 1368 BPF_F_RDONLY = (1U << 3), 1369 BPF_F_WRONLY = (1U << 4), 1370 1371/* Flag for stack_map, store build_id+offset instead of pointer */ 1372 BPF_F_STACK_BUILD_ID = (1U << 5), 1373 1374/* Zero-initialize hash function seed. This should only be used for testing. */ 1375 BPF_F_ZERO_SEED = (1U << 6), 1376 1377/* Flags for accessing BPF object from program side. */ 1378 BPF_F_RDONLY_PROG = (1U << 7), 1379 BPF_F_WRONLY_PROG = (1U << 8), 1380 1381/* Clone map from listener for newly accepted socket */ 1382 BPF_F_CLONE = (1U << 9), 1383 1384/* Enable memory-mapping BPF map */ 1385 BPF_F_MMAPABLE = (1U << 10), 1386 1387/* Share perf_event among processes */ 1388 BPF_F_PRESERVE_ELEMS = (1U << 11), 1389 1390/* Create a map that is suitable to be an inner map with dynamic max entries */ 1391 BPF_F_INNER_MAP = (1U << 12), 1392 1393/* Create a map that will be registered/unregesitered by the backed bpf_link */ 1394 BPF_F_LINK = (1U << 13), 1395 1396/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1397 BPF_F_PATH_FD = (1U << 14), 1398 1399/* Flag for value_type_btf_obj_fd, the fd is available */ 1400 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1401 1402/* BPF token FD is passed in a corresponding command's token_fd field */ 1403 BPF_F_TOKEN_FD = (1U << 16), 1404 1405/* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */ 1406 BPF_F_SEGV_ON_FAULT = (1U << 17), 1407 1408/* Do not translate kernel bpf_arena pointers to user pointers */ 1409 BPF_F_NO_USER_CONV = (1U << 18), 1410}; 1411 1412/* Flags for BPF_PROG_QUERY. */ 1413 1414/* Query effective (directly attached + inherited from ancestor cgroups) 1415 * programs that will be executed for events within a cgroup. 1416 * attach_flags with this flag are always returned 0. 1417 */ 1418#define BPF_F_QUERY_EFFECTIVE (1U << 0) 1419 1420/* Flags for BPF_PROG_TEST_RUN */ 1421 1422/* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1423#define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1424/* If set, XDP frames will be transmitted after processing */ 1425#define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1426 1427/* type for BPF_ENABLE_STATS */ 1428enum bpf_stats_type { 1429 /* enabled run_time_ns and run_cnt */ 1430 BPF_STATS_RUN_TIME = 0, 1431}; 1432 1433enum bpf_stack_build_id_status { 1434 /* user space need an empty entry to identify end of a trace */ 1435 BPF_STACK_BUILD_ID_EMPTY = 0, 1436 /* with valid build_id and offset */ 1437 BPF_STACK_BUILD_ID_VALID = 1, 1438 /* couldn't get build_id, fallback to ip */ 1439 BPF_STACK_BUILD_ID_IP = 2, 1440}; 1441 1442#define BPF_BUILD_ID_SIZE 20 1443struct bpf_stack_build_id { 1444 __s32 status; 1445 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1446 union { 1447 __u64 offset; 1448 __u64 ip; 1449 }; 1450}; 1451 1452#define BPF_OBJ_NAME_LEN 16U 1453 1454union bpf_attr { 1455 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1456 __u32 map_type; /* one of enum bpf_map_type */ 1457 __u32 key_size; /* size of key in bytes */ 1458 __u32 value_size; /* size of value in bytes */ 1459 __u32 max_entries; /* max number of entries in a map */ 1460 __u32 map_flags; /* BPF_MAP_CREATE related 1461 * flags defined above. 1462 */ 1463 __u32 inner_map_fd; /* fd pointing to the inner map */ 1464 __u32 numa_node; /* numa node (effective only if 1465 * BPF_F_NUMA_NODE is set). 1466 */ 1467 char map_name[BPF_OBJ_NAME_LEN]; 1468 __u32 map_ifindex; /* ifindex of netdev to create on */ 1469 __u32 btf_fd; /* fd pointing to a BTF type data */ 1470 __u32 btf_key_type_id; /* BTF type_id of the key */ 1471 __u32 btf_value_type_id; /* BTF type_id of the value */ 1472 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1473 * struct stored as the 1474 * map value 1475 */ 1476 /* Any per-map-type extra fields 1477 * 1478 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1479 * number of hash functions (if 0, the bloom filter will default 1480 * to using 5 hash functions). 1481 * 1482 * BPF_MAP_TYPE_ARENA - contains the address where user space 1483 * is going to mmap() the arena. It has to be page aligned. 1484 */ 1485 __u64 map_extra; 1486 1487 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1488 * type data for 1489 * btf_vmlinux_value_type_id. 1490 */ 1491 /* BPF token FD to use with BPF_MAP_CREATE operation. 1492 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1493 */ 1494 __s32 map_token_fd; 1495 }; 1496 1497 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1498 __u32 map_fd; 1499 __aligned_u64 key; 1500 union { 1501 __aligned_u64 value; 1502 __aligned_u64 next_key; 1503 }; 1504 __u64 flags; 1505 }; 1506 1507 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1508 __aligned_u64 in_batch; /* start batch, 1509 * NULL to start from beginning 1510 */ 1511 __aligned_u64 out_batch; /* output: next start batch */ 1512 __aligned_u64 keys; 1513 __aligned_u64 values; 1514 __u32 count; /* input/output: 1515 * input: # of key/value 1516 * elements 1517 * output: # of filled elements 1518 */ 1519 __u32 map_fd; 1520 __u64 elem_flags; 1521 __u64 flags; 1522 } batch; 1523 1524 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1525 __u32 prog_type; /* one of enum bpf_prog_type */ 1526 __u32 insn_cnt; 1527 __aligned_u64 insns; 1528 __aligned_u64 license; 1529 __u32 log_level; /* verbosity level of verifier */ 1530 __u32 log_size; /* size of user buffer */ 1531 __aligned_u64 log_buf; /* user supplied buffer */ 1532 __u32 kern_version; /* not used */ 1533 __u32 prog_flags; 1534 char prog_name[BPF_OBJ_NAME_LEN]; 1535 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1536 /* For some prog types expected attach type must be known at 1537 * load time to verify attach type specific parts of prog 1538 * (context accesses, allowed helpers, etc). 1539 */ 1540 __u32 expected_attach_type; 1541 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1542 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1543 __aligned_u64 func_info; /* func info */ 1544 __u32 func_info_cnt; /* number of bpf_func_info records */ 1545 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1546 __aligned_u64 line_info; /* line info */ 1547 __u32 line_info_cnt; /* number of bpf_line_info records */ 1548 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1549 union { 1550 /* valid prog_fd to attach to bpf prog */ 1551 __u32 attach_prog_fd; 1552 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1553 __u32 attach_btf_obj_fd; 1554 }; 1555 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1556 __aligned_u64 fd_array; /* array of FDs */ 1557 __aligned_u64 core_relos; 1558 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1559 /* output: actual total log contents size (including termintaing zero). 1560 * It could be both larger than original log_size (if log was 1561 * truncated), or smaller (if log buffer wasn't filled completely). 1562 */ 1563 __u32 log_true_size; 1564 /* BPF token FD to use with BPF_PROG_LOAD operation. 1565 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1566 */ 1567 __s32 prog_token_fd; 1568 }; 1569 1570 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1571 __aligned_u64 pathname; 1572 __u32 bpf_fd; 1573 __u32 file_flags; 1574 /* Same as dirfd in openat() syscall; see openat(2) 1575 * manpage for details of path FD and pathname semantics; 1576 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1577 * file_flags field, otherwise it should be set to zero; 1578 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1579 */ 1580 __s32 path_fd; 1581 }; 1582 1583 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1584 union { 1585 __u32 target_fd; /* target object to attach to or ... */ 1586 __u32 target_ifindex; /* target ifindex */ 1587 }; 1588 __u32 attach_bpf_fd; 1589 __u32 attach_type; 1590 __u32 attach_flags; 1591 __u32 replace_bpf_fd; 1592 union { 1593 __u32 relative_fd; 1594 __u32 relative_id; 1595 }; 1596 __u64 expected_revision; 1597 }; 1598 1599 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1600 __u32 prog_fd; 1601 __u32 retval; 1602 __u32 data_size_in; /* input: len of data_in */ 1603 __u32 data_size_out; /* input/output: len of data_out 1604 * returns ENOSPC if data_out 1605 * is too small. 1606 */ 1607 __aligned_u64 data_in; 1608 __aligned_u64 data_out; 1609 __u32 repeat; 1610 __u32 duration; 1611 __u32 ctx_size_in; /* input: len of ctx_in */ 1612 __u32 ctx_size_out; /* input/output: len of ctx_out 1613 * returns ENOSPC if ctx_out 1614 * is too small. 1615 */ 1616 __aligned_u64 ctx_in; 1617 __aligned_u64 ctx_out; 1618 __u32 flags; 1619 __u32 cpu; 1620 __u32 batch_size; 1621 } test; 1622 1623 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1624 union { 1625 __u32 start_id; 1626 __u32 prog_id; 1627 __u32 map_id; 1628 __u32 btf_id; 1629 __u32 link_id; 1630 }; 1631 __u32 next_id; 1632 __u32 open_flags; 1633 }; 1634 1635 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1636 __u32 bpf_fd; 1637 __u32 info_len; 1638 __aligned_u64 info; 1639 } info; 1640 1641 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1642 union { 1643 __u32 target_fd; /* target object to query or ... */ 1644 __u32 target_ifindex; /* target ifindex */ 1645 }; 1646 __u32 attach_type; 1647 __u32 query_flags; 1648 __u32 attach_flags; 1649 __aligned_u64 prog_ids; 1650 union { 1651 __u32 prog_cnt; 1652 __u32 count; 1653 }; 1654 __u32 :32; 1655 /* output: per-program attach_flags. 1656 * not allowed to be set during effective query. 1657 */ 1658 __aligned_u64 prog_attach_flags; 1659 __aligned_u64 link_ids; 1660 __aligned_u64 link_attach_flags; 1661 __u64 revision; 1662 } query; 1663 1664 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1665 __u64 name; 1666 __u32 prog_fd; 1667 } raw_tracepoint; 1668 1669 struct { /* anonymous struct for BPF_BTF_LOAD */ 1670 __aligned_u64 btf; 1671 __aligned_u64 btf_log_buf; 1672 __u32 btf_size; 1673 __u32 btf_log_size; 1674 __u32 btf_log_level; 1675 /* output: actual total log contents size (including termintaing zero). 1676 * It could be both larger than original log_size (if log was 1677 * truncated), or smaller (if log buffer wasn't filled completely). 1678 */ 1679 __u32 btf_log_true_size; 1680 __u32 btf_flags; 1681 /* BPF token FD to use with BPF_BTF_LOAD operation. 1682 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1683 */ 1684 __s32 btf_token_fd; 1685 }; 1686 1687 struct { 1688 __u32 pid; /* input: pid */ 1689 __u32 fd; /* input: fd */ 1690 __u32 flags; /* input: flags */ 1691 __u32 buf_len; /* input/output: buf len */ 1692 __aligned_u64 buf; /* input/output: 1693 * tp_name for tracepoint 1694 * symbol for kprobe 1695 * filename for uprobe 1696 */ 1697 __u32 prog_id; /* output: prod_id */ 1698 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1699 __u64 probe_offset; /* output: probe_offset */ 1700 __u64 probe_addr; /* output: probe_addr */ 1701 } task_fd_query; 1702 1703 struct { /* struct used by BPF_LINK_CREATE command */ 1704 union { 1705 __u32 prog_fd; /* eBPF program to attach */ 1706 __u32 map_fd; /* struct_ops to attach */ 1707 }; 1708 union { 1709 __u32 target_fd; /* target object to attach to or ... */ 1710 __u32 target_ifindex; /* target ifindex */ 1711 }; 1712 __u32 attach_type; /* attach type */ 1713 __u32 flags; /* extra flags */ 1714 union { 1715 __u32 target_btf_id; /* btf_id of target to attach to */ 1716 struct { 1717 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1718 __u32 iter_info_len; /* iter_info length */ 1719 }; 1720 struct { 1721 /* black box user-provided value passed through 1722 * to BPF program at the execution time and 1723 * accessible through bpf_get_attach_cookie() BPF helper 1724 */ 1725 __u64 bpf_cookie; 1726 } perf_event; 1727 struct { 1728 __u32 flags; 1729 __u32 cnt; 1730 __aligned_u64 syms; 1731 __aligned_u64 addrs; 1732 __aligned_u64 cookies; 1733 } kprobe_multi; 1734 struct { 1735 /* this is overlaid with the target_btf_id above. */ 1736 __u32 target_btf_id; 1737 /* black box user-provided value passed through 1738 * to BPF program at the execution time and 1739 * accessible through bpf_get_attach_cookie() BPF helper 1740 */ 1741 __u64 cookie; 1742 } tracing; 1743 struct { 1744 __u32 pf; 1745 __u32 hooknum; 1746 __s32 priority; 1747 __u32 flags; 1748 } netfilter; 1749 struct { 1750 union { 1751 __u32 relative_fd; 1752 __u32 relative_id; 1753 }; 1754 __u64 expected_revision; 1755 } tcx; 1756 struct { 1757 __aligned_u64 path; 1758 __aligned_u64 offsets; 1759 __aligned_u64 ref_ctr_offsets; 1760 __aligned_u64 cookies; 1761 __u32 cnt; 1762 __u32 flags; 1763 __u32 pid; 1764 } uprobe_multi; 1765 struct { 1766 union { 1767 __u32 relative_fd; 1768 __u32 relative_id; 1769 }; 1770 __u64 expected_revision; 1771 } netkit; 1772 }; 1773 } link_create; 1774 1775 struct { /* struct used by BPF_LINK_UPDATE command */ 1776 __u32 link_fd; /* link fd */ 1777 union { 1778 /* new program fd to update link with */ 1779 __u32 new_prog_fd; 1780 /* new struct_ops map fd to update link with */ 1781 __u32 new_map_fd; 1782 }; 1783 __u32 flags; /* extra flags */ 1784 union { 1785 /* expected link's program fd; is specified only if 1786 * BPF_F_REPLACE flag is set in flags. 1787 */ 1788 __u32 old_prog_fd; 1789 /* expected link's map fd; is specified only 1790 * if BPF_F_REPLACE flag is set. 1791 */ 1792 __u32 old_map_fd; 1793 }; 1794 } link_update; 1795 1796 struct { 1797 __u32 link_fd; 1798 } link_detach; 1799 1800 struct { /* struct used by BPF_ENABLE_STATS command */ 1801 __u32 type; 1802 } enable_stats; 1803 1804 struct { /* struct used by BPF_ITER_CREATE command */ 1805 __u32 link_fd; 1806 __u32 flags; 1807 } iter_create; 1808 1809 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1810 __u32 prog_fd; 1811 __u32 map_fd; 1812 __u32 flags; /* extra flags */ 1813 } prog_bind_map; 1814 1815 struct { /* struct used by BPF_TOKEN_CREATE command */ 1816 __u32 flags; 1817 __u32 bpffs_fd; 1818 } token_create; 1819 1820} __attribute__((aligned(8))); 1821 1822/* The description below is an attempt at providing documentation to eBPF 1823 * developers about the multiple available eBPF helper functions. It can be 1824 * parsed and used to produce a manual page. The workflow is the following, 1825 * and requires the rst2man utility: 1826 * 1827 * $ ./scripts/bpf_doc.py \ 1828 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1829 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1830 * $ man /tmp/bpf-helpers.7 1831 * 1832 * Note that in order to produce this external documentation, some RST 1833 * formatting is used in the descriptions to get "bold" and "italics" in 1834 * manual pages. Also note that the few trailing white spaces are 1835 * intentional, removing them would break paragraphs for rst2man. 1836 * 1837 * Start of BPF helper function descriptions: 1838 * 1839 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1840 * Description 1841 * Perform a lookup in *map* for an entry associated to *key*. 1842 * Return 1843 * Map value associated to *key*, or **NULL** if no entry was 1844 * found. 1845 * 1846 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1847 * Description 1848 * Add or update the value of the entry associated to *key* in 1849 * *map* with *value*. *flags* is one of: 1850 * 1851 * **BPF_NOEXIST** 1852 * The entry for *key* must not exist in the map. 1853 * **BPF_EXIST** 1854 * The entry for *key* must already exist in the map. 1855 * **BPF_ANY** 1856 * No condition on the existence of the entry for *key*. 1857 * 1858 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1859 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1860 * elements always exist), the helper would return an error. 1861 * Return 1862 * 0 on success, or a negative error in case of failure. 1863 * 1864 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1865 * Description 1866 * Delete entry with *key* from *map*. 1867 * Return 1868 * 0 on success, or a negative error in case of failure. 1869 * 1870 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1871 * Description 1872 * For tracing programs, safely attempt to read *size* bytes from 1873 * kernel space address *unsafe_ptr* and store the data in *dst*. 1874 * 1875 * Generally, use **bpf_probe_read_user**\ () or 1876 * **bpf_probe_read_kernel**\ () instead. 1877 * Return 1878 * 0 on success, or a negative error in case of failure. 1879 * 1880 * u64 bpf_ktime_get_ns(void) 1881 * Description 1882 * Return the time elapsed since system boot, in nanoseconds. 1883 * Does not include time the system was suspended. 1884 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1885 * Return 1886 * Current *ktime*. 1887 * 1888 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1889 * Description 1890 * This helper is a "printk()-like" facility for debugging. It 1891 * prints a message defined by format *fmt* (of size *fmt_size*) 1892 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1893 * available. It can take up to three additional **u64** 1894 * arguments (as an eBPF helpers, the total number of arguments is 1895 * limited to five). 1896 * 1897 * Each time the helper is called, it appends a line to the trace. 1898 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1899 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1900 * The format of the trace is customizable, and the exact output 1901 * one will get depends on the options set in 1902 * *\/sys/kernel/tracing/trace_options* (see also the 1903 * *README* file under the same directory). However, it usually 1904 * defaults to something like: 1905 * 1906 * :: 1907 * 1908 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1909 * 1910 * In the above: 1911 * 1912 * * ``telnet`` is the name of the current task. 1913 * * ``470`` is the PID of the current task. 1914 * * ``001`` is the CPU number on which the task is 1915 * running. 1916 * * In ``.N..``, each character refers to a set of 1917 * options (whether irqs are enabled, scheduling 1918 * options, whether hard/softirqs are running, level of 1919 * preempt_disabled respectively). **N** means that 1920 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1921 * are set. 1922 * * ``419421.045894`` is a timestamp. 1923 * * ``0x00000001`` is a fake value used by BPF for the 1924 * instruction pointer register. 1925 * * ``<formatted msg>`` is the message formatted with 1926 * *fmt*. 1927 * 1928 * The conversion specifiers supported by *fmt* are similar, but 1929 * more limited than for printk(). They are **%d**, **%i**, 1930 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1931 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1932 * of field, padding with zeroes, etc.) is available, and the 1933 * helper will return **-EINVAL** (but print nothing) if it 1934 * encounters an unknown specifier. 1935 * 1936 * Also, note that **bpf_trace_printk**\ () is slow, and should 1937 * only be used for debugging purposes. For this reason, a notice 1938 * block (spanning several lines) is printed to kernel logs and 1939 * states that the helper should not be used "for production use" 1940 * the first time this helper is used (or more precisely, when 1941 * **trace_printk**\ () buffers are allocated). For passing values 1942 * to user space, perf events should be preferred. 1943 * Return 1944 * The number of bytes written to the buffer, or a negative error 1945 * in case of failure. 1946 * 1947 * u32 bpf_get_prandom_u32(void) 1948 * Description 1949 * Get a pseudo-random number. 1950 * 1951 * From a security point of view, this helper uses its own 1952 * pseudo-random internal state, and cannot be used to infer the 1953 * seed of other random functions in the kernel. However, it is 1954 * essential to note that the generator used by the helper is not 1955 * cryptographically secure. 1956 * Return 1957 * A random 32-bit unsigned value. 1958 * 1959 * u32 bpf_get_smp_processor_id(void) 1960 * Description 1961 * Get the SMP (symmetric multiprocessing) processor id. Note that 1962 * all programs run with migration disabled, which means that the 1963 * SMP processor id is stable during all the execution of the 1964 * program. 1965 * Return 1966 * The SMP id of the processor running the program. 1967 * 1968 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1969 * Description 1970 * Store *len* bytes from address *from* into the packet 1971 * associated to *skb*, at *offset*. *flags* are a combination of 1972 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1973 * checksum for the packet after storing the bytes) and 1974 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1975 * **->swhash** and *skb*\ **->l4hash** to 0). 1976 * 1977 * A call to this helper is susceptible to change the underlying 1978 * packet buffer. Therefore, at load time, all checks on pointers 1979 * previously done by the verifier are invalidated and must be 1980 * performed again, if the helper is used in combination with 1981 * direct packet access. 1982 * Return 1983 * 0 on success, or a negative error in case of failure. 1984 * 1985 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1986 * Description 1987 * Recompute the layer 3 (e.g. IP) checksum for the packet 1988 * associated to *skb*. Computation is incremental, so the helper 1989 * must know the former value of the header field that was 1990 * modified (*from*), the new value of this field (*to*), and the 1991 * number of bytes (2 or 4) for this field, stored in *size*. 1992 * Alternatively, it is possible to store the difference between 1993 * the previous and the new values of the header field in *to*, by 1994 * setting *from* and *size* to 0. For both methods, *offset* 1995 * indicates the location of the IP checksum within the packet. 1996 * 1997 * This helper works in combination with **bpf_csum_diff**\ (), 1998 * which does not update the checksum in-place, but offers more 1999 * flexibility and can handle sizes larger than 2 or 4 for the 2000 * checksum to update. 2001 * 2002 * A call to this helper is susceptible to change the underlying 2003 * packet buffer. Therefore, at load time, all checks on pointers 2004 * previously done by the verifier are invalidated and must be 2005 * performed again, if the helper is used in combination with 2006 * direct packet access. 2007 * Return 2008 * 0 on success, or a negative error in case of failure. 2009 * 2010 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 2011 * Description 2012 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 2013 * packet associated to *skb*. Computation is incremental, so the 2014 * helper must know the former value of the header field that was 2015 * modified (*from*), the new value of this field (*to*), and the 2016 * number of bytes (2 or 4) for this field, stored on the lowest 2017 * four bits of *flags*. Alternatively, it is possible to store 2018 * the difference between the previous and the new values of the 2019 * header field in *to*, by setting *from* and the four lowest 2020 * bits of *flags* to 0. For both methods, *offset* indicates the 2021 * location of the IP checksum within the packet. In addition to 2022 * the size of the field, *flags* can be added (bitwise OR) actual 2023 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 2024 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 2025 * for updates resulting in a null checksum the value is set to 2026 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 2027 * the checksum is to be computed against a pseudo-header. 2028 * 2029 * This helper works in combination with **bpf_csum_diff**\ (), 2030 * which does not update the checksum in-place, but offers more 2031 * flexibility and can handle sizes larger than 2 or 4 for the 2032 * checksum to update. 2033 * 2034 * A call to this helper is susceptible to change the underlying 2035 * packet buffer. Therefore, at load time, all checks on pointers 2036 * previously done by the verifier are invalidated and must be 2037 * performed again, if the helper is used in combination with 2038 * direct packet access. 2039 * Return 2040 * 0 on success, or a negative error in case of failure. 2041 * 2042 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2043 * Description 2044 * This special helper is used to trigger a "tail call", or in 2045 * other words, to jump into another eBPF program. The same stack 2046 * frame is used (but values on stack and in registers for the 2047 * caller are not accessible to the callee). This mechanism allows 2048 * for program chaining, either for raising the maximum number of 2049 * available eBPF instructions, or to execute given programs in 2050 * conditional blocks. For security reasons, there is an upper 2051 * limit to the number of successive tail calls that can be 2052 * performed. 2053 * 2054 * Upon call of this helper, the program attempts to jump into a 2055 * program referenced at index *index* in *prog_array_map*, a 2056 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2057 * *ctx*, a pointer to the context. 2058 * 2059 * If the call succeeds, the kernel immediately runs the first 2060 * instruction of the new program. This is not a function call, 2061 * and it never returns to the previous program. If the call 2062 * fails, then the helper has no effect, and the caller continues 2063 * to run its subsequent instructions. A call can fail if the 2064 * destination program for the jump does not exist (i.e. *index* 2065 * is superior to the number of entries in *prog_array_map*), or 2066 * if the maximum number of tail calls has been reached for this 2067 * chain of programs. This limit is defined in the kernel by the 2068 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2069 * which is currently set to 33. 2070 * Return 2071 * 0 on success, or a negative error in case of failure. 2072 * 2073 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2074 * Description 2075 * Clone and redirect the packet associated to *skb* to another 2076 * net device of index *ifindex*. Both ingress and egress 2077 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2078 * value in *flags* is used to make the distinction (ingress path 2079 * is selected if the flag is present, egress path otherwise). 2080 * This is the only flag supported for now. 2081 * 2082 * In comparison with **bpf_redirect**\ () helper, 2083 * **bpf_clone_redirect**\ () has the associated cost of 2084 * duplicating the packet buffer, but this can be executed out of 2085 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2086 * efficient, but it is handled through an action code where the 2087 * redirection happens only after the eBPF program has returned. 2088 * 2089 * A call to this helper is susceptible to change the underlying 2090 * packet buffer. Therefore, at load time, all checks on pointers 2091 * previously done by the verifier are invalidated and must be 2092 * performed again, if the helper is used in combination with 2093 * direct packet access. 2094 * Return 2095 * 0 on success, or a negative error in case of failure. Positive 2096 * error indicates a potential drop or congestion in the target 2097 * device. The particular positive error codes are not defined. 2098 * 2099 * u64 bpf_get_current_pid_tgid(void) 2100 * Description 2101 * Get the current pid and tgid. 2102 * Return 2103 * A 64-bit integer containing the current tgid and pid, and 2104 * created as such: 2105 * *current_task*\ **->tgid << 32 \|** 2106 * *current_task*\ **->pid**. 2107 * 2108 * u64 bpf_get_current_uid_gid(void) 2109 * Description 2110 * Get the current uid and gid. 2111 * Return 2112 * A 64-bit integer containing the current GID and UID, and 2113 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2114 * 2115 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2116 * Description 2117 * Copy the **comm** attribute of the current task into *buf* of 2118 * *size_of_buf*. The **comm** attribute contains the name of 2119 * the executable (excluding the path) for the current task. The 2120 * *size_of_buf* must be strictly positive. On success, the 2121 * helper makes sure that the *buf* is NUL-terminated. On failure, 2122 * it is filled with zeroes. 2123 * Return 2124 * 0 on success, or a negative error in case of failure. 2125 * 2126 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2127 * Description 2128 * Retrieve the classid for the current task, i.e. for the net_cls 2129 * cgroup to which *skb* belongs. 2130 * 2131 * This helper can be used on TC egress path, but not on ingress. 2132 * 2133 * The net_cls cgroup provides an interface to tag network packets 2134 * based on a user-provided identifier for all traffic coming from 2135 * the tasks belonging to the related cgroup. See also the related 2136 * kernel documentation, available from the Linux sources in file 2137 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2138 * 2139 * The Linux kernel has two versions for cgroups: there are 2140 * cgroups v1 and cgroups v2. Both are available to users, who can 2141 * use a mixture of them, but note that the net_cls cgroup is for 2142 * cgroup v1 only. This makes it incompatible with BPF programs 2143 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2144 * only hold data for one version of cgroups at a time). 2145 * 2146 * This helper is only available is the kernel was compiled with 2147 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2148 * "**y**" or to "**m**". 2149 * Return 2150 * The classid, or 0 for the default unconfigured classid. 2151 * 2152 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2153 * Description 2154 * Push a *vlan_tci* (VLAN tag control information) of protocol 2155 * *vlan_proto* to the packet associated to *skb*, then update 2156 * the checksum. Note that if *vlan_proto* is different from 2157 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2158 * be **ETH_P_8021Q**. 2159 * 2160 * A call to this helper is susceptible to change the underlying 2161 * packet buffer. Therefore, at load time, all checks on pointers 2162 * previously done by the verifier are invalidated and must be 2163 * performed again, if the helper is used in combination with 2164 * direct packet access. 2165 * Return 2166 * 0 on success, or a negative error in case of failure. 2167 * 2168 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2169 * Description 2170 * Pop a VLAN header from the packet associated to *skb*. 2171 * 2172 * A call to this helper is susceptible to change the underlying 2173 * packet buffer. Therefore, at load time, all checks on pointers 2174 * previously done by the verifier are invalidated and must be 2175 * performed again, if the helper is used in combination with 2176 * direct packet access. 2177 * Return 2178 * 0 on success, or a negative error in case of failure. 2179 * 2180 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2181 * Description 2182 * Get tunnel metadata. This helper takes a pointer *key* to an 2183 * empty **struct bpf_tunnel_key** of **size**, that will be 2184 * filled with tunnel metadata for the packet associated to *skb*. 2185 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2186 * indicates that the tunnel is based on IPv6 protocol instead of 2187 * IPv4. 2188 * 2189 * The **struct bpf_tunnel_key** is an object that generalizes the 2190 * principal parameters used by various tunneling protocols into a 2191 * single struct. This way, it can be used to easily make a 2192 * decision based on the contents of the encapsulation header, 2193 * "summarized" in this struct. In particular, it holds the IP 2194 * address of the remote end (IPv4 or IPv6, depending on the case) 2195 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2196 * this struct exposes the *key*\ **->tunnel_id**, which is 2197 * generally mapped to a VNI (Virtual Network Identifier), making 2198 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2199 * () helper. 2200 * 2201 * Let's imagine that the following code is part of a program 2202 * attached to the TC ingress interface, on one end of a GRE 2203 * tunnel, and is supposed to filter out all messages coming from 2204 * remote ends with IPv4 address other than 10.0.0.1: 2205 * 2206 * :: 2207 * 2208 * int ret; 2209 * struct bpf_tunnel_key key = {}; 2210 * 2211 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2212 * if (ret < 0) 2213 * return TC_ACT_SHOT; // drop packet 2214 * 2215 * if (key.remote_ipv4 != 0x0a000001) 2216 * return TC_ACT_SHOT; // drop packet 2217 * 2218 * return TC_ACT_OK; // accept packet 2219 * 2220 * This interface can also be used with all encapsulation devices 2221 * that can operate in "collect metadata" mode: instead of having 2222 * one network device per specific configuration, the "collect 2223 * metadata" mode only requires a single device where the 2224 * configuration can be extracted from this helper. 2225 * 2226 * This can be used together with various tunnels such as VXLan, 2227 * Geneve, GRE or IP in IP (IPIP). 2228 * Return 2229 * 0 on success, or a negative error in case of failure. 2230 * 2231 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2232 * Description 2233 * Populate tunnel metadata for packet associated to *skb.* The 2234 * tunnel metadata is set to the contents of *key*, of *size*. The 2235 * *flags* can be set to a combination of the following values: 2236 * 2237 * **BPF_F_TUNINFO_IPV6** 2238 * Indicate that the tunnel is based on IPv6 protocol 2239 * instead of IPv4. 2240 * **BPF_F_ZERO_CSUM_TX** 2241 * For IPv4 packets, add a flag to tunnel metadata 2242 * indicating that checksum computation should be skipped 2243 * and checksum set to zeroes. 2244 * **BPF_F_DONT_FRAGMENT** 2245 * Add a flag to tunnel metadata indicating that the 2246 * packet should not be fragmented. 2247 * **BPF_F_SEQ_NUMBER** 2248 * Add a flag to tunnel metadata indicating that a 2249 * sequence number should be added to tunnel header before 2250 * sending the packet. This flag was added for GRE 2251 * encapsulation, but might be used with other protocols 2252 * as well in the future. 2253 * **BPF_F_NO_TUNNEL_KEY** 2254 * Add a flag to tunnel metadata indicating that no tunnel 2255 * key should be set in the resulting tunnel header. 2256 * 2257 * Here is a typical usage on the transmit path: 2258 * 2259 * :: 2260 * 2261 * struct bpf_tunnel_key key; 2262 * populate key ... 2263 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2264 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2265 * 2266 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2267 * helper for additional information. 2268 * Return 2269 * 0 on success, or a negative error in case of failure. 2270 * 2271 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2272 * Description 2273 * Read the value of a perf event counter. This helper relies on a 2274 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2275 * the perf event counter is selected when *map* is updated with 2276 * perf event file descriptors. The *map* is an array whose size 2277 * is the number of available CPUs, and each cell contains a value 2278 * relative to one CPU. The value to retrieve is indicated by 2279 * *flags*, that contains the index of the CPU to look up, masked 2280 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2281 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2282 * current CPU should be retrieved. 2283 * 2284 * Note that before Linux 4.13, only hardware perf event can be 2285 * retrieved. 2286 * 2287 * Also, be aware that the newer helper 2288 * **bpf_perf_event_read_value**\ () is recommended over 2289 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2290 * quirks where error and counter value are used as a return code 2291 * (which is wrong to do since ranges may overlap). This issue is 2292 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2293 * time provides more features over the **bpf_perf_event_read**\ 2294 * () interface. Please refer to the description of 2295 * **bpf_perf_event_read_value**\ () for details. 2296 * Return 2297 * The value of the perf event counter read from the map, or a 2298 * negative error code in case of failure. 2299 * 2300 * long bpf_redirect(u32 ifindex, u64 flags) 2301 * Description 2302 * Redirect the packet to another net device of index *ifindex*. 2303 * This helper is somewhat similar to **bpf_clone_redirect**\ 2304 * (), except that the packet is not cloned, which provides 2305 * increased performance. 2306 * 2307 * Except for XDP, both ingress and egress interfaces can be used 2308 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2309 * to make the distinction (ingress path is selected if the flag 2310 * is present, egress path otherwise). Currently, XDP only 2311 * supports redirection to the egress interface, and accepts no 2312 * flag at all. 2313 * 2314 * The same effect can also be attained with the more generic 2315 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2316 * redirect target instead of providing it directly to the helper. 2317 * Return 2318 * For XDP, the helper returns **XDP_REDIRECT** on success or 2319 * **XDP_ABORTED** on error. For other program types, the values 2320 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2321 * error. 2322 * 2323 * u32 bpf_get_route_realm(struct sk_buff *skb) 2324 * Description 2325 * Retrieve the realm or the route, that is to say the 2326 * **tclassid** field of the destination for the *skb*. The 2327 * identifier retrieved is a user-provided tag, similar to the 2328 * one used with the net_cls cgroup (see description for 2329 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2330 * held by a route (a destination entry), not by a task. 2331 * 2332 * Retrieving this identifier works with the clsact TC egress hook 2333 * (see also **tc-bpf(8)**), or alternatively on conventional 2334 * classful egress qdiscs, but not on TC ingress path. In case of 2335 * clsact TC egress hook, this has the advantage that, internally, 2336 * the destination entry has not been dropped yet in the transmit 2337 * path. Therefore, the destination entry does not need to be 2338 * artificially held via **netif_keep_dst**\ () for a classful 2339 * qdisc until the *skb* is freed. 2340 * 2341 * This helper is available only if the kernel was compiled with 2342 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2343 * Return 2344 * The realm of the route for the packet associated to *skb*, or 0 2345 * if none was found. 2346 * 2347 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2348 * Description 2349 * Write raw *data* blob into a special BPF perf event held by 2350 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2351 * event must have the following attributes: **PERF_SAMPLE_RAW** 2352 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2353 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2354 * 2355 * The *flags* are used to indicate the index in *map* for which 2356 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2357 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2358 * to indicate that the index of the current CPU core should be 2359 * used. 2360 * 2361 * The value to write, of *size*, is passed through eBPF stack and 2362 * pointed by *data*. 2363 * 2364 * The context of the program *ctx* needs also be passed to the 2365 * helper. 2366 * 2367 * On user space, a program willing to read the values needs to 2368 * call **perf_event_open**\ () on the perf event (either for 2369 * one or for all CPUs) and to store the file descriptor into the 2370 * *map*. This must be done before the eBPF program can send data 2371 * into it. An example is available in file 2372 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2373 * tree (the eBPF program counterpart is in 2374 * *samples/bpf/trace_output_kern.c*). 2375 * 2376 * **bpf_perf_event_output**\ () achieves better performance 2377 * than **bpf_trace_printk**\ () for sharing data with user 2378 * space, and is much better suitable for streaming data from eBPF 2379 * programs. 2380 * 2381 * Note that this helper is not restricted to tracing use cases 2382 * and can be used with programs attached to TC or XDP as well, 2383 * where it allows for passing data to user space listeners. Data 2384 * can be: 2385 * 2386 * * Only custom structs, 2387 * * Only the packet payload, or 2388 * * A combination of both. 2389 * Return 2390 * 0 on success, or a negative error in case of failure. 2391 * 2392 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2393 * Description 2394 * This helper was provided as an easy way to load data from a 2395 * packet. It can be used to load *len* bytes from *offset* from 2396 * the packet associated to *skb*, into the buffer pointed by 2397 * *to*. 2398 * 2399 * Since Linux 4.7, usage of this helper has mostly been replaced 2400 * by "direct packet access", enabling packet data to be 2401 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2402 * pointing respectively to the first byte of packet data and to 2403 * the byte after the last byte of packet data. However, it 2404 * remains useful if one wishes to read large quantities of data 2405 * at once from a packet into the eBPF stack. 2406 * Return 2407 * 0 on success, or a negative error in case of failure. 2408 * 2409 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2410 * Description 2411 * Walk a user or a kernel stack and return its id. To achieve 2412 * this, the helper needs *ctx*, which is a pointer to the context 2413 * on which the tracing program is executed, and a pointer to a 2414 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2415 * 2416 * The last argument, *flags*, holds the number of stack frames to 2417 * skip (from 0 to 255), masked with 2418 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2419 * a combination of the following flags: 2420 * 2421 * **BPF_F_USER_STACK** 2422 * Collect a user space stack instead of a kernel stack. 2423 * **BPF_F_FAST_STACK_CMP** 2424 * Compare stacks by hash only. 2425 * **BPF_F_REUSE_STACKID** 2426 * If two different stacks hash into the same *stackid*, 2427 * discard the old one. 2428 * 2429 * The stack id retrieved is a 32 bit long integer handle which 2430 * can be further combined with other data (including other stack 2431 * ids) and used as a key into maps. This can be useful for 2432 * generating a variety of graphs (such as flame graphs or off-cpu 2433 * graphs). 2434 * 2435 * For walking a stack, this helper is an improvement over 2436 * **bpf_probe_read**\ (), which can be used with unrolled loops 2437 * but is not efficient and consumes a lot of eBPF instructions. 2438 * Instead, **bpf_get_stackid**\ () can collect up to 2439 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2440 * this limit can be controlled with the **sysctl** program, and 2441 * that it should be manually increased in order to profile long 2442 * user stacks (such as stacks for Java programs). To do so, use: 2443 * 2444 * :: 2445 * 2446 * # sysctl kernel.perf_event_max_stack=<new value> 2447 * Return 2448 * The positive or null stack id on success, or a negative error 2449 * in case of failure. 2450 * 2451 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2452 * Description 2453 * Compute a checksum difference, from the raw buffer pointed by 2454 * *from*, of length *from_size* (that must be a multiple of 4), 2455 * towards the raw buffer pointed by *to*, of size *to_size* 2456 * (same remark). An optional *seed* can be added to the value 2457 * (this can be cascaded, the seed may come from a previous call 2458 * to the helper). 2459 * 2460 * This is flexible enough to be used in several ways: 2461 * 2462 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2463 * checksum, it can be used when pushing new data. 2464 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2465 * checksum, it can be used when removing data from a packet. 2466 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2467 * can be used to compute a diff. Note that *from_size* and 2468 * *to_size* do not need to be equal. 2469 * 2470 * This helper can be used in combination with 2471 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2472 * which one can feed in the difference computed with 2473 * **bpf_csum_diff**\ (). 2474 * Return 2475 * The checksum result, or a negative error code in case of 2476 * failure. 2477 * 2478 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2479 * Description 2480 * Retrieve tunnel options metadata for the packet associated to 2481 * *skb*, and store the raw tunnel option data to the buffer *opt* 2482 * of *size*. 2483 * 2484 * This helper can be used with encapsulation devices that can 2485 * operate in "collect metadata" mode (please refer to the related 2486 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2487 * more details). A particular example where this can be used is 2488 * in combination with the Geneve encapsulation protocol, where it 2489 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2490 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2491 * the eBPF program. This allows for full customization of these 2492 * headers. 2493 * Return 2494 * The size of the option data retrieved. 2495 * 2496 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2497 * Description 2498 * Set tunnel options metadata for the packet associated to *skb* 2499 * to the option data contained in the raw buffer *opt* of *size*. 2500 * 2501 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2502 * helper for additional information. 2503 * Return 2504 * 0 on success, or a negative error in case of failure. 2505 * 2506 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2507 * Description 2508 * Change the protocol of the *skb* to *proto*. Currently 2509 * supported are transition from IPv4 to IPv6, and from IPv6 to 2510 * IPv4. The helper takes care of the groundwork for the 2511 * transition, including resizing the socket buffer. The eBPF 2512 * program is expected to fill the new headers, if any, via 2513 * **skb_store_bytes**\ () and to recompute the checksums with 2514 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2515 * (). The main case for this helper is to perform NAT64 2516 * operations out of an eBPF program. 2517 * 2518 * Internally, the GSO type is marked as dodgy so that headers are 2519 * checked and segments are recalculated by the GSO/GRO engine. 2520 * The size for GSO target is adapted as well. 2521 * 2522 * All values for *flags* are reserved for future usage, and must 2523 * be left at zero. 2524 * 2525 * A call to this helper is susceptible to change the underlying 2526 * packet buffer. Therefore, at load time, all checks on pointers 2527 * previously done by the verifier are invalidated and must be 2528 * performed again, if the helper is used in combination with 2529 * direct packet access. 2530 * Return 2531 * 0 on success, or a negative error in case of failure. 2532 * 2533 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2534 * Description 2535 * Change the packet type for the packet associated to *skb*. This 2536 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2537 * the eBPF program does not have a write access to *skb*\ 2538 * **->pkt_type** beside this helper. Using a helper here allows 2539 * for graceful handling of errors. 2540 * 2541 * The major use case is to change incoming *skb*s to 2542 * **PACKET_HOST** in a programmatic way instead of having to 2543 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2544 * example. 2545 * 2546 * Note that *type* only allows certain values. At this time, they 2547 * are: 2548 * 2549 * **PACKET_HOST** 2550 * Packet is for us. 2551 * **PACKET_BROADCAST** 2552 * Send packet to all. 2553 * **PACKET_MULTICAST** 2554 * Send packet to group. 2555 * **PACKET_OTHERHOST** 2556 * Send packet to someone else. 2557 * Return 2558 * 0 on success, or a negative error in case of failure. 2559 * 2560 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2561 * Description 2562 * Check whether *skb* is a descendant of the cgroup2 held by 2563 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2564 * Return 2565 * The return value depends on the result of the test, and can be: 2566 * 2567 * * 0, if the *skb* failed the cgroup2 descendant test. 2568 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2569 * * A negative error code, if an error occurred. 2570 * 2571 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2572 * Description 2573 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2574 * not set, in particular if the hash was cleared due to mangling, 2575 * recompute this hash. Later accesses to the hash can be done 2576 * directly with *skb*\ **->hash**. 2577 * 2578 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2579 * prototype with **bpf_skb_change_proto**\ (), or calling 2580 * **bpf_skb_store_bytes**\ () with the 2581 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2582 * the hash and to trigger a new computation for the next call to 2583 * **bpf_get_hash_recalc**\ (). 2584 * Return 2585 * The 32-bit hash. 2586 * 2587 * u64 bpf_get_current_task(void) 2588 * Description 2589 * Get the current task. 2590 * Return 2591 * A pointer to the current task struct. 2592 * 2593 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2594 * Description 2595 * Attempt in a safe way to write *len* bytes from the buffer 2596 * *src* to *dst* in memory. It only works for threads that are in 2597 * user context, and *dst* must be a valid user space address. 2598 * 2599 * This helper should not be used to implement any kind of 2600 * security mechanism because of TOC-TOU attacks, but rather to 2601 * debug, divert, and manipulate execution of semi-cooperative 2602 * processes. 2603 * 2604 * Keep in mind that this feature is meant for experiments, and it 2605 * has a risk of crashing the system and running programs. 2606 * Therefore, when an eBPF program using this helper is attached, 2607 * a warning including PID and process name is printed to kernel 2608 * logs. 2609 * Return 2610 * 0 on success, or a negative error in case of failure. 2611 * 2612 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2613 * Description 2614 * Check whether the probe is being run is the context of a given 2615 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2616 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2617 * Return 2618 * The return value depends on the result of the test, and can be: 2619 * 2620 * * 1, if current task belongs to the cgroup2. 2621 * * 0, if current task does not belong to the cgroup2. 2622 * * A negative error code, if an error occurred. 2623 * 2624 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2625 * Description 2626 * Resize (trim or grow) the packet associated to *skb* to the 2627 * new *len*. The *flags* are reserved for future usage, and must 2628 * be left at zero. 2629 * 2630 * The basic idea is that the helper performs the needed work to 2631 * change the size of the packet, then the eBPF program rewrites 2632 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2633 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2634 * and others. This helper is a slow path utility intended for 2635 * replies with control messages. And because it is targeted for 2636 * slow path, the helper itself can afford to be slow: it 2637 * implicitly linearizes, unclones and drops offloads from the 2638 * *skb*. 2639 * 2640 * A call to this helper is susceptible to change the underlying 2641 * packet buffer. Therefore, at load time, all checks on pointers 2642 * previously done by the verifier are invalidated and must be 2643 * performed again, if the helper is used in combination with 2644 * direct packet access. 2645 * Return 2646 * 0 on success, or a negative error in case of failure. 2647 * 2648 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2649 * Description 2650 * Pull in non-linear data in case the *skb* is non-linear and not 2651 * all of *len* are part of the linear section. Make *len* bytes 2652 * from *skb* readable and writable. If a zero value is passed for 2653 * *len*, then all bytes in the linear part of *skb* will be made 2654 * readable and writable. 2655 * 2656 * This helper is only needed for reading and writing with direct 2657 * packet access. 2658 * 2659 * For direct packet access, testing that offsets to access 2660 * are within packet boundaries (test on *skb*\ **->data_end**) is 2661 * susceptible to fail if offsets are invalid, or if the requested 2662 * data is in non-linear parts of the *skb*. On failure the 2663 * program can just bail out, or in the case of a non-linear 2664 * buffer, use a helper to make the data available. The 2665 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2666 * the data. Another one consists in using **bpf_skb_pull_data** 2667 * to pull in once the non-linear parts, then retesting and 2668 * eventually access the data. 2669 * 2670 * At the same time, this also makes sure the *skb* is uncloned, 2671 * which is a necessary condition for direct write. As this needs 2672 * to be an invariant for the write part only, the verifier 2673 * detects writes and adds a prologue that is calling 2674 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2675 * the very beginning in case it is indeed cloned. 2676 * 2677 * A call to this helper is susceptible to change the underlying 2678 * packet buffer. Therefore, at load time, all checks on pointers 2679 * previously done by the verifier are invalidated and must be 2680 * performed again, if the helper is used in combination with 2681 * direct packet access. 2682 * Return 2683 * 0 on success, or a negative error in case of failure. 2684 * 2685 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2686 * Description 2687 * Add the checksum *csum* into *skb*\ **->csum** in case the 2688 * driver has supplied a checksum for the entire packet into that 2689 * field. Return an error otherwise. This helper is intended to be 2690 * used in combination with **bpf_csum_diff**\ (), in particular 2691 * when the checksum needs to be updated after data has been 2692 * written into the packet through direct packet access. 2693 * Return 2694 * The checksum on success, or a negative error code in case of 2695 * failure. 2696 * 2697 * void bpf_set_hash_invalid(struct sk_buff *skb) 2698 * Description 2699 * Invalidate the current *skb*\ **->hash**. It can be used after 2700 * mangling on headers through direct packet access, in order to 2701 * indicate that the hash is outdated and to trigger a 2702 * recalculation the next time the kernel tries to access this 2703 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2704 * Return 2705 * void. 2706 * 2707 * long bpf_get_numa_node_id(void) 2708 * Description 2709 * Return the id of the current NUMA node. The primary use case 2710 * for this helper is the selection of sockets for the local NUMA 2711 * node, when the program is attached to sockets using the 2712 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2713 * but the helper is also available to other eBPF program types, 2714 * similarly to **bpf_get_smp_processor_id**\ (). 2715 * Return 2716 * The id of current NUMA node. 2717 * 2718 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2719 * Description 2720 * Grows headroom of packet associated to *skb* and adjusts the 2721 * offset of the MAC header accordingly, adding *len* bytes of 2722 * space. It automatically extends and reallocates memory as 2723 * required. 2724 * 2725 * This helper can be used on a layer 3 *skb* to push a MAC header 2726 * for redirection into a layer 2 device. 2727 * 2728 * All values for *flags* are reserved for future usage, and must 2729 * be left at zero. 2730 * 2731 * A call to this helper is susceptible to change the underlying 2732 * packet buffer. Therefore, at load time, all checks on pointers 2733 * previously done by the verifier are invalidated and must be 2734 * performed again, if the helper is used in combination with 2735 * direct packet access. 2736 * Return 2737 * 0 on success, or a negative error in case of failure. 2738 * 2739 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2740 * Description 2741 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2742 * it is possible to use a negative value for *delta*. This helper 2743 * can be used to prepare the packet for pushing or popping 2744 * headers. 2745 * 2746 * A call to this helper is susceptible to change the underlying 2747 * packet buffer. Therefore, at load time, all checks on pointers 2748 * previously done by the verifier are invalidated and must be 2749 * performed again, if the helper is used in combination with 2750 * direct packet access. 2751 * Return 2752 * 0 on success, or a negative error in case of failure. 2753 * 2754 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2755 * Description 2756 * Copy a NUL terminated string from an unsafe kernel address 2757 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2758 * more details. 2759 * 2760 * Generally, use **bpf_probe_read_user_str**\ () or 2761 * **bpf_probe_read_kernel_str**\ () instead. 2762 * Return 2763 * On success, the strictly positive length of the string, 2764 * including the trailing NUL character. On error, a negative 2765 * value. 2766 * 2767 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2768 * Description 2769 * If the **struct sk_buff** pointed by *skb* has a known socket, 2770 * retrieve the cookie (generated by the kernel) of this socket. 2771 * If no cookie has been set yet, generate a new cookie. Once 2772 * generated, the socket cookie remains stable for the life of the 2773 * socket. This helper can be useful for monitoring per socket 2774 * networking traffic statistics as it provides a global socket 2775 * identifier that can be assumed unique. 2776 * Return 2777 * A 8-byte long unique number on success, or 0 if the socket 2778 * field is missing inside *skb*. 2779 * 2780 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2781 * Description 2782 * Equivalent to bpf_get_socket_cookie() helper that accepts 2783 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2784 * Return 2785 * A 8-byte long unique number. 2786 * 2787 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2788 * Description 2789 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2790 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2791 * Return 2792 * A 8-byte long unique number. 2793 * 2794 * u64 bpf_get_socket_cookie(struct sock *sk) 2795 * Description 2796 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2797 * *sk*, but gets socket from a BTF **struct sock**. This helper 2798 * also works for sleepable programs. 2799 * Return 2800 * A 8-byte long unique number or 0 if *sk* is NULL. 2801 * 2802 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2803 * Description 2804 * Get the owner UID of the socked associated to *skb*. 2805 * Return 2806 * The owner UID of the socket associated to *skb*. If the socket 2807 * is **NULL**, or if it is not a full socket (i.e. if it is a 2808 * time-wait or a request socket instead), **overflowuid** value 2809 * is returned (note that **overflowuid** might also be the actual 2810 * UID value for the socket). 2811 * 2812 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2813 * Description 2814 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2815 * to value *hash*. 2816 * Return 2817 * 0 2818 * 2819 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2820 * Description 2821 * Emulate a call to **setsockopt()** on the socket associated to 2822 * *bpf_socket*, which must be a full socket. The *level* at 2823 * which the option resides and the name *optname* of the option 2824 * must be specified, see **setsockopt(2)** for more information. 2825 * The option value of length *optlen* is pointed by *optval*. 2826 * 2827 * *bpf_socket* should be one of the following: 2828 * 2829 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2830 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2831 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2832 * 2833 * This helper actually implements a subset of **setsockopt()**. 2834 * It supports the following *level*\ s: 2835 * 2836 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2837 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2838 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2839 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2840 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2841 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2842 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2843 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2844 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2845 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2846 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2847 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2848 * **TCP_BPF_RTO_MIN**. 2849 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2850 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2851 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2852 * Return 2853 * 0 on success, or a negative error in case of failure. 2854 * 2855 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2856 * Description 2857 * Grow or shrink the room for data in the packet associated to 2858 * *skb* by *len_diff*, and according to the selected *mode*. 2859 * 2860 * By default, the helper will reset any offloaded checksum 2861 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2862 * by the following flag: 2863 * 2864 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2865 * checksum data of the skb to CHECKSUM_NONE. 2866 * 2867 * There are two supported modes at this time: 2868 * 2869 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2870 * (room space is added or removed between the layer 2 and 2871 * layer 3 headers). 2872 * 2873 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2874 * (room space is added or removed between the layer 3 and 2875 * layer 4 headers). 2876 * 2877 * The following flags are supported at this time: 2878 * 2879 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2880 * Adjusting mss in this way is not allowed for datagrams. 2881 * 2882 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2883 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2884 * Any new space is reserved to hold a tunnel header. 2885 * Configure skb offsets and other fields accordingly. 2886 * 2887 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2888 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2889 * Use with ENCAP_L3 flags to further specify the tunnel type. 2890 * 2891 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2892 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2893 * type; *len* is the length of the inner MAC header. 2894 * 2895 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2896 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2897 * L2 type as Ethernet. 2898 * 2899 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2900 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2901 * Indicate the new IP header version after decapsulating the outer 2902 * IP header. Used when the inner and outer IP versions are different. 2903 * 2904 * A call to this helper is susceptible to change the underlying 2905 * packet buffer. Therefore, at load time, all checks on pointers 2906 * previously done by the verifier are invalidated and must be 2907 * performed again, if the helper is used in combination with 2908 * direct packet access. 2909 * Return 2910 * 0 on success, or a negative error in case of failure. 2911 * 2912 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2913 * Description 2914 * Redirect the packet to the endpoint referenced by *map* at 2915 * index *key*. Depending on its type, this *map* can contain 2916 * references to net devices (for forwarding packets through other 2917 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2918 * but this is only implemented for native XDP (with driver 2919 * support) as of this writing). 2920 * 2921 * The lower two bits of *flags* are used as the return code if 2922 * the map lookup fails. This is so that the return value can be 2923 * one of the XDP program return codes up to **XDP_TX**, as chosen 2924 * by the caller. The higher bits of *flags* can be set to 2925 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2926 * 2927 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2928 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2929 * interface will be excluded when do broadcasting. 2930 * 2931 * See also **bpf_redirect**\ (), which only supports redirecting 2932 * to an ifindex, but doesn't require a map to do so. 2933 * Return 2934 * **XDP_REDIRECT** on success, or the value of the two lower bits 2935 * of the *flags* argument on error. 2936 * 2937 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2938 * Description 2939 * Redirect the packet to the socket referenced by *map* (of type 2940 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2941 * egress interfaces can be used for redirection. The 2942 * **BPF_F_INGRESS** value in *flags* is used to make the 2943 * distinction (ingress path is selected if the flag is present, 2944 * egress path otherwise). This is the only flag supported for now. 2945 * Return 2946 * **SK_PASS** on success, or **SK_DROP** on error. 2947 * 2948 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2949 * Description 2950 * Add an entry to, or update a *map* referencing sockets. The 2951 * *skops* is used as a new value for the entry associated to 2952 * *key*. *flags* is one of: 2953 * 2954 * **BPF_NOEXIST** 2955 * The entry for *key* must not exist in the map. 2956 * **BPF_EXIST** 2957 * The entry for *key* must already exist in the map. 2958 * **BPF_ANY** 2959 * No condition on the existence of the entry for *key*. 2960 * 2961 * If the *map* has eBPF programs (parser and verdict), those will 2962 * be inherited by the socket being added. If the socket is 2963 * already attached to eBPF programs, this results in an error. 2964 * Return 2965 * 0 on success, or a negative error in case of failure. 2966 * 2967 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2968 * Description 2969 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2970 * *delta* (which can be positive or negative). Note that this 2971 * operation modifies the address stored in *xdp_md*\ **->data**, 2972 * so the latter must be loaded only after the helper has been 2973 * called. 2974 * 2975 * The use of *xdp_md*\ **->data_meta** is optional and programs 2976 * are not required to use it. The rationale is that when the 2977 * packet is processed with XDP (e.g. as DoS filter), it is 2978 * possible to push further meta data along with it before passing 2979 * to the stack, and to give the guarantee that an ingress eBPF 2980 * program attached as a TC classifier on the same device can pick 2981 * this up for further post-processing. Since TC works with socket 2982 * buffers, it remains possible to set from XDP the **mark** or 2983 * **priority** pointers, or other pointers for the socket buffer. 2984 * Having this scratch space generic and programmable allows for 2985 * more flexibility as the user is free to store whatever meta 2986 * data they need. 2987 * 2988 * A call to this helper is susceptible to change the underlying 2989 * packet buffer. Therefore, at load time, all checks on pointers 2990 * previously done by the verifier are invalidated and must be 2991 * performed again, if the helper is used in combination with 2992 * direct packet access. 2993 * Return 2994 * 0 on success, or a negative error in case of failure. 2995 * 2996 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2997 * Description 2998 * Read the value of a perf event counter, and store it into *buf* 2999 * of size *buf_size*. This helper relies on a *map* of type 3000 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 3001 * counter is selected when *map* is updated with perf event file 3002 * descriptors. The *map* is an array whose size is the number of 3003 * available CPUs, and each cell contains a value relative to one 3004 * CPU. The value to retrieve is indicated by *flags*, that 3005 * contains the index of the CPU to look up, masked with 3006 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 3007 * **BPF_F_CURRENT_CPU** to indicate that the value for the 3008 * current CPU should be retrieved. 3009 * 3010 * This helper behaves in a way close to 3011 * **bpf_perf_event_read**\ () helper, save that instead of 3012 * just returning the value observed, it fills the *buf* 3013 * structure. This allows for additional data to be retrieved: in 3014 * particular, the enabled and running times (in *buf*\ 3015 * **->enabled** and *buf*\ **->running**, respectively) are 3016 * copied. In general, **bpf_perf_event_read_value**\ () is 3017 * recommended over **bpf_perf_event_read**\ (), which has some 3018 * ABI issues and provides fewer functionalities. 3019 * 3020 * These values are interesting, because hardware PMU (Performance 3021 * Monitoring Unit) counters are limited resources. When there are 3022 * more PMU based perf events opened than available counters, 3023 * kernel will multiplex these events so each event gets certain 3024 * percentage (but not all) of the PMU time. In case that 3025 * multiplexing happens, the number of samples or counter value 3026 * will not reflect the case compared to when no multiplexing 3027 * occurs. This makes comparison between different runs difficult. 3028 * Typically, the counter value should be normalized before 3029 * comparing to other experiments. The usual normalization is done 3030 * as follows. 3031 * 3032 * :: 3033 * 3034 * normalized_counter = counter * t_enabled / t_running 3035 * 3036 * Where t_enabled is the time enabled for event and t_running is 3037 * the time running for event since last normalization. The 3038 * enabled and running times are accumulated since the perf event 3039 * open. To achieve scaling factor between two invocations of an 3040 * eBPF program, users can use CPU id as the key (which is 3041 * typical for perf array usage model) to remember the previous 3042 * value and do the calculation inside the eBPF program. 3043 * Return 3044 * 0 on success, or a negative error in case of failure. 3045 * 3046 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3047 * Description 3048 * For an eBPF program attached to a perf event, retrieve the 3049 * value of the event counter associated to *ctx* and store it in 3050 * the structure pointed by *buf* and of size *buf_size*. Enabled 3051 * and running times are also stored in the structure (see 3052 * description of helper **bpf_perf_event_read_value**\ () for 3053 * more details). 3054 * Return 3055 * 0 on success, or a negative error in case of failure. 3056 * 3057 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3058 * Description 3059 * Emulate a call to **getsockopt()** on the socket associated to 3060 * *bpf_socket*, which must be a full socket. The *level* at 3061 * which the option resides and the name *optname* of the option 3062 * must be specified, see **getsockopt(2)** for more information. 3063 * The retrieved value is stored in the structure pointed by 3064 * *opval* and of length *optlen*. 3065 * 3066 * *bpf_socket* should be one of the following: 3067 * 3068 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3069 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3070 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3071 * 3072 * This helper actually implements a subset of **getsockopt()**. 3073 * It supports the same set of *optname*\ s that is supported by 3074 * the **bpf_setsockopt**\ () helper. The exceptions are 3075 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3076 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3077 * Return 3078 * 0 on success, or a negative error in case of failure. 3079 * 3080 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3081 * Description 3082 * Used for error injection, this helper uses kprobes to override 3083 * the return value of the probed function, and to set it to *rc*. 3084 * The first argument is the context *regs* on which the kprobe 3085 * works. 3086 * 3087 * This helper works by setting the PC (program counter) 3088 * to an override function which is run in place of the original 3089 * probed function. This means the probed function is not run at 3090 * all. The replacement function just returns with the required 3091 * value. 3092 * 3093 * This helper has security implications, and thus is subject to 3094 * restrictions. It is only available if the kernel was compiled 3095 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3096 * option, and in this case it only works on functions tagged with 3097 * **ALLOW_ERROR_INJECTION** in the kernel code. 3098 * 3099 * Also, the helper is only available for the architectures having 3100 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 3101 * x86 architecture is the only one to support this feature. 3102 * Return 3103 * 0 3104 * 3105 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3106 * Description 3107 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3108 * for the full TCP socket associated to *bpf_sock_ops* to 3109 * *argval*. 3110 * 3111 * The primary use of this field is to determine if there should 3112 * be calls to eBPF programs of type 3113 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3114 * code. A program of the same type can change its value, per 3115 * connection and as necessary, when the connection is 3116 * established. This field is directly accessible for reading, but 3117 * this helper must be used for updates in order to return an 3118 * error if an eBPF program tries to set a callback that is not 3119 * supported in the current kernel. 3120 * 3121 * *argval* is a flag array which can combine these flags: 3122 * 3123 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3124 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3125 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3126 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3127 * 3128 * Therefore, this function can be used to clear a callback flag by 3129 * setting the appropriate bit to zero. e.g. to disable the RTO 3130 * callback: 3131 * 3132 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3133 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3134 * 3135 * Here are some examples of where one could call such eBPF 3136 * program: 3137 * 3138 * * When RTO fires. 3139 * * When a packet is retransmitted. 3140 * * When the connection terminates. 3141 * * When a packet is sent. 3142 * * When a packet is received. 3143 * Return 3144 * Code **-EINVAL** if the socket is not a full TCP socket; 3145 * otherwise, a positive number containing the bits that could not 3146 * be set is returned (which comes down to 0 if all bits were set 3147 * as required). 3148 * 3149 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3150 * Description 3151 * This helper is used in programs implementing policies at the 3152 * socket level. If the message *msg* is allowed to pass (i.e. if 3153 * the verdict eBPF program returns **SK_PASS**), redirect it to 3154 * the socket referenced by *map* (of type 3155 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3156 * egress interfaces can be used for redirection. The 3157 * **BPF_F_INGRESS** value in *flags* is used to make the 3158 * distinction (ingress path is selected if the flag is present, 3159 * egress path otherwise). This is the only flag supported for now. 3160 * Return 3161 * **SK_PASS** on success, or **SK_DROP** on error. 3162 * 3163 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3164 * Description 3165 * For socket policies, apply the verdict of the eBPF program to 3166 * the next *bytes* (number of bytes) of message *msg*. 3167 * 3168 * For example, this helper can be used in the following cases: 3169 * 3170 * * A single **sendmsg**\ () or **sendfile**\ () system call 3171 * contains multiple logical messages that the eBPF program is 3172 * supposed to read and for which it should apply a verdict. 3173 * * An eBPF program only cares to read the first *bytes* of a 3174 * *msg*. If the message has a large payload, then setting up 3175 * and calling the eBPF program repeatedly for all bytes, even 3176 * though the verdict is already known, would create unnecessary 3177 * overhead. 3178 * 3179 * When called from within an eBPF program, the helper sets a 3180 * counter internal to the BPF infrastructure, that is used to 3181 * apply the last verdict to the next *bytes*. If *bytes* is 3182 * smaller than the current data being processed from a 3183 * **sendmsg**\ () or **sendfile**\ () system call, the first 3184 * *bytes* will be sent and the eBPF program will be re-run with 3185 * the pointer for start of data pointing to byte number *bytes* 3186 * **+ 1**. If *bytes* is larger than the current data being 3187 * processed, then the eBPF verdict will be applied to multiple 3188 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3189 * consumed. 3190 * 3191 * Note that if a socket closes with the internal counter holding 3192 * a non-zero value, this is not a problem because data is not 3193 * being buffered for *bytes* and is sent as it is received. 3194 * Return 3195 * 0 3196 * 3197 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3198 * Description 3199 * For socket policies, prevent the execution of the verdict eBPF 3200 * program for message *msg* until *bytes* (byte number) have been 3201 * accumulated. 3202 * 3203 * This can be used when one needs a specific number of bytes 3204 * before a verdict can be assigned, even if the data spans 3205 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3206 * case would be a user calling **sendmsg**\ () repeatedly with 3207 * 1-byte long message segments. Obviously, this is bad for 3208 * performance, but it is still valid. If the eBPF program needs 3209 * *bytes* bytes to validate a header, this helper can be used to 3210 * prevent the eBPF program to be called again until *bytes* have 3211 * been accumulated. 3212 * Return 3213 * 0 3214 * 3215 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3216 * Description 3217 * For socket policies, pull in non-linear data from user space 3218 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3219 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3220 * respectively. 3221 * 3222 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3223 * *msg* it can only parse data that the (**data**, **data_end**) 3224 * pointers have already consumed. For **sendmsg**\ () hooks this 3225 * is likely the first scatterlist element. But for calls relying 3226 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3227 * be the range (**0**, **0**) because the data is shared with 3228 * user space and by default the objective is to avoid allowing 3229 * user space to modify data while (or after) eBPF verdict is 3230 * being decided. This helper can be used to pull in data and to 3231 * set the start and end pointer to given values. Data will be 3232 * copied if necessary (i.e. if data was not linear and if start 3233 * and end pointers do not point to the same chunk). 3234 * 3235 * A call to this helper is susceptible to change the underlying 3236 * packet buffer. Therefore, at load time, all checks on pointers 3237 * previously done by the verifier are invalidated and must be 3238 * performed again, if the helper is used in combination with 3239 * direct packet access. 3240 * 3241 * All values for *flags* are reserved for future usage, and must 3242 * be left at zero. 3243 * Return 3244 * 0 on success, or a negative error in case of failure. 3245 * 3246 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3247 * Description 3248 * Bind the socket associated to *ctx* to the address pointed by 3249 * *addr*, of length *addr_len*. This allows for making outgoing 3250 * connection from the desired IP address, which can be useful for 3251 * example when all processes inside a cgroup should use one 3252 * single IP address on a host that has multiple IP configured. 3253 * 3254 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3255 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3256 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3257 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3258 * behavior and lets the kernel efficiently pick up an unused 3259 * port as long as 4-tuple is unique. Passing non-zero port might 3260 * lead to degraded performance. 3261 * Return 3262 * 0 on success, or a negative error in case of failure. 3263 * 3264 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3265 * Description 3266 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3267 * possible to both shrink and grow the packet tail. 3268 * Shrink done via *delta* being a negative integer. 3269 * 3270 * A call to this helper is susceptible to change the underlying 3271 * packet buffer. Therefore, at load time, all checks on pointers 3272 * previously done by the verifier are invalidated and must be 3273 * performed again, if the helper is used in combination with 3274 * direct packet access. 3275 * Return 3276 * 0 on success, or a negative error in case of failure. 3277 * 3278 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3279 * Description 3280 * Retrieve the XFRM state (IP transform framework, see also 3281 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3282 * 3283 * The retrieved value is stored in the **struct bpf_xfrm_state** 3284 * pointed by *xfrm_state* and of length *size*. 3285 * 3286 * All values for *flags* are reserved for future usage, and must 3287 * be left at zero. 3288 * 3289 * This helper is available only if the kernel was compiled with 3290 * **CONFIG_XFRM** configuration option. 3291 * Return 3292 * 0 on success, or a negative error in case of failure. 3293 * 3294 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3295 * Description 3296 * Return a user or a kernel stack in bpf program provided buffer. 3297 * To achieve this, the helper needs *ctx*, which is a pointer 3298 * to the context on which the tracing program is executed. 3299 * To store the stacktrace, the bpf program provides *buf* with 3300 * a nonnegative *size*. 3301 * 3302 * The last argument, *flags*, holds the number of stack frames to 3303 * skip (from 0 to 255), masked with 3304 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3305 * the following flags: 3306 * 3307 * **BPF_F_USER_STACK** 3308 * Collect a user space stack instead of a kernel stack. 3309 * **BPF_F_USER_BUILD_ID** 3310 * Collect (build_id, file_offset) instead of ips for user 3311 * stack, only valid if **BPF_F_USER_STACK** is also 3312 * specified. 3313 * 3314 * *file_offset* is an offset relative to the beginning 3315 * of the executable or shared object file backing the vma 3316 * which the *ip* falls in. It is *not* an offset relative 3317 * to that object's base address. Accordingly, it must be 3318 * adjusted by adding (sh_addr - sh_offset), where 3319 * sh_{addr,offset} correspond to the executable section 3320 * containing *file_offset* in the object, for comparisons 3321 * to symbols' st_value to be valid. 3322 * 3323 * **bpf_get_stack**\ () can collect up to 3324 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3325 * to sufficient large buffer size. Note that 3326 * this limit can be controlled with the **sysctl** program, and 3327 * that it should be manually increased in order to profile long 3328 * user stacks (such as stacks for Java programs). To do so, use: 3329 * 3330 * :: 3331 * 3332 * # sysctl kernel.perf_event_max_stack=<new value> 3333 * Return 3334 * The non-negative copied *buf* length equal to or less than 3335 * *size* on success, or a negative error in case of failure. 3336 * 3337 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3338 * Description 3339 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3340 * it provides an easy way to load *len* bytes from *offset* 3341 * from the packet associated to *skb*, into the buffer pointed 3342 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3343 * a fifth argument *start_header* exists in order to select a 3344 * base offset to start from. *start_header* can be one of: 3345 * 3346 * **BPF_HDR_START_MAC** 3347 * Base offset to load data from is *skb*'s mac header. 3348 * **BPF_HDR_START_NET** 3349 * Base offset to load data from is *skb*'s network header. 3350 * 3351 * In general, "direct packet access" is the preferred method to 3352 * access packet data, however, this helper is in particular useful 3353 * in socket filters where *skb*\ **->data** does not always point 3354 * to the start of the mac header and where "direct packet access" 3355 * is not available. 3356 * Return 3357 * 0 on success, or a negative error in case of failure. 3358 * 3359 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3360 * Description 3361 * Do FIB lookup in kernel tables using parameters in *params*. 3362 * If lookup is successful and result shows packet is to be 3363 * forwarded, the neighbor tables are searched for the nexthop. 3364 * If successful (ie., FIB lookup shows forwarding and nexthop 3365 * is resolved), the nexthop address is returned in ipv4_dst 3366 * or ipv6_dst based on family, smac is set to mac address of 3367 * egress device, dmac is set to nexthop mac address, rt_metric 3368 * is set to metric from route (IPv4/IPv6 only), and ifindex 3369 * is set to the device index of the nexthop from the FIB lookup. 3370 * 3371 * *plen* argument is the size of the passed in struct. 3372 * *flags* argument can be a combination of one or more of the 3373 * following values: 3374 * 3375 * **BPF_FIB_LOOKUP_DIRECT** 3376 * Do a direct table lookup vs full lookup using FIB 3377 * rules. 3378 * **BPF_FIB_LOOKUP_TBID** 3379 * Used with BPF_FIB_LOOKUP_DIRECT. 3380 * Use the routing table ID present in *params*->tbid 3381 * for the fib lookup. 3382 * **BPF_FIB_LOOKUP_OUTPUT** 3383 * Perform lookup from an egress perspective (default is 3384 * ingress). 3385 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3386 * Skip the neighbour table lookup. *params*->dmac 3387 * and *params*->smac will not be set as output. A common 3388 * use case is to call **bpf_redirect_neigh**\ () after 3389 * doing **bpf_fib_lookup**\ (). 3390 * **BPF_FIB_LOOKUP_SRC** 3391 * Derive and set source IP addr in *params*->ipv{4,6}_src 3392 * for the nexthop. If the src addr cannot be derived, 3393 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3394 * case, *params*->dmac and *params*->smac are not set either. 3395 * 3396 * *ctx* is either **struct xdp_md** for XDP programs or 3397 * **struct sk_buff** tc cls_act programs. 3398 * Return 3399 * * < 0 if any input argument is invalid 3400 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3401 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3402 * packet is not forwarded or needs assist from full stack 3403 * 3404 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3405 * was exceeded and output params->mtu_result contains the MTU. 3406 * 3407 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3408 * Description 3409 * Add an entry to, or update a sockhash *map* referencing sockets. 3410 * The *skops* is used as a new value for the entry associated to 3411 * *key*. *flags* is one of: 3412 * 3413 * **BPF_NOEXIST** 3414 * The entry for *key* must not exist in the map. 3415 * **BPF_EXIST** 3416 * The entry for *key* must already exist in the map. 3417 * **BPF_ANY** 3418 * No condition on the existence of the entry for *key*. 3419 * 3420 * If the *map* has eBPF programs (parser and verdict), those will 3421 * be inherited by the socket being added. If the socket is 3422 * already attached to eBPF programs, this results in an error. 3423 * Return 3424 * 0 on success, or a negative error in case of failure. 3425 * 3426 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3427 * Description 3428 * This helper is used in programs implementing policies at the 3429 * socket level. If the message *msg* is allowed to pass (i.e. if 3430 * the verdict eBPF program returns **SK_PASS**), redirect it to 3431 * the socket referenced by *map* (of type 3432 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3433 * egress interfaces can be used for redirection. The 3434 * **BPF_F_INGRESS** value in *flags* is used to make the 3435 * distinction (ingress path is selected if the flag is present, 3436 * egress path otherwise). This is the only flag supported for now. 3437 * Return 3438 * **SK_PASS** on success, or **SK_DROP** on error. 3439 * 3440 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3441 * Description 3442 * This helper is used in programs implementing policies at the 3443 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3444 * if the verdict eBPF program returns **SK_PASS**), redirect it 3445 * to the socket referenced by *map* (of type 3446 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3447 * egress interfaces can be used for redirection. The 3448 * **BPF_F_INGRESS** value in *flags* is used to make the 3449 * distinction (ingress path is selected if the flag is present, 3450 * egress otherwise). This is the only flag supported for now. 3451 * Return 3452 * **SK_PASS** on success, or **SK_DROP** on error. 3453 * 3454 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3455 * Description 3456 * Encapsulate the packet associated to *skb* within a Layer 3 3457 * protocol header. This header is provided in the buffer at 3458 * address *hdr*, with *len* its size in bytes. *type* indicates 3459 * the protocol of the header and can be one of: 3460 * 3461 * **BPF_LWT_ENCAP_SEG6** 3462 * IPv6 encapsulation with Segment Routing Header 3463 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3464 * the IPv6 header is computed by the kernel. 3465 * **BPF_LWT_ENCAP_SEG6_INLINE** 3466 * Only works if *skb* contains an IPv6 packet. Insert a 3467 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3468 * the IPv6 header. 3469 * **BPF_LWT_ENCAP_IP** 3470 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3471 * must be IPv4 or IPv6, followed by zero or more 3472 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3473 * total bytes in all prepended headers. Please note that 3474 * if **skb_is_gso**\ (*skb*) is true, no more than two 3475 * headers can be prepended, and the inner header, if 3476 * present, should be either GRE or UDP/GUE. 3477 * 3478 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3479 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3480 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3481 * **BPF_PROG_TYPE_LWT_XMIT**. 3482 * 3483 * A call to this helper is susceptible to change the underlying 3484 * packet buffer. Therefore, at load time, all checks on pointers 3485 * previously done by the verifier are invalidated and must be 3486 * performed again, if the helper is used in combination with 3487 * direct packet access. 3488 * Return 3489 * 0 on success, or a negative error in case of failure. 3490 * 3491 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3492 * Description 3493 * Store *len* bytes from address *from* into the packet 3494 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3495 * inside the outermost IPv6 Segment Routing Header can be 3496 * modified through this helper. 3497 * 3498 * A call to this helper is susceptible to change the underlying 3499 * packet buffer. Therefore, at load time, all checks on pointers 3500 * previously done by the verifier are invalidated and must be 3501 * performed again, if the helper is used in combination with 3502 * direct packet access. 3503 * Return 3504 * 0 on success, or a negative error in case of failure. 3505 * 3506 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3507 * Description 3508 * Adjust the size allocated to TLVs in the outermost IPv6 3509 * Segment Routing Header contained in the packet associated to 3510 * *skb*, at position *offset* by *delta* bytes. Only offsets 3511 * after the segments are accepted. *delta* can be as well 3512 * positive (growing) as negative (shrinking). 3513 * 3514 * A call to this helper is susceptible to change the underlying 3515 * packet buffer. Therefore, at load time, all checks on pointers 3516 * previously done by the verifier are invalidated and must be 3517 * performed again, if the helper is used in combination with 3518 * direct packet access. 3519 * Return 3520 * 0 on success, or a negative error in case of failure. 3521 * 3522 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3523 * Description 3524 * Apply an IPv6 Segment Routing action of type *action* to the 3525 * packet associated to *skb*. Each action takes a parameter 3526 * contained at address *param*, and of length *param_len* bytes. 3527 * *action* can be one of: 3528 * 3529 * **SEG6_LOCAL_ACTION_END_X** 3530 * End.X action: Endpoint with Layer-3 cross-connect. 3531 * Type of *param*: **struct in6_addr**. 3532 * **SEG6_LOCAL_ACTION_END_T** 3533 * End.T action: Endpoint with specific IPv6 table lookup. 3534 * Type of *param*: **int**. 3535 * **SEG6_LOCAL_ACTION_END_B6** 3536 * End.B6 action: Endpoint bound to an SRv6 policy. 3537 * Type of *param*: **struct ipv6_sr_hdr**. 3538 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3539 * End.B6.Encap action: Endpoint bound to an SRv6 3540 * encapsulation policy. 3541 * Type of *param*: **struct ipv6_sr_hdr**. 3542 * 3543 * A call to this helper is susceptible to change the underlying 3544 * packet buffer. Therefore, at load time, all checks on pointers 3545 * previously done by the verifier are invalidated and must be 3546 * performed again, if the helper is used in combination with 3547 * direct packet access. 3548 * Return 3549 * 0 on success, or a negative error in case of failure. 3550 * 3551 * long bpf_rc_repeat(void *ctx) 3552 * Description 3553 * This helper is used in programs implementing IR decoding, to 3554 * report a successfully decoded repeat key message. This delays 3555 * the generation of a key up event for previously generated 3556 * key down event. 3557 * 3558 * Some IR protocols like NEC have a special IR message for 3559 * repeating last button, for when a button is held down. 3560 * 3561 * The *ctx* should point to the lirc sample as passed into 3562 * the program. 3563 * 3564 * This helper is only available is the kernel was compiled with 3565 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3566 * "**y**". 3567 * Return 3568 * 0 3569 * 3570 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3571 * Description 3572 * This helper is used in programs implementing IR decoding, to 3573 * report a successfully decoded key press with *scancode*, 3574 * *toggle* value in the given *protocol*. The scancode will be 3575 * translated to a keycode using the rc keymap, and reported as 3576 * an input key down event. After a period a key up event is 3577 * generated. This period can be extended by calling either 3578 * **bpf_rc_keydown**\ () again with the same values, or calling 3579 * **bpf_rc_repeat**\ (). 3580 * 3581 * Some protocols include a toggle bit, in case the button was 3582 * released and pressed again between consecutive scancodes. 3583 * 3584 * The *ctx* should point to the lirc sample as passed into 3585 * the program. 3586 * 3587 * The *protocol* is the decoded protocol number (see 3588 * **enum rc_proto** for some predefined values). 3589 * 3590 * This helper is only available is the kernel was compiled with 3591 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3592 * "**y**". 3593 * Return 3594 * 0 3595 * 3596 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3597 * Description 3598 * Return the cgroup v2 id of the socket associated with the *skb*. 3599 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3600 * helper for cgroup v1 by providing a tag resp. identifier that 3601 * can be matched on or used for map lookups e.g. to implement 3602 * policy. The cgroup v2 id of a given path in the hierarchy is 3603 * exposed in user space through the f_handle API in order to get 3604 * to the same 64-bit id. 3605 * 3606 * This helper can be used on TC egress path, but not on ingress, 3607 * and is available only if the kernel was compiled with the 3608 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3609 * Return 3610 * The id is returned or 0 in case the id could not be retrieved. 3611 * 3612 * u64 bpf_get_current_cgroup_id(void) 3613 * Description 3614 * Get the current cgroup id based on the cgroup within which 3615 * the current task is running. 3616 * Return 3617 * A 64-bit integer containing the current cgroup id based 3618 * on the cgroup within which the current task is running. 3619 * 3620 * void *bpf_get_local_storage(void *map, u64 flags) 3621 * Description 3622 * Get the pointer to the local storage area. 3623 * The type and the size of the local storage is defined 3624 * by the *map* argument. 3625 * The *flags* meaning is specific for each map type, 3626 * and has to be 0 for cgroup local storage. 3627 * 3628 * Depending on the BPF program type, a local storage area 3629 * can be shared between multiple instances of the BPF program, 3630 * running simultaneously. 3631 * 3632 * A user should care about the synchronization by himself. 3633 * For example, by using the **BPF_ATOMIC** instructions to alter 3634 * the shared data. 3635 * Return 3636 * A pointer to the local storage area. 3637 * 3638 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3639 * Description 3640 * Select a **SO_REUSEPORT** socket from a 3641 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3642 * It checks the selected socket is matching the incoming 3643 * request in the socket buffer. 3644 * Return 3645 * 0 on success, or a negative error in case of failure. 3646 * 3647 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3648 * Description 3649 * Return id of cgroup v2 that is ancestor of cgroup associated 3650 * with the *skb* at the *ancestor_level*. The root cgroup is at 3651 * *ancestor_level* zero and each step down the hierarchy 3652 * increments the level. If *ancestor_level* == level of cgroup 3653 * associated with *skb*, then return value will be same as that 3654 * of **bpf_skb_cgroup_id**\ (). 3655 * 3656 * The helper is useful to implement policies based on cgroups 3657 * that are upper in hierarchy than immediate cgroup associated 3658 * with *skb*. 3659 * 3660 * The format of returned id and helper limitations are same as in 3661 * **bpf_skb_cgroup_id**\ (). 3662 * Return 3663 * The id is returned or 0 in case the id could not be retrieved. 3664 * 3665 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3666 * Description 3667 * Look for TCP socket matching *tuple*, optionally in a child 3668 * network namespace *netns*. The return value must be checked, 3669 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3670 * 3671 * The *ctx* should point to the context of the program, such as 3672 * the skb or socket (depending on the hook in use). This is used 3673 * to determine the base network namespace for the lookup. 3674 * 3675 * *tuple_size* must be one of: 3676 * 3677 * **sizeof**\ (*tuple*\ **->ipv4**) 3678 * Look for an IPv4 socket. 3679 * **sizeof**\ (*tuple*\ **->ipv6**) 3680 * Look for an IPv6 socket. 3681 * 3682 * If the *netns* is a negative signed 32-bit integer, then the 3683 * socket lookup table in the netns associated with the *ctx* 3684 * will be used. For the TC hooks, this is the netns of the device 3685 * in the skb. For socket hooks, this is the netns of the socket. 3686 * If *netns* is any other signed 32-bit value greater than or 3687 * equal to zero then it specifies the ID of the netns relative to 3688 * the netns associated with the *ctx*. *netns* values beyond the 3689 * range of 32-bit integers are reserved for future use. 3690 * 3691 * All values for *flags* are reserved for future usage, and must 3692 * be left at zero. 3693 * 3694 * This helper is available only if the kernel was compiled with 3695 * **CONFIG_NET** configuration option. 3696 * Return 3697 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3698 * For sockets with reuseport option, the **struct bpf_sock** 3699 * result is from *reuse*\ **->socks**\ [] using the hash of the 3700 * tuple. 3701 * 3702 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3703 * Description 3704 * Look for UDP socket matching *tuple*, optionally in a child 3705 * network namespace *netns*. The return value must be checked, 3706 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3707 * 3708 * The *ctx* should point to the context of the program, such as 3709 * the skb or socket (depending on the hook in use). This is used 3710 * to determine the base network namespace for the lookup. 3711 * 3712 * *tuple_size* must be one of: 3713 * 3714 * **sizeof**\ (*tuple*\ **->ipv4**) 3715 * Look for an IPv4 socket. 3716 * **sizeof**\ (*tuple*\ **->ipv6**) 3717 * Look for an IPv6 socket. 3718 * 3719 * If the *netns* is a negative signed 32-bit integer, then the 3720 * socket lookup table in the netns associated with the *ctx* 3721 * will be used. For the TC hooks, this is the netns of the device 3722 * in the skb. For socket hooks, this is the netns of the socket. 3723 * If *netns* is any other signed 32-bit value greater than or 3724 * equal to zero then it specifies the ID of the netns relative to 3725 * the netns associated with the *ctx*. *netns* values beyond the 3726 * range of 32-bit integers are reserved for future use. 3727 * 3728 * All values for *flags* are reserved for future usage, and must 3729 * be left at zero. 3730 * 3731 * This helper is available only if the kernel was compiled with 3732 * **CONFIG_NET** configuration option. 3733 * Return 3734 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3735 * For sockets with reuseport option, the **struct bpf_sock** 3736 * result is from *reuse*\ **->socks**\ [] using the hash of the 3737 * tuple. 3738 * 3739 * long bpf_sk_release(void *sock) 3740 * Description 3741 * Release the reference held by *sock*. *sock* must be a 3742 * non-**NULL** pointer that was returned from 3743 * **bpf_sk_lookup_xxx**\ (). 3744 * Return 3745 * 0 on success, or a negative error in case of failure. 3746 * 3747 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3748 * Description 3749 * Push an element *value* in *map*. *flags* is one of: 3750 * 3751 * **BPF_EXIST** 3752 * If the queue/stack is full, the oldest element is 3753 * removed to make room for this. 3754 * Return 3755 * 0 on success, or a negative error in case of failure. 3756 * 3757 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3758 * Description 3759 * Pop an element from *map*. 3760 * Return 3761 * 0 on success, or a negative error in case of failure. 3762 * 3763 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3764 * Description 3765 * Get an element from *map* without removing it. 3766 * Return 3767 * 0 on success, or a negative error in case of failure. 3768 * 3769 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3770 * Description 3771 * For socket policies, insert *len* bytes into *msg* at offset 3772 * *start*. 3773 * 3774 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3775 * *msg* it may want to insert metadata or options into the *msg*. 3776 * This can later be read and used by any of the lower layer BPF 3777 * hooks. 3778 * 3779 * This helper may fail if under memory pressure (a malloc 3780 * fails) in these cases BPF programs will get an appropriate 3781 * error and BPF programs will need to handle them. 3782 * Return 3783 * 0 on success, or a negative error in case of failure. 3784 * 3785 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3786 * Description 3787 * Will remove *len* bytes from a *msg* starting at byte *start*. 3788 * This may result in **ENOMEM** errors under certain situations if 3789 * an allocation and copy are required due to a full ring buffer. 3790 * However, the helper will try to avoid doing the allocation 3791 * if possible. Other errors can occur if input parameters are 3792 * invalid either due to *start* byte not being valid part of *msg* 3793 * payload and/or *pop* value being to large. 3794 * Return 3795 * 0 on success, or a negative error in case of failure. 3796 * 3797 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3798 * Description 3799 * This helper is used in programs implementing IR decoding, to 3800 * report a successfully decoded pointer movement. 3801 * 3802 * The *ctx* should point to the lirc sample as passed into 3803 * the program. 3804 * 3805 * This helper is only available is the kernel was compiled with 3806 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3807 * "**y**". 3808 * Return 3809 * 0 3810 * 3811 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3812 * Description 3813 * Acquire a spinlock represented by the pointer *lock*, which is 3814 * stored as part of a value of a map. Taking the lock allows to 3815 * safely update the rest of the fields in that value. The 3816 * spinlock can (and must) later be released with a call to 3817 * **bpf_spin_unlock**\ (\ *lock*\ ). 3818 * 3819 * Spinlocks in BPF programs come with a number of restrictions 3820 * and constraints: 3821 * 3822 * * **bpf_spin_lock** objects are only allowed inside maps of 3823 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3824 * list could be extended in the future). 3825 * * BTF description of the map is mandatory. 3826 * * The BPF program can take ONE lock at a time, since taking two 3827 * or more could cause dead locks. 3828 * * Only one **struct bpf_spin_lock** is allowed per map element. 3829 * * When the lock is taken, calls (either BPF to BPF or helpers) 3830 * are not allowed. 3831 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3832 * allowed inside a spinlock-ed region. 3833 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3834 * the lock, on all execution paths, before it returns. 3835 * * The BPF program can access **struct bpf_spin_lock** only via 3836 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3837 * helpers. Loading or storing data into the **struct 3838 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3839 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3840 * of the map value must be a struct and have **struct 3841 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3842 * Nested lock inside another struct is not allowed. 3843 * * The **struct bpf_spin_lock** *lock* field in a map value must 3844 * be aligned on a multiple of 4 bytes in that value. 3845 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3846 * the **bpf_spin_lock** field to user space. 3847 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3848 * a BPF program, do not update the **bpf_spin_lock** field. 3849 * * **bpf_spin_lock** cannot be on the stack or inside a 3850 * networking packet (it can only be inside of a map values). 3851 * * **bpf_spin_lock** is available to root only. 3852 * * Tracing programs and socket filter programs cannot use 3853 * **bpf_spin_lock**\ () due to insufficient preemption checks 3854 * (but this may change in the future). 3855 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3856 * Return 3857 * 0 3858 * 3859 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3860 * Description 3861 * Release the *lock* previously locked by a call to 3862 * **bpf_spin_lock**\ (\ *lock*\ ). 3863 * Return 3864 * 0 3865 * 3866 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3867 * Description 3868 * This helper gets a **struct bpf_sock** pointer such 3869 * that all the fields in this **bpf_sock** can be accessed. 3870 * Return 3871 * A **struct bpf_sock** pointer on success, or **NULL** in 3872 * case of failure. 3873 * 3874 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3875 * Description 3876 * This helper gets a **struct bpf_tcp_sock** pointer from a 3877 * **struct bpf_sock** pointer. 3878 * Return 3879 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3880 * case of failure. 3881 * 3882 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3883 * Description 3884 * Set ECN (Explicit Congestion Notification) field of IP header 3885 * to **CE** (Congestion Encountered) if current value is **ECT** 3886 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3887 * and IPv4. 3888 * Return 3889 * 1 if the **CE** flag is set (either by the current helper call 3890 * or because it was already present), 0 if it is not set. 3891 * 3892 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3893 * Description 3894 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3895 * **bpf_sk_release**\ () is unnecessary and not allowed. 3896 * Return 3897 * A **struct bpf_sock** pointer on success, or **NULL** in 3898 * case of failure. 3899 * 3900 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3901 * Description 3902 * Look for TCP socket matching *tuple*, optionally in a child 3903 * network namespace *netns*. The return value must be checked, 3904 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3905 * 3906 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3907 * that it also returns timewait or request sockets. Use 3908 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3909 * full structure. 3910 * 3911 * This helper is available only if the kernel was compiled with 3912 * **CONFIG_NET** configuration option. 3913 * Return 3914 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3915 * For sockets with reuseport option, the **struct bpf_sock** 3916 * result is from *reuse*\ **->socks**\ [] using the hash of the 3917 * tuple. 3918 * 3919 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3920 * Description 3921 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3922 * the listening socket in *sk*. 3923 * 3924 * *iph* points to the start of the IPv4 or IPv6 header, while 3925 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3926 * **sizeof**\ (**struct ipv6hdr**). 3927 * 3928 * *th* points to the start of the TCP header, while *th_len* 3929 * contains the length of the TCP header (at least 3930 * **sizeof**\ (**struct tcphdr**)). 3931 * Return 3932 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3933 * error otherwise. 3934 * 3935 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3936 * Description 3937 * Get name of sysctl in /proc/sys/ and copy it into provided by 3938 * program buffer *buf* of size *buf_len*. 3939 * 3940 * The buffer is always NUL terminated, unless it's zero-sized. 3941 * 3942 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3943 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3944 * only (e.g. "tcp_mem"). 3945 * Return 3946 * Number of character copied (not including the trailing NUL). 3947 * 3948 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3949 * truncated name in this case). 3950 * 3951 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3952 * Description 3953 * Get current value of sysctl as it is presented in /proc/sys 3954 * (incl. newline, etc), and copy it as a string into provided 3955 * by program buffer *buf* of size *buf_len*. 3956 * 3957 * The whole value is copied, no matter what file position user 3958 * space issued e.g. sys_read at. 3959 * 3960 * The buffer is always NUL terminated, unless it's zero-sized. 3961 * Return 3962 * Number of character copied (not including the trailing NUL). 3963 * 3964 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3965 * truncated name in this case). 3966 * 3967 * **-EINVAL** if current value was unavailable, e.g. because 3968 * sysctl is uninitialized and read returns -EIO for it. 3969 * 3970 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3971 * Description 3972 * Get new value being written by user space to sysctl (before 3973 * the actual write happens) and copy it as a string into 3974 * provided by program buffer *buf* of size *buf_len*. 3975 * 3976 * User space may write new value at file position > 0. 3977 * 3978 * The buffer is always NUL terminated, unless it's zero-sized. 3979 * Return 3980 * Number of character copied (not including the trailing NUL). 3981 * 3982 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3983 * truncated name in this case). 3984 * 3985 * **-EINVAL** if sysctl is being read. 3986 * 3987 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3988 * Description 3989 * Override new value being written by user space to sysctl with 3990 * value provided by program in buffer *buf* of size *buf_len*. 3991 * 3992 * *buf* should contain a string in same form as provided by user 3993 * space on sysctl write. 3994 * 3995 * User space may write new value at file position > 0. To override 3996 * the whole sysctl value file position should be set to zero. 3997 * Return 3998 * 0 on success. 3999 * 4000 * **-E2BIG** if the *buf_len* is too big. 4001 * 4002 * **-EINVAL** if sysctl is being read. 4003 * 4004 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 4005 * Description 4006 * Convert the initial part of the string from buffer *buf* of 4007 * size *buf_len* to a long integer according to the given base 4008 * and save the result in *res*. 4009 * 4010 * The string may begin with an arbitrary amount of white space 4011 * (as determined by **isspace**\ (3)) followed by a single 4012 * optional '**-**' sign. 4013 * 4014 * Five least significant bits of *flags* encode base, other bits 4015 * are currently unused. 4016 * 4017 * Base must be either 8, 10, 16 or 0 to detect it automatically 4018 * similar to user space **strtol**\ (3). 4019 * Return 4020 * Number of characters consumed on success. Must be positive but 4021 * no more than *buf_len*. 4022 * 4023 * **-EINVAL** if no valid digits were found or unsupported base 4024 * was provided. 4025 * 4026 * **-ERANGE** if resulting value was out of range. 4027 * 4028 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 4029 * Description 4030 * Convert the initial part of the string from buffer *buf* of 4031 * size *buf_len* to an unsigned long integer according to the 4032 * given base and save the result in *res*. 4033 * 4034 * The string may begin with an arbitrary amount of white space 4035 * (as determined by **isspace**\ (3)). 4036 * 4037 * Five least significant bits of *flags* encode base, other bits 4038 * are currently unused. 4039 * 4040 * Base must be either 8, 10, 16 or 0 to detect it automatically 4041 * similar to user space **strtoul**\ (3). 4042 * Return 4043 * Number of characters consumed on success. Must be positive but 4044 * no more than *buf_len*. 4045 * 4046 * **-EINVAL** if no valid digits were found or unsupported base 4047 * was provided. 4048 * 4049 * **-ERANGE** if resulting value was out of range. 4050 * 4051 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4052 * Description 4053 * Get a bpf-local-storage from a *sk*. 4054 * 4055 * Logically, it could be thought of getting the value from 4056 * a *map* with *sk* as the **key**. From this 4057 * perspective, the usage is not much different from 4058 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4059 * helper enforces the key must be a full socket and the map must 4060 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4061 * 4062 * Underneath, the value is stored locally at *sk* instead of 4063 * the *map*. The *map* is used as the bpf-local-storage 4064 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4065 * searched against all bpf-local-storages residing at *sk*. 4066 * 4067 * *sk* is a kernel **struct sock** pointer for LSM program. 4068 * *sk* is a **struct bpf_sock** pointer for other program types. 4069 * 4070 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4071 * used such that a new bpf-local-storage will be 4072 * created if one does not exist. *value* can be used 4073 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4074 * the initial value of a bpf-local-storage. If *value* is 4075 * **NULL**, the new bpf-local-storage will be zero initialized. 4076 * Return 4077 * A bpf-local-storage pointer is returned on success. 4078 * 4079 * **NULL** if not found or there was an error in adding 4080 * a new bpf-local-storage. 4081 * 4082 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4083 * Description 4084 * Delete a bpf-local-storage from a *sk*. 4085 * Return 4086 * 0 on success. 4087 * 4088 * **-ENOENT** if the bpf-local-storage cannot be found. 4089 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4090 * 4091 * long bpf_send_signal(u32 sig) 4092 * Description 4093 * Send signal *sig* to the process of the current task. 4094 * The signal may be delivered to any of this process's threads. 4095 * Return 4096 * 0 on success or successfully queued. 4097 * 4098 * **-EBUSY** if work queue under nmi is full. 4099 * 4100 * **-EINVAL** if *sig* is invalid. 4101 * 4102 * **-EPERM** if no permission to send the *sig*. 4103 * 4104 * **-EAGAIN** if bpf program can try again. 4105 * 4106 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4107 * Description 4108 * Try to issue a SYN cookie for the packet with corresponding 4109 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4110 * 4111 * *iph* points to the start of the IPv4 or IPv6 header, while 4112 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4113 * **sizeof**\ (**struct ipv6hdr**). 4114 * 4115 * *th* points to the start of the TCP header, while *th_len* 4116 * contains the length of the TCP header with options (at least 4117 * **sizeof**\ (**struct tcphdr**)). 4118 * Return 4119 * On success, lower 32 bits hold the generated SYN cookie in 4120 * followed by 16 bits which hold the MSS value for that cookie, 4121 * and the top 16 bits are unused. 4122 * 4123 * On failure, the returned value is one of the following: 4124 * 4125 * **-EINVAL** SYN cookie cannot be issued due to error 4126 * 4127 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4128 * 4129 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4130 * 4131 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4132 * 4133 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4134 * Description 4135 * Write raw *data* blob into a special BPF perf event held by 4136 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4137 * event must have the following attributes: **PERF_SAMPLE_RAW** 4138 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4139 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4140 * 4141 * The *flags* are used to indicate the index in *map* for which 4142 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4143 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4144 * to indicate that the index of the current CPU core should be 4145 * used. 4146 * 4147 * The value to write, of *size*, is passed through eBPF stack and 4148 * pointed by *data*. 4149 * 4150 * *ctx* is a pointer to in-kernel struct sk_buff. 4151 * 4152 * This helper is similar to **bpf_perf_event_output**\ () but 4153 * restricted to raw_tracepoint bpf programs. 4154 * Return 4155 * 0 on success, or a negative error in case of failure. 4156 * 4157 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4158 * Description 4159 * Safely attempt to read *size* bytes from user space address 4160 * *unsafe_ptr* and store the data in *dst*. 4161 * Return 4162 * 0 on success, or a negative error in case of failure. 4163 * 4164 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4165 * Description 4166 * Safely attempt to read *size* bytes from kernel space address 4167 * *unsafe_ptr* and store the data in *dst*. 4168 * Return 4169 * 0 on success, or a negative error in case of failure. 4170 * 4171 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4172 * Description 4173 * Copy a NUL terminated string from an unsafe user address 4174 * *unsafe_ptr* to *dst*. The *size* should include the 4175 * terminating NUL byte. In case the string length is smaller than 4176 * *size*, the target is not padded with further NUL bytes. If the 4177 * string length is larger than *size*, just *size*-1 bytes are 4178 * copied and the last byte is set to NUL. 4179 * 4180 * On success, returns the number of bytes that were written, 4181 * including the terminal NUL. This makes this helper useful in 4182 * tracing programs for reading strings, and more importantly to 4183 * get its length at runtime. See the following snippet: 4184 * 4185 * :: 4186 * 4187 * SEC("kprobe/sys_open") 4188 * void bpf_sys_open(struct pt_regs *ctx) 4189 * { 4190 * char buf[PATHLEN]; // PATHLEN is defined to 256 4191 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4192 * ctx->di); 4193 * 4194 * // Consume buf, for example push it to 4195 * // userspace via bpf_perf_event_output(); we 4196 * // can use res (the string length) as event 4197 * // size, after checking its boundaries. 4198 * } 4199 * 4200 * In comparison, using **bpf_probe_read_user**\ () helper here 4201 * instead to read the string would require to estimate the length 4202 * at compile time, and would often result in copying more memory 4203 * than necessary. 4204 * 4205 * Another useful use case is when parsing individual process 4206 * arguments or individual environment variables navigating 4207 * *current*\ **->mm->arg_start** and *current*\ 4208 * **->mm->env_start**: using this helper and the return value, 4209 * one can quickly iterate at the right offset of the memory area. 4210 * Return 4211 * On success, the strictly positive length of the output string, 4212 * including the trailing NUL character. On error, a negative 4213 * value. 4214 * 4215 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4216 * Description 4217 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4218 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4219 * Return 4220 * On success, the strictly positive length of the string, including 4221 * the trailing NUL character. On error, a negative value. 4222 * 4223 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4224 * Description 4225 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4226 * *rcv_nxt* is the ack_seq to be sent out. 4227 * Return 4228 * 0 on success, or a negative error in case of failure. 4229 * 4230 * long bpf_send_signal_thread(u32 sig) 4231 * Description 4232 * Send signal *sig* to the thread corresponding to the current task. 4233 * Return 4234 * 0 on success or successfully queued. 4235 * 4236 * **-EBUSY** if work queue under nmi is full. 4237 * 4238 * **-EINVAL** if *sig* is invalid. 4239 * 4240 * **-EPERM** if no permission to send the *sig*. 4241 * 4242 * **-EAGAIN** if bpf program can try again. 4243 * 4244 * u64 bpf_jiffies64(void) 4245 * Description 4246 * Obtain the 64bit jiffies 4247 * Return 4248 * The 64 bit jiffies 4249 * 4250 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4251 * Description 4252 * For an eBPF program attached to a perf event, retrieve the 4253 * branch records (**struct perf_branch_entry**) associated to *ctx* 4254 * and store it in the buffer pointed by *buf* up to size 4255 * *size* bytes. 4256 * Return 4257 * On success, number of bytes written to *buf*. On error, a 4258 * negative value. 4259 * 4260 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4261 * instead return the number of bytes required to store all the 4262 * branch entries. If this flag is set, *buf* may be NULL. 4263 * 4264 * **-EINVAL** if arguments invalid or **size** not a multiple 4265 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4266 * 4267 * **-ENOENT** if architecture does not support branch records. 4268 * 4269 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4270 * Description 4271 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4272 * *namespace* will be returned in *nsdata*. 4273 * Return 4274 * 0 on success, or one of the following in case of failure: 4275 * 4276 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4277 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4278 * 4279 * **-ENOENT** if pidns does not exists for the current task. 4280 * 4281 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4282 * Description 4283 * Write raw *data* blob into a special BPF perf event held by 4284 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4285 * event must have the following attributes: **PERF_SAMPLE_RAW** 4286 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4287 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4288 * 4289 * The *flags* are used to indicate the index in *map* for which 4290 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4291 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4292 * to indicate that the index of the current CPU core should be 4293 * used. 4294 * 4295 * The value to write, of *size*, is passed through eBPF stack and 4296 * pointed by *data*. 4297 * 4298 * *ctx* is a pointer to in-kernel struct xdp_buff. 4299 * 4300 * This helper is similar to **bpf_perf_eventoutput**\ () but 4301 * restricted to raw_tracepoint bpf programs. 4302 * Return 4303 * 0 on success, or a negative error in case of failure. 4304 * 4305 * u64 bpf_get_netns_cookie(void *ctx) 4306 * Description 4307 * Retrieve the cookie (generated by the kernel) of the network 4308 * namespace the input *ctx* is associated with. The network 4309 * namespace cookie remains stable for its lifetime and provides 4310 * a global identifier that can be assumed unique. If *ctx* is 4311 * NULL, then the helper returns the cookie for the initial 4312 * network namespace. The cookie itself is very similar to that 4313 * of **bpf_get_socket_cookie**\ () helper, but for network 4314 * namespaces instead of sockets. 4315 * Return 4316 * A 8-byte long opaque number. 4317 * 4318 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4319 * Description 4320 * Return id of cgroup v2 that is ancestor of the cgroup associated 4321 * with the current task at the *ancestor_level*. The root cgroup 4322 * is at *ancestor_level* zero and each step down the hierarchy 4323 * increments the level. If *ancestor_level* == level of cgroup 4324 * associated with the current task, then return value will be the 4325 * same as that of **bpf_get_current_cgroup_id**\ (). 4326 * 4327 * The helper is useful to implement policies based on cgroups 4328 * that are upper in hierarchy than immediate cgroup associated 4329 * with the current task. 4330 * 4331 * The format of returned id and helper limitations are same as in 4332 * **bpf_get_current_cgroup_id**\ (). 4333 * Return 4334 * The id is returned or 0 in case the id could not be retrieved. 4335 * 4336 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4337 * Description 4338 * Helper is overloaded depending on BPF program type. This 4339 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4340 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4341 * 4342 * Assign the *sk* to the *skb*. When combined with appropriate 4343 * routing configuration to receive the packet towards the socket, 4344 * will cause *skb* to be delivered to the specified socket. 4345 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4346 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4347 * interfere with successful delivery to the socket. 4348 * 4349 * This operation is only valid from TC ingress path. 4350 * 4351 * The *flags* argument must be zero. 4352 * Return 4353 * 0 on success, or a negative error in case of failure: 4354 * 4355 * **-EINVAL** if specified *flags* are not supported. 4356 * 4357 * **-ENOENT** if the socket is unavailable for assignment. 4358 * 4359 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4360 * 4361 * **-EOPNOTSUPP** if the operation is not supported, for example 4362 * a call from outside of TC ingress. 4363 * 4364 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4365 * Description 4366 * Helper is overloaded depending on BPF program type. This 4367 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4368 * 4369 * Select the *sk* as a result of a socket lookup. 4370 * 4371 * For the operation to succeed passed socket must be compatible 4372 * with the packet description provided by the *ctx* object. 4373 * 4374 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4375 * be an exact match. While IP family (**AF_INET** or 4376 * **AF_INET6**) must be compatible, that is IPv6 sockets 4377 * that are not v6-only can be selected for IPv4 packets. 4378 * 4379 * Only TCP listeners and UDP unconnected sockets can be 4380 * selected. *sk* can also be NULL to reset any previous 4381 * selection. 4382 * 4383 * *flags* argument can combination of following values: 4384 * 4385 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4386 * socket selection, potentially done by a BPF program 4387 * that ran before us. 4388 * 4389 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4390 * load-balancing within reuseport group for the socket 4391 * being selected. 4392 * 4393 * On success *ctx->sk* will point to the selected socket. 4394 * 4395 * Return 4396 * 0 on success, or a negative errno in case of failure. 4397 * 4398 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4399 * not compatible with packet family (*ctx->family*). 4400 * 4401 * * **-EEXIST** if socket has been already selected, 4402 * potentially by another program, and 4403 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4404 * 4405 * * **-EINVAL** if unsupported flags were specified. 4406 * 4407 * * **-EPROTOTYPE** if socket L4 protocol 4408 * (*sk->protocol*) doesn't match packet protocol 4409 * (*ctx->protocol*). 4410 * 4411 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4412 * state (TCP listening or UDP unconnected). 4413 * 4414 * u64 bpf_ktime_get_boot_ns(void) 4415 * Description 4416 * Return the time elapsed since system boot, in nanoseconds. 4417 * Does include the time the system was suspended. 4418 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4419 * Return 4420 * Current *ktime*. 4421 * 4422 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4423 * Description 4424 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4425 * out the format string. 4426 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4427 * the format string itself. The *data* and *data_len* are format string 4428 * arguments. The *data* are a **u64** array and corresponding format string 4429 * values are stored in the array. For strings and pointers where pointees 4430 * are accessed, only the pointer values are stored in the *data* array. 4431 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4432 * 4433 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4434 * Reading kernel memory may fail due to either invalid address or 4435 * valid address but requiring a major memory fault. If reading kernel memory 4436 * fails, the string for **%s** will be an empty string, and the ip 4437 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4438 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4439 * Return 4440 * 0 on success, or a negative error in case of failure: 4441 * 4442 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4443 * by returning 1 from bpf program. 4444 * 4445 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4446 * 4447 * **-E2BIG** if *fmt* contains too many format specifiers. 4448 * 4449 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4450 * 4451 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4452 * Description 4453 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4454 * The *m* represents the seq_file. The *data* and *len* represent the 4455 * data to write in bytes. 4456 * Return 4457 * 0 on success, or a negative error in case of failure: 4458 * 4459 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4460 * 4461 * u64 bpf_sk_cgroup_id(void *sk) 4462 * Description 4463 * Return the cgroup v2 id of the socket *sk*. 4464 * 4465 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4466 * returned from **bpf_sk_lookup_xxx**\ (), 4467 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4468 * same as in **bpf_skb_cgroup_id**\ (). 4469 * 4470 * This helper is available only if the kernel was compiled with 4471 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4472 * Return 4473 * The id is returned or 0 in case the id could not be retrieved. 4474 * 4475 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4476 * Description 4477 * Return id of cgroup v2 that is ancestor of cgroup associated 4478 * with the *sk* at the *ancestor_level*. The root cgroup is at 4479 * *ancestor_level* zero and each step down the hierarchy 4480 * increments the level. If *ancestor_level* == level of cgroup 4481 * associated with *sk*, then return value will be same as that 4482 * of **bpf_sk_cgroup_id**\ (). 4483 * 4484 * The helper is useful to implement policies based on cgroups 4485 * that are upper in hierarchy than immediate cgroup associated 4486 * with *sk*. 4487 * 4488 * The format of returned id and helper limitations are same as in 4489 * **bpf_sk_cgroup_id**\ (). 4490 * Return 4491 * The id is returned or 0 in case the id could not be retrieved. 4492 * 4493 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4494 * Description 4495 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4496 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4497 * of new data availability is sent. 4498 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4499 * of new data availability is sent unconditionally. 4500 * If **0** is specified in *flags*, an adaptive notification 4501 * of new data availability is sent. 4502 * 4503 * An adaptive notification is a notification sent whenever the user-space 4504 * process has caught up and consumed all available payloads. In case the user-space 4505 * process is still processing a previous payload, then no notification is needed 4506 * as it will process the newly added payload automatically. 4507 * Return 4508 * 0 on success, or a negative error in case of failure. 4509 * 4510 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4511 * Description 4512 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4513 * *flags* must be 0. 4514 * Return 4515 * Valid pointer with *size* bytes of memory available; NULL, 4516 * otherwise. 4517 * 4518 * void bpf_ringbuf_submit(void *data, u64 flags) 4519 * Description 4520 * Submit reserved ring buffer sample, pointed to by *data*. 4521 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4522 * of new data availability is sent. 4523 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4524 * of new data availability is sent unconditionally. 4525 * If **0** is specified in *flags*, an adaptive notification 4526 * of new data availability is sent. 4527 * 4528 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4529 * Return 4530 * Nothing. Always succeeds. 4531 * 4532 * void bpf_ringbuf_discard(void *data, u64 flags) 4533 * Description 4534 * Discard reserved ring buffer sample, pointed to by *data*. 4535 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4536 * of new data availability is sent. 4537 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4538 * of new data availability is sent unconditionally. 4539 * If **0** is specified in *flags*, an adaptive notification 4540 * of new data availability is sent. 4541 * 4542 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4543 * Return 4544 * Nothing. Always succeeds. 4545 * 4546 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4547 * Description 4548 * Query various characteristics of provided ring buffer. What 4549 * exactly is queries is determined by *flags*: 4550 * 4551 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4552 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4553 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4554 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4555 * 4556 * Data returned is just a momentary snapshot of actual values 4557 * and could be inaccurate, so this facility should be used to 4558 * power heuristics and for reporting, not to make 100% correct 4559 * calculation. 4560 * Return 4561 * Requested value, or 0, if *flags* are not recognized. 4562 * 4563 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4564 * Description 4565 * Change the skbs checksum level by one layer up or down, or 4566 * reset it entirely to none in order to have the stack perform 4567 * checksum validation. The level is applicable to the following 4568 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4569 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4570 * through **bpf_skb_adjust_room**\ () helper with passing in 4571 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4572 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4573 * the UDP header is removed. Similarly, an encap of the latter 4574 * into the former could be accompanied by a helper call to 4575 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4576 * skb is still intended to be processed in higher layers of the 4577 * stack instead of just egressing at tc. 4578 * 4579 * There are three supported level settings at this time: 4580 * 4581 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4582 * with CHECKSUM_UNNECESSARY. 4583 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4584 * with CHECKSUM_UNNECESSARY. 4585 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4586 * sets CHECKSUM_NONE to force checksum validation by the stack. 4587 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4588 * skb->csum_level. 4589 * Return 4590 * 0 on success, or a negative error in case of failure. In the 4591 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4592 * is returned or the error code -EACCES in case the skb is not 4593 * subject to CHECKSUM_UNNECESSARY. 4594 * 4595 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4596 * Description 4597 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4598 * Return 4599 * *sk* if casting is valid, or **NULL** otherwise. 4600 * 4601 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4602 * Description 4603 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4604 * Return 4605 * *sk* if casting is valid, or **NULL** otherwise. 4606 * 4607 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4608 * Description 4609 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4610 * Return 4611 * *sk* if casting is valid, or **NULL** otherwise. 4612 * 4613 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4614 * Description 4615 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4616 * Return 4617 * *sk* if casting is valid, or **NULL** otherwise. 4618 * 4619 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4620 * Description 4621 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4622 * Return 4623 * *sk* if casting is valid, or **NULL** otherwise. 4624 * 4625 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4626 * Description 4627 * Return a user or a kernel stack in bpf program provided buffer. 4628 * Note: the user stack will only be populated if the *task* is 4629 * the current task; all other tasks will return -EOPNOTSUPP. 4630 * To achieve this, the helper needs *task*, which is a valid 4631 * pointer to **struct task_struct**. To store the stacktrace, the 4632 * bpf program provides *buf* with a nonnegative *size*. 4633 * 4634 * The last argument, *flags*, holds the number of stack frames to 4635 * skip (from 0 to 255), masked with 4636 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4637 * the following flags: 4638 * 4639 * **BPF_F_USER_STACK** 4640 * Collect a user space stack instead of a kernel stack. 4641 * The *task* must be the current task. 4642 * **BPF_F_USER_BUILD_ID** 4643 * Collect buildid+offset instead of ips for user stack, 4644 * only valid if **BPF_F_USER_STACK** is also specified. 4645 * 4646 * **bpf_get_task_stack**\ () can collect up to 4647 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4648 * to sufficient large buffer size. Note that 4649 * this limit can be controlled with the **sysctl** program, and 4650 * that it should be manually increased in order to profile long 4651 * user stacks (such as stacks for Java programs). To do so, use: 4652 * 4653 * :: 4654 * 4655 * # sysctl kernel.perf_event_max_stack=<new value> 4656 * Return 4657 * The non-negative copied *buf* length equal to or less than 4658 * *size* on success, or a negative error in case of failure. 4659 * 4660 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4661 * Description 4662 * Load header option. Support reading a particular TCP header 4663 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4664 * 4665 * If *flags* is 0, it will search the option from the 4666 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4667 * has details on what skb_data contains under different 4668 * *skops*\ **->op**. 4669 * 4670 * The first byte of the *searchby_res* specifies the 4671 * kind that it wants to search. 4672 * 4673 * If the searching kind is an experimental kind 4674 * (i.e. 253 or 254 according to RFC6994). It also 4675 * needs to specify the "magic" which is either 4676 * 2 bytes or 4 bytes. It then also needs to 4677 * specify the size of the magic by using 4678 * the 2nd byte which is "kind-length" of a TCP 4679 * header option and the "kind-length" also 4680 * includes the first 2 bytes "kind" and "kind-length" 4681 * itself as a normal TCP header option also does. 4682 * 4683 * For example, to search experimental kind 254 with 4684 * 2 byte magic 0xeB9F, the searchby_res should be 4685 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4686 * 4687 * To search for the standard window scale option (3), 4688 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4689 * Note, kind-length must be 0 for regular option. 4690 * 4691 * Searching for No-Op (0) and End-of-Option-List (1) are 4692 * not supported. 4693 * 4694 * *len* must be at least 2 bytes which is the minimal size 4695 * of a header option. 4696 * 4697 * Supported flags: 4698 * 4699 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4700 * saved_syn packet or the just-received syn packet. 4701 * 4702 * Return 4703 * > 0 when found, the header option is copied to *searchby_res*. 4704 * The return value is the total length copied. On failure, a 4705 * negative error code is returned: 4706 * 4707 * **-EINVAL** if a parameter is invalid. 4708 * 4709 * **-ENOMSG** if the option is not found. 4710 * 4711 * **-ENOENT** if no syn packet is available when 4712 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4713 * 4714 * **-ENOSPC** if there is not enough space. Only *len* number of 4715 * bytes are copied. 4716 * 4717 * **-EFAULT** on failure to parse the header options in the 4718 * packet. 4719 * 4720 * **-EPERM** if the helper cannot be used under the current 4721 * *skops*\ **->op**. 4722 * 4723 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4724 * Description 4725 * Store header option. The data will be copied 4726 * from buffer *from* with length *len* to the TCP header. 4727 * 4728 * The buffer *from* should have the whole option that 4729 * includes the kind, kind-length, and the actual 4730 * option data. The *len* must be at least kind-length 4731 * long. The kind-length does not have to be 4 byte 4732 * aligned. The kernel will take care of the padding 4733 * and setting the 4 bytes aligned value to th->doff. 4734 * 4735 * This helper will check for duplicated option 4736 * by searching the same option in the outgoing skb. 4737 * 4738 * This helper can only be called during 4739 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4740 * 4741 * Return 4742 * 0 on success, or negative error in case of failure: 4743 * 4744 * **-EINVAL** If param is invalid. 4745 * 4746 * **-ENOSPC** if there is not enough space in the header. 4747 * Nothing has been written 4748 * 4749 * **-EEXIST** if the option already exists. 4750 * 4751 * **-EFAULT** on failure to parse the existing header options. 4752 * 4753 * **-EPERM** if the helper cannot be used under the current 4754 * *skops*\ **->op**. 4755 * 4756 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4757 * Description 4758 * Reserve *len* bytes for the bpf header option. The 4759 * space will be used by **bpf_store_hdr_opt**\ () later in 4760 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4761 * 4762 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4763 * the total number of bytes will be reserved. 4764 * 4765 * This helper can only be called during 4766 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4767 * 4768 * Return 4769 * 0 on success, or negative error in case of failure: 4770 * 4771 * **-EINVAL** if a parameter is invalid. 4772 * 4773 * **-ENOSPC** if there is not enough space in the header. 4774 * 4775 * **-EPERM** if the helper cannot be used under the current 4776 * *skops*\ **->op**. 4777 * 4778 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4779 * Description 4780 * Get a bpf_local_storage from an *inode*. 4781 * 4782 * Logically, it could be thought of as getting the value from 4783 * a *map* with *inode* as the **key**. From this 4784 * perspective, the usage is not much different from 4785 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4786 * helper enforces the key must be an inode and the map must also 4787 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4788 * 4789 * Underneath, the value is stored locally at *inode* instead of 4790 * the *map*. The *map* is used as the bpf-local-storage 4791 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4792 * searched against all bpf_local_storage residing at *inode*. 4793 * 4794 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4795 * used such that a new bpf_local_storage will be 4796 * created if one does not exist. *value* can be used 4797 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4798 * the initial value of a bpf_local_storage. If *value* is 4799 * **NULL**, the new bpf_local_storage will be zero initialized. 4800 * Return 4801 * A bpf_local_storage pointer is returned on success. 4802 * 4803 * **NULL** if not found or there was an error in adding 4804 * a new bpf_local_storage. 4805 * 4806 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4807 * Description 4808 * Delete a bpf_local_storage from an *inode*. 4809 * Return 4810 * 0 on success. 4811 * 4812 * **-ENOENT** if the bpf_local_storage cannot be found. 4813 * 4814 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4815 * Description 4816 * Return full path for given **struct path** object, which 4817 * needs to be the kernel BTF *path* object. The path is 4818 * returned in the provided buffer *buf* of size *sz* and 4819 * is zero terminated. 4820 * 4821 * Return 4822 * On success, the strictly positive length of the string, 4823 * including the trailing NUL character. On error, a negative 4824 * value. 4825 * 4826 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4827 * Description 4828 * Read *size* bytes from user space address *user_ptr* and store 4829 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4830 * Return 4831 * 0 on success, or a negative error in case of failure. 4832 * 4833 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4834 * Description 4835 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4836 * using *ptr*->type_id. This value should specify the type 4837 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4838 * can be used to look up vmlinux BTF type ids. Traversing the 4839 * data structure using BTF, the type information and values are 4840 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4841 * the pointer data is carried out to avoid kernel crashes during 4842 * operation. Smaller types can use string space on the stack; 4843 * larger programs can use map data to store the string 4844 * representation. 4845 * 4846 * The string can be subsequently shared with userspace via 4847 * bpf_perf_event_output() or ring buffer interfaces. 4848 * bpf_trace_printk() is to be avoided as it places too small 4849 * a limit on string size to be useful. 4850 * 4851 * *flags* is a combination of 4852 * 4853 * **BTF_F_COMPACT** 4854 * no formatting around type information 4855 * **BTF_F_NONAME** 4856 * no struct/union member names/types 4857 * **BTF_F_PTR_RAW** 4858 * show raw (unobfuscated) pointer values; 4859 * equivalent to printk specifier %px. 4860 * **BTF_F_ZERO** 4861 * show zero-valued struct/union members; they 4862 * are not displayed by default 4863 * 4864 * Return 4865 * The number of bytes that were written (or would have been 4866 * written if output had to be truncated due to string size), 4867 * or a negative error in cases of failure. 4868 * 4869 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4870 * Description 4871 * Use BTF to write to seq_write a string representation of 4872 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4873 * *flags* are identical to those used for bpf_snprintf_btf. 4874 * Return 4875 * 0 on success or a negative error in case of failure. 4876 * 4877 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4878 * Description 4879 * See **bpf_get_cgroup_classid**\ () for the main description. 4880 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4881 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4882 * associated socket instead of the current process. 4883 * Return 4884 * The id is returned or 0 in case the id could not be retrieved. 4885 * 4886 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4887 * Description 4888 * Redirect the packet to another net device of index *ifindex* 4889 * and fill in L2 addresses from neighboring subsystem. This helper 4890 * is somewhat similar to **bpf_redirect**\ (), except that it 4891 * populates L2 addresses as well, meaning, internally, the helper 4892 * relies on the neighbor lookup for the L2 address of the nexthop. 4893 * 4894 * The helper will perform a FIB lookup based on the skb's 4895 * networking header to get the address of the next hop, unless 4896 * this is supplied by the caller in the *params* argument. The 4897 * *plen* argument indicates the len of *params* and should be set 4898 * to 0 if *params* is NULL. 4899 * 4900 * The *flags* argument is reserved and must be 0. The helper is 4901 * currently only supported for tc BPF program types, and enabled 4902 * for IPv4 and IPv6 protocols. 4903 * Return 4904 * The helper returns **TC_ACT_REDIRECT** on success or 4905 * **TC_ACT_SHOT** on error. 4906 * 4907 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4908 * Description 4909 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4910 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4911 * extern variable decorated with '__ksym'. For ksym, there is a 4912 * global var (either static or global) defined of the same name 4913 * in the kernel. The ksym is percpu if the global var is percpu. 4914 * The returned pointer points to the global percpu var on *cpu*. 4915 * 4916 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4917 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4918 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4919 * bpf_per_cpu_ptr() must check the returned value. 4920 * Return 4921 * A pointer pointing to the kernel percpu variable on *cpu*, or 4922 * NULL, if *cpu* is invalid. 4923 * 4924 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4925 * Description 4926 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4927 * pointer to the percpu kernel variable on this cpu. See the 4928 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4929 * 4930 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4931 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4932 * never return NULL. 4933 * Return 4934 * A pointer pointing to the kernel percpu variable on this cpu. 4935 * 4936 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4937 * Description 4938 * Redirect the packet to another net device of index *ifindex*. 4939 * This helper is somewhat similar to **bpf_redirect**\ (), except 4940 * that the redirection happens to the *ifindex*' peer device and 4941 * the netns switch takes place from ingress to ingress without 4942 * going through the CPU's backlog queue. 4943 * 4944 * The *flags* argument is reserved and must be 0. The helper is 4945 * currently only supported for tc BPF program types at the 4946 * ingress hook and for veth and netkit target device types. The 4947 * peer device must reside in a different network namespace. 4948 * Return 4949 * The helper returns **TC_ACT_REDIRECT** on success or 4950 * **TC_ACT_SHOT** on error. 4951 * 4952 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4953 * Description 4954 * Get a bpf_local_storage from the *task*. 4955 * 4956 * Logically, it could be thought of as getting the value from 4957 * a *map* with *task* as the **key**. From this 4958 * perspective, the usage is not much different from 4959 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4960 * helper enforces the key must be a task_struct and the map must also 4961 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4962 * 4963 * Underneath, the value is stored locally at *task* instead of 4964 * the *map*. The *map* is used as the bpf-local-storage 4965 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4966 * searched against all bpf_local_storage residing at *task*. 4967 * 4968 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4969 * used such that a new bpf_local_storage will be 4970 * created if one does not exist. *value* can be used 4971 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4972 * the initial value of a bpf_local_storage. If *value* is 4973 * **NULL**, the new bpf_local_storage will be zero initialized. 4974 * Return 4975 * A bpf_local_storage pointer is returned on success. 4976 * 4977 * **NULL** if not found or there was an error in adding 4978 * a new bpf_local_storage. 4979 * 4980 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4981 * Description 4982 * Delete a bpf_local_storage from a *task*. 4983 * Return 4984 * 0 on success. 4985 * 4986 * **-ENOENT** if the bpf_local_storage cannot be found. 4987 * 4988 * struct task_struct *bpf_get_current_task_btf(void) 4989 * Description 4990 * Return a BTF pointer to the "current" task. 4991 * This pointer can also be used in helpers that accept an 4992 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4993 * Return 4994 * Pointer to the current task. 4995 * 4996 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4997 * Description 4998 * Set or clear certain options on *bprm*: 4999 * 5000 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 5001 * which sets the **AT_SECURE** auxv for glibc. The bit 5002 * is cleared if the flag is not specified. 5003 * Return 5004 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 5005 * 5006 * u64 bpf_ktime_get_coarse_ns(void) 5007 * Description 5008 * Return a coarse-grained version of the time elapsed since 5009 * system boot, in nanoseconds. Does not include time the system 5010 * was suspended. 5011 * 5012 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 5013 * Return 5014 * Current *ktime*. 5015 * 5016 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 5017 * Description 5018 * Returns the stored IMA hash of the *inode* (if it's available). 5019 * If the hash is larger than *size*, then only *size* 5020 * bytes will be copied to *dst* 5021 * Return 5022 * The **hash_algo** is returned on success, 5023 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 5024 * invalid arguments are passed. 5025 * 5026 * struct socket *bpf_sock_from_file(struct file *file) 5027 * Description 5028 * If the given file represents a socket, returns the associated 5029 * socket. 5030 * Return 5031 * A pointer to a struct socket on success or NULL if the file is 5032 * not a socket. 5033 * 5034 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 5035 * Description 5036 * Check packet size against exceeding MTU of net device (based 5037 * on *ifindex*). This helper will likely be used in combination 5038 * with helpers that adjust/change the packet size. 5039 * 5040 * The argument *len_diff* can be used for querying with a planned 5041 * size change. This allows to check MTU prior to changing packet 5042 * ctx. Providing a *len_diff* adjustment that is larger than the 5043 * actual packet size (resulting in negative packet size) will in 5044 * principle not exceed the MTU, which is why it is not considered 5045 * a failure. Other BPF helpers are needed for performing the 5046 * planned size change; therefore the responsibility for catching 5047 * a negative packet size belongs in those helpers. 5048 * 5049 * Specifying *ifindex* zero means the MTU check is performed 5050 * against the current net device. This is practical if this isn't 5051 * used prior to redirect. 5052 * 5053 * On input *mtu_len* must be a valid pointer, else verifier will 5054 * reject BPF program. If the value *mtu_len* is initialized to 5055 * zero then the ctx packet size is use. When value *mtu_len* is 5056 * provided as input this specify the L3 length that the MTU check 5057 * is done against. Remember XDP and TC length operate at L2, but 5058 * this value is L3 as this correlate to MTU and IP-header tot_len 5059 * values which are L3 (similar behavior as bpf_fib_lookup). 5060 * 5061 * The Linux kernel route table can configure MTUs on a more 5062 * specific per route level, which is not provided by this helper. 5063 * For route level MTU checks use the **bpf_fib_lookup**\ () 5064 * helper. 5065 * 5066 * *ctx* is either **struct xdp_md** for XDP programs or 5067 * **struct sk_buff** for tc cls_act programs. 5068 * 5069 * The *flags* argument can be a combination of one or more of the 5070 * following values: 5071 * 5072 * **BPF_MTU_CHK_SEGS** 5073 * This flag will only works for *ctx* **struct sk_buff**. 5074 * If packet context contains extra packet segment buffers 5075 * (often knows as GSO skb), then MTU check is harder to 5076 * check at this point, because in transmit path it is 5077 * possible for the skb packet to get re-segmented 5078 * (depending on net device features). This could still be 5079 * a MTU violation, so this flag enables performing MTU 5080 * check against segments, with a different violation 5081 * return code to tell it apart. Check cannot use len_diff. 5082 * 5083 * On return *mtu_len* pointer contains the MTU value of the net 5084 * device. Remember the net device configured MTU is the L3 size, 5085 * which is returned here and XDP and TC length operate at L2. 5086 * Helper take this into account for you, but remember when using 5087 * MTU value in your BPF-code. 5088 * 5089 * Return 5090 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5091 * 5092 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5093 * 5094 * MTU violations return positive values, but also populate MTU 5095 * value in *mtu_len* pointer, as this can be needed for 5096 * implementing PMTU handing: 5097 * 5098 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5099 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5100 * 5101 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5102 * Description 5103 * For each element in **map**, call **callback_fn** function with 5104 * **map**, **callback_ctx** and other map-specific parameters. 5105 * The **callback_fn** should be a static function and 5106 * the **callback_ctx** should be a pointer to the stack. 5107 * The **flags** is used to control certain aspects of the helper. 5108 * Currently, the **flags** must be 0. 5109 * 5110 * The following are a list of supported map types and their 5111 * respective expected callback signatures: 5112 * 5113 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5114 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5115 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5116 * 5117 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5118 * 5119 * For per_cpu maps, the map_value is the value on the cpu where the 5120 * bpf_prog is running. 5121 * 5122 * If **callback_fn** return 0, the helper will continue to the next 5123 * element. If return value is 1, the helper will skip the rest of 5124 * elements and return. Other return values are not used now. 5125 * 5126 * Return 5127 * The number of traversed map elements for success, **-EINVAL** for 5128 * invalid **flags**. 5129 * 5130 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5131 * Description 5132 * Outputs a string into the **str** buffer of size **str_size** 5133 * based on a format string stored in a read-only map pointed by 5134 * **fmt**. 5135 * 5136 * Each format specifier in **fmt** corresponds to one u64 element 5137 * in the **data** array. For strings and pointers where pointees 5138 * are accessed, only the pointer values are stored in the *data* 5139 * array. The *data_len* is the size of *data* in bytes - must be 5140 * a multiple of 8. 5141 * 5142 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5143 * memory. Reading kernel memory may fail due to either invalid 5144 * address or valid address but requiring a major memory fault. If 5145 * reading kernel memory fails, the string for **%s** will be an 5146 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5147 * Not returning error to bpf program is consistent with what 5148 * **bpf_trace_printk**\ () does for now. 5149 * 5150 * Return 5151 * The strictly positive length of the formatted string, including 5152 * the trailing zero character. If the return value is greater than 5153 * **str_size**, **str** contains a truncated string, guaranteed to 5154 * be zero-terminated except when **str_size** is 0. 5155 * 5156 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5157 * 5158 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5159 * Description 5160 * Execute bpf syscall with given arguments. 5161 * Return 5162 * A syscall result. 5163 * 5164 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5165 * Description 5166 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5167 * Return 5168 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5169 * 5170 * long bpf_sys_close(u32 fd) 5171 * Description 5172 * Execute close syscall for given FD. 5173 * Return 5174 * A syscall result. 5175 * 5176 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5177 * Description 5178 * Initialize the timer. 5179 * First 4 bits of *flags* specify clockid. 5180 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5181 * All other bits of *flags* are reserved. 5182 * The verifier will reject the program if *timer* is not from 5183 * the same *map*. 5184 * Return 5185 * 0 on success. 5186 * **-EBUSY** if *timer* is already initialized. 5187 * **-EINVAL** if invalid *flags* are passed. 5188 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5189 * The user space should either hold a file descriptor to a map with timers 5190 * or pin such map in bpffs. When map is unpinned or file descriptor is 5191 * closed all timers in the map will be cancelled and freed. 5192 * 5193 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5194 * Description 5195 * Configure the timer to call *callback_fn* static function. 5196 * Return 5197 * 0 on success. 5198 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5199 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5200 * The user space should either hold a file descriptor to a map with timers 5201 * or pin such map in bpffs. When map is unpinned or file descriptor is 5202 * closed all timers in the map will be cancelled and freed. 5203 * 5204 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5205 * Description 5206 * Set timer expiration N nanoseconds from the current time. The 5207 * configured callback will be invoked in soft irq context on some cpu 5208 * and will not repeat unless another bpf_timer_start() is made. 5209 * In such case the next invocation can migrate to a different cpu. 5210 * Since struct bpf_timer is a field inside map element the map 5211 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5212 * of BPF program to make sure that callback_fn code stays valid. 5213 * When user space reference to a map reaches zero all timers 5214 * in a map are cancelled and corresponding program's refcnts are 5215 * decremented. This is done to make sure that Ctrl-C of a user 5216 * process doesn't leave any timers running. If map is pinned in 5217 * bpffs the callback_fn can re-arm itself indefinitely. 5218 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5219 * cancel and free the timer in the given map element. 5220 * The map can contain timers that invoke callback_fn-s from different 5221 * programs. The same callback_fn can serve different timers from 5222 * different maps if key/value layout matches across maps. 5223 * Every bpf_timer_set_callback() can have different callback_fn. 5224 * 5225 * *flags* can be one of: 5226 * 5227 * **BPF_F_TIMER_ABS** 5228 * Start the timer in absolute expire value instead of the 5229 * default relative one. 5230 * **BPF_F_TIMER_CPU_PIN** 5231 * Timer will be pinned to the CPU of the caller. 5232 * 5233 * Return 5234 * 0 on success. 5235 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5236 * or invalid *flags* are passed. 5237 * 5238 * long bpf_timer_cancel(struct bpf_timer *timer) 5239 * Description 5240 * Cancel the timer and wait for callback_fn to finish if it was running. 5241 * Return 5242 * 0 if the timer was not active. 5243 * 1 if the timer was active. 5244 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5245 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5246 * own timer which would have led to a deadlock otherwise. 5247 * 5248 * u64 bpf_get_func_ip(void *ctx) 5249 * Description 5250 * Get address of the traced function (for tracing and kprobe programs). 5251 * 5252 * When called for kprobe program attached as uprobe it returns 5253 * probe address for both entry and return uprobe. 5254 * 5255 * Return 5256 * Address of the traced function for kprobe. 5257 * 0 for kprobes placed within the function (not at the entry). 5258 * Address of the probe for uprobe and return uprobe. 5259 * 5260 * u64 bpf_get_attach_cookie(void *ctx) 5261 * Description 5262 * Get bpf_cookie value provided (optionally) during the program 5263 * attachment. It might be different for each individual 5264 * attachment, even if BPF program itself is the same. 5265 * Expects BPF program context *ctx* as a first argument. 5266 * 5267 * Supported for the following program types: 5268 * - kprobe/uprobe; 5269 * - tracepoint; 5270 * - perf_event. 5271 * Return 5272 * Value specified by user at BPF link creation/attachment time 5273 * or 0, if it was not specified. 5274 * 5275 * long bpf_task_pt_regs(struct task_struct *task) 5276 * Description 5277 * Get the struct pt_regs associated with **task**. 5278 * Return 5279 * A pointer to struct pt_regs. 5280 * 5281 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5282 * Description 5283 * Get branch trace from hardware engines like Intel LBR. The 5284 * hardware engine is stopped shortly after the helper is 5285 * called. Therefore, the user need to filter branch entries 5286 * based on the actual use case. To capture branch trace 5287 * before the trigger point of the BPF program, the helper 5288 * should be called at the beginning of the BPF program. 5289 * 5290 * The data is stored as struct perf_branch_entry into output 5291 * buffer *entries*. *size* is the size of *entries* in bytes. 5292 * *flags* is reserved for now and must be zero. 5293 * 5294 * Return 5295 * On success, number of bytes written to *buf*. On error, a 5296 * negative value. 5297 * 5298 * **-EINVAL** if *flags* is not zero. 5299 * 5300 * **-ENOENT** if architecture does not support branch records. 5301 * 5302 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5303 * Description 5304 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5305 * to format and can handle more format args as a result. 5306 * 5307 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5308 * Return 5309 * The number of bytes written to the buffer, or a negative error 5310 * in case of failure. 5311 * 5312 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5313 * Description 5314 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5315 * Return 5316 * *sk* if casting is valid, or **NULL** otherwise. 5317 * 5318 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5319 * Description 5320 * Get the address of a kernel symbol, returned in *res*. *res* is 5321 * set to 0 if the symbol is not found. 5322 * Return 5323 * On success, zero. On error, a negative value. 5324 * 5325 * **-EINVAL** if *flags* is not zero. 5326 * 5327 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5328 * 5329 * **-ENOENT** if symbol is not found. 5330 * 5331 * **-EPERM** if caller does not have permission to obtain kernel address. 5332 * 5333 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5334 * Description 5335 * Find vma of *task* that contains *addr*, call *callback_fn* 5336 * function with *task*, *vma*, and *callback_ctx*. 5337 * The *callback_fn* should be a static function and 5338 * the *callback_ctx* should be a pointer to the stack. 5339 * The *flags* is used to control certain aspects of the helper. 5340 * Currently, the *flags* must be 0. 5341 * 5342 * The expected callback signature is 5343 * 5344 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5345 * 5346 * Return 5347 * 0 on success. 5348 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5349 * **-EBUSY** if failed to try lock mmap_lock. 5350 * **-EINVAL** for invalid **flags**. 5351 * 5352 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5353 * Description 5354 * For **nr_loops**, call **callback_fn** function 5355 * with **callback_ctx** as the context parameter. 5356 * The **callback_fn** should be a static function and 5357 * the **callback_ctx** should be a pointer to the stack. 5358 * The **flags** is used to control certain aspects of the helper. 5359 * Currently, the **flags** must be 0. Currently, nr_loops is 5360 * limited to 1 << 23 (~8 million) loops. 5361 * 5362 * long (\*callback_fn)(u32 index, void \*ctx); 5363 * 5364 * where **index** is the current index in the loop. The index 5365 * is zero-indexed. 5366 * 5367 * If **callback_fn** returns 0, the helper will continue to the next 5368 * loop. If return value is 1, the helper will skip the rest of 5369 * the loops and return. Other return values are not used now, 5370 * and will be rejected by the verifier. 5371 * 5372 * Return 5373 * The number of loops performed, **-EINVAL** for invalid **flags**, 5374 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5375 * 5376 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5377 * Description 5378 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5379 * to be null-terminated and **s1_sz** is the maximum storage 5380 * size of **s1**. **s2** must be a read-only string. 5381 * Return 5382 * An integer less than, equal to, or greater than zero 5383 * if the first **s1_sz** bytes of **s1** is found to be 5384 * less than, to match, or be greater than **s2**. 5385 * 5386 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5387 * Description 5388 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5389 * returned in **value**. 5390 * 5391 * Return 5392 * 0 on success. 5393 * **-EINVAL** if n >= argument register count of traced function. 5394 * 5395 * long bpf_get_func_ret(void *ctx, u64 *value) 5396 * Description 5397 * Get return value of the traced function (for tracing programs) 5398 * in **value**. 5399 * 5400 * Return 5401 * 0 on success. 5402 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5403 * 5404 * long bpf_get_func_arg_cnt(void *ctx) 5405 * Description 5406 * Get number of registers of the traced function (for tracing programs) where 5407 * function arguments are stored in these registers. 5408 * 5409 * Return 5410 * The number of argument registers of the traced function. 5411 * 5412 * int bpf_get_retval(void) 5413 * Description 5414 * Get the BPF program's return value that will be returned to the upper layers. 5415 * 5416 * This helper is currently supported by cgroup programs and only by the hooks 5417 * where BPF program's return value is returned to the userspace via errno. 5418 * Return 5419 * The BPF program's return value. 5420 * 5421 * int bpf_set_retval(int retval) 5422 * Description 5423 * Set the BPF program's return value that will be returned to the upper layers. 5424 * 5425 * This helper is currently supported by cgroup programs and only by the hooks 5426 * where BPF program's return value is returned to the userspace via errno. 5427 * 5428 * Note that there is the following corner case where the program exports an error 5429 * via bpf_set_retval but signals success via 'return 1': 5430 * 5431 * bpf_set_retval(-EPERM); 5432 * return 1; 5433 * 5434 * In this case, the BPF program's return value will use helper's -EPERM. This 5435 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5436 * 5437 * Return 5438 * 0 on success, or a negative error in case of failure. 5439 * 5440 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5441 * Description 5442 * Get the total size of a given xdp buff (linear and paged area) 5443 * Return 5444 * The total size of a given xdp buffer. 5445 * 5446 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5447 * Description 5448 * This helper is provided as an easy way to load data from a 5449 * xdp buffer. It can be used to load *len* bytes from *offset* from 5450 * the frame associated to *xdp_md*, into the buffer pointed by 5451 * *buf*. 5452 * Return 5453 * 0 on success, or a negative error in case of failure. 5454 * 5455 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5456 * Description 5457 * Store *len* bytes from buffer *buf* into the frame 5458 * associated to *xdp_md*, at *offset*. 5459 * Return 5460 * 0 on success, or a negative error in case of failure. 5461 * 5462 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5463 * Description 5464 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5465 * address space, and stores the data in *dst*. *flags* is not 5466 * used yet and is provided for future extensibility. This helper 5467 * can only be used by sleepable programs. 5468 * Return 5469 * 0 on success, or a negative error in case of failure. On error 5470 * *dst* buffer is zeroed out. 5471 * 5472 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5473 * Description 5474 * Change the __sk_buff->tstamp_type to *tstamp_type* 5475 * and set *tstamp* to the __sk_buff->tstamp together. 5476 * 5477 * If there is no need to change the __sk_buff->tstamp_type, 5478 * the tstamp value can be directly written to __sk_buff->tstamp 5479 * instead. 5480 * 5481 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5482 * will be kept during bpf_redirect_*(). A non zero 5483 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5484 * *tstamp_type*. 5485 * 5486 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5487 * with a zero *tstamp*. 5488 * 5489 * Only IPv4 and IPv6 skb->protocol are supported. 5490 * 5491 * This function is most useful when it needs to set a 5492 * mono delivery time to __sk_buff->tstamp and then 5493 * bpf_redirect_*() to the egress of an iface. For example, 5494 * changing the (rcv) timestamp in __sk_buff->tstamp at 5495 * ingress to a mono delivery time and then bpf_redirect_*() 5496 * to sch_fq@phy-dev. 5497 * Return 5498 * 0 on success. 5499 * **-EINVAL** for invalid input 5500 * **-EOPNOTSUPP** for unsupported protocol 5501 * 5502 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5503 * Description 5504 * Returns a calculated IMA hash of the *file*. 5505 * If the hash is larger than *size*, then only *size* 5506 * bytes will be copied to *dst* 5507 * Return 5508 * The **hash_algo** is returned on success, 5509 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5510 * invalid arguments are passed. 5511 * 5512 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5513 * Description 5514 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5515 * old value. *ptr* can be NULL, otherwise it must be a referenced 5516 * pointer which will be released when this helper is called. 5517 * Return 5518 * The old value of kptr (which can be NULL). The returned pointer 5519 * if not NULL, is a reference which must be released using its 5520 * corresponding release function, or moved into a BPF map before 5521 * program exit. 5522 * 5523 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5524 * Description 5525 * Perform a lookup in *percpu map* for an entry associated to 5526 * *key* on *cpu*. 5527 * Return 5528 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5529 * was found or *cpu* is invalid. 5530 * 5531 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5532 * Description 5533 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5534 * Return 5535 * *sk* if casting is valid, or **NULL** otherwise. 5536 * 5537 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5538 * Description 5539 * Get a dynptr to local memory *data*. 5540 * 5541 * *data* must be a ptr to a map value. 5542 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5543 * *flags* is currently unused. 5544 * Return 5545 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5546 * -EINVAL if flags is not 0. 5547 * 5548 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5549 * Description 5550 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5551 * through the dynptr interface. *flags* must be 0. 5552 * 5553 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5554 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5555 * reservation fails. This is enforced by the verifier. 5556 * Return 5557 * 0 on success, or a negative error in case of failure. 5558 * 5559 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5560 * Description 5561 * Submit reserved ring buffer sample, pointed to by *data*, 5562 * through the dynptr interface. This is a no-op if the dynptr is 5563 * invalid/null. 5564 * 5565 * For more information on *flags*, please see 5566 * 'bpf_ringbuf_submit'. 5567 * Return 5568 * Nothing. Always succeeds. 5569 * 5570 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5571 * Description 5572 * Discard reserved ring buffer sample through the dynptr 5573 * interface. This is a no-op if the dynptr is invalid/null. 5574 * 5575 * For more information on *flags*, please see 5576 * 'bpf_ringbuf_discard'. 5577 * Return 5578 * Nothing. Always succeeds. 5579 * 5580 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5581 * Description 5582 * Read *len* bytes from *src* into *dst*, starting from *offset* 5583 * into *src*. 5584 * *flags* is currently unused. 5585 * Return 5586 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5587 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5588 * *flags* is not 0. 5589 * 5590 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5591 * Description 5592 * Write *len* bytes from *src* into *dst*, starting from *offset* 5593 * into *dst*. 5594 * 5595 * *flags* must be 0 except for skb-type dynptrs. 5596 * 5597 * For skb-type dynptrs: 5598 * * All data slices of the dynptr are automatically 5599 * invalidated after **bpf_dynptr_write**\ (). This is 5600 * because writing may pull the skb and change the 5601 * underlying packet buffer. 5602 * 5603 * * For *flags*, please see the flags accepted by 5604 * **bpf_skb_store_bytes**\ (). 5605 * Return 5606 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5607 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5608 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5609 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5610 * 5611 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5612 * Description 5613 * Get a pointer to the underlying dynptr data. 5614 * 5615 * *len* must be a statically known value. The returned data slice 5616 * is invalidated whenever the dynptr is invalidated. 5617 * 5618 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5619 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5620 * Return 5621 * Pointer to the underlying dynptr data, NULL if the dynptr is 5622 * read-only, if the dynptr is invalid, or if the offset and length 5623 * is out of bounds. 5624 * 5625 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5626 * Description 5627 * Try to issue a SYN cookie for the packet with corresponding 5628 * IPv4/TCP headers, *iph* and *th*, without depending on a 5629 * listening socket. 5630 * 5631 * *iph* points to the IPv4 header. 5632 * 5633 * *th* points to the start of the TCP header, while *th_len* 5634 * contains the length of the TCP header (at least 5635 * **sizeof**\ (**struct tcphdr**)). 5636 * Return 5637 * On success, lower 32 bits hold the generated SYN cookie in 5638 * followed by 16 bits which hold the MSS value for that cookie, 5639 * and the top 16 bits are unused. 5640 * 5641 * On failure, the returned value is one of the following: 5642 * 5643 * **-EINVAL** if *th_len* is invalid. 5644 * 5645 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5646 * Description 5647 * Try to issue a SYN cookie for the packet with corresponding 5648 * IPv6/TCP headers, *iph* and *th*, without depending on a 5649 * listening socket. 5650 * 5651 * *iph* points to the IPv6 header. 5652 * 5653 * *th* points to the start of the TCP header, while *th_len* 5654 * contains the length of the TCP header (at least 5655 * **sizeof**\ (**struct tcphdr**)). 5656 * Return 5657 * On success, lower 32 bits hold the generated SYN cookie in 5658 * followed by 16 bits which hold the MSS value for that cookie, 5659 * and the top 16 bits are unused. 5660 * 5661 * On failure, the returned value is one of the following: 5662 * 5663 * **-EINVAL** if *th_len* is invalid. 5664 * 5665 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5666 * 5667 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5668 * Description 5669 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5670 * without depending on a listening socket. 5671 * 5672 * *iph* points to the IPv4 header. 5673 * 5674 * *th* points to the TCP header. 5675 * Return 5676 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5677 * 5678 * On failure, the returned value is one of the following: 5679 * 5680 * **-EACCES** if the SYN cookie is not valid. 5681 * 5682 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5683 * Description 5684 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5685 * without depending on a listening socket. 5686 * 5687 * *iph* points to the IPv6 header. 5688 * 5689 * *th* points to the TCP header. 5690 * Return 5691 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5692 * 5693 * On failure, the returned value is one of the following: 5694 * 5695 * **-EACCES** if the SYN cookie is not valid. 5696 * 5697 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5698 * 5699 * u64 bpf_ktime_get_tai_ns(void) 5700 * Description 5701 * A nonsettable system-wide clock derived from wall-clock time but 5702 * ignoring leap seconds. This clock does not experience 5703 * discontinuities and backwards jumps caused by NTP inserting leap 5704 * seconds as CLOCK_REALTIME does. 5705 * 5706 * See: **clock_gettime**\ (**CLOCK_TAI**) 5707 * Return 5708 * Current *ktime*. 5709 * 5710 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5711 * Description 5712 * Drain samples from the specified user ring buffer, and invoke 5713 * the provided callback for each such sample: 5714 * 5715 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5716 * 5717 * If **callback_fn** returns 0, the helper will continue to try 5718 * and drain the next sample, up to a maximum of 5719 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5720 * the helper will skip the rest of the samples and return. Other 5721 * return values are not used now, and will be rejected by the 5722 * verifier. 5723 * Return 5724 * The number of drained samples if no error was encountered while 5725 * draining samples, or 0 if no samples were present in the ring 5726 * buffer. If a user-space producer was epoll-waiting on this map, 5727 * and at least one sample was drained, they will receive an event 5728 * notification notifying them of available space in the ring 5729 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5730 * function, no wakeup notification will be sent. If the 5731 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5732 * be sent even if no sample was drained. 5733 * 5734 * On failure, the returned value is one of the following: 5735 * 5736 * **-EBUSY** if the ring buffer is contended, and another calling 5737 * context was concurrently draining the ring buffer. 5738 * 5739 * **-EINVAL** if user-space is not properly tracking the ring 5740 * buffer due to the producer position not being aligned to 8 5741 * bytes, a sample not being aligned to 8 bytes, or the producer 5742 * position not matching the advertised length of a sample. 5743 * 5744 * **-E2BIG** if user-space has tried to publish a sample which is 5745 * larger than the size of the ring buffer, or which cannot fit 5746 * within a struct bpf_dynptr. 5747 * 5748 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5749 * Description 5750 * Get a bpf_local_storage from the *cgroup*. 5751 * 5752 * Logically, it could be thought of as getting the value from 5753 * a *map* with *cgroup* as the **key**. From this 5754 * perspective, the usage is not much different from 5755 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5756 * helper enforces the key must be a cgroup struct and the map must also 5757 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5758 * 5759 * In reality, the local-storage value is embedded directly inside of the 5760 * *cgroup* object itself, rather than being located in the 5761 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5762 * queried for some *map* on a *cgroup* object, the kernel will perform an 5763 * O(n) iteration over all of the live local-storage values for that 5764 * *cgroup* object until the local-storage value for the *map* is found. 5765 * 5766 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5767 * used such that a new bpf_local_storage will be 5768 * created if one does not exist. *value* can be used 5769 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5770 * the initial value of a bpf_local_storage. If *value* is 5771 * **NULL**, the new bpf_local_storage will be zero initialized. 5772 * Return 5773 * A bpf_local_storage pointer is returned on success. 5774 * 5775 * **NULL** if not found or there was an error in adding 5776 * a new bpf_local_storage. 5777 * 5778 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5779 * Description 5780 * Delete a bpf_local_storage from a *cgroup*. 5781 * Return 5782 * 0 on success. 5783 * 5784 * **-ENOENT** if the bpf_local_storage cannot be found. 5785 */ 5786#define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5787 FN(unspec, 0, ##ctx) \ 5788 FN(map_lookup_elem, 1, ##ctx) \ 5789 FN(map_update_elem, 2, ##ctx) \ 5790 FN(map_delete_elem, 3, ##ctx) \ 5791 FN(probe_read, 4, ##ctx) \ 5792 FN(ktime_get_ns, 5, ##ctx) \ 5793 FN(trace_printk, 6, ##ctx) \ 5794 FN(get_prandom_u32, 7, ##ctx) \ 5795 FN(get_smp_processor_id, 8, ##ctx) \ 5796 FN(skb_store_bytes, 9, ##ctx) \ 5797 FN(l3_csum_replace, 10, ##ctx) \ 5798 FN(l4_csum_replace, 11, ##ctx) \ 5799 FN(tail_call, 12, ##ctx) \ 5800 FN(clone_redirect, 13, ##ctx) \ 5801 FN(get_current_pid_tgid, 14, ##ctx) \ 5802 FN(get_current_uid_gid, 15, ##ctx) \ 5803 FN(get_current_comm, 16, ##ctx) \ 5804 FN(get_cgroup_classid, 17, ##ctx) \ 5805 FN(skb_vlan_push, 18, ##ctx) \ 5806 FN(skb_vlan_pop, 19, ##ctx) \ 5807 FN(skb_get_tunnel_key, 20, ##ctx) \ 5808 FN(skb_set_tunnel_key, 21, ##ctx) \ 5809 FN(perf_event_read, 22, ##ctx) \ 5810 FN(redirect, 23, ##ctx) \ 5811 FN(get_route_realm, 24, ##ctx) \ 5812 FN(perf_event_output, 25, ##ctx) \ 5813 FN(skb_load_bytes, 26, ##ctx) \ 5814 FN(get_stackid, 27, ##ctx) \ 5815 FN(csum_diff, 28, ##ctx) \ 5816 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5817 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5818 FN(skb_change_proto, 31, ##ctx) \ 5819 FN(skb_change_type, 32, ##ctx) \ 5820 FN(skb_under_cgroup, 33, ##ctx) \ 5821 FN(get_hash_recalc, 34, ##ctx) \ 5822 FN(get_current_task, 35, ##ctx) \ 5823 FN(probe_write_user, 36, ##ctx) \ 5824 FN(current_task_under_cgroup, 37, ##ctx) \ 5825 FN(skb_change_tail, 38, ##ctx) \ 5826 FN(skb_pull_data, 39, ##ctx) \ 5827 FN(csum_update, 40, ##ctx) \ 5828 FN(set_hash_invalid, 41, ##ctx) \ 5829 FN(get_numa_node_id, 42, ##ctx) \ 5830 FN(skb_change_head, 43, ##ctx) \ 5831 FN(xdp_adjust_head, 44, ##ctx) \ 5832 FN(probe_read_str, 45, ##ctx) \ 5833 FN(get_socket_cookie, 46, ##ctx) \ 5834 FN(get_socket_uid, 47, ##ctx) \ 5835 FN(set_hash, 48, ##ctx) \ 5836 FN(setsockopt, 49, ##ctx) \ 5837 FN(skb_adjust_room, 50, ##ctx) \ 5838 FN(redirect_map, 51, ##ctx) \ 5839 FN(sk_redirect_map, 52, ##ctx) \ 5840 FN(sock_map_update, 53, ##ctx) \ 5841 FN(xdp_adjust_meta, 54, ##ctx) \ 5842 FN(perf_event_read_value, 55, ##ctx) \ 5843 FN(perf_prog_read_value, 56, ##ctx) \ 5844 FN(getsockopt, 57, ##ctx) \ 5845 FN(override_return, 58, ##ctx) \ 5846 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5847 FN(msg_redirect_map, 60, ##ctx) \ 5848 FN(msg_apply_bytes, 61, ##ctx) \ 5849 FN(msg_cork_bytes, 62, ##ctx) \ 5850 FN(msg_pull_data, 63, ##ctx) \ 5851 FN(bind, 64, ##ctx) \ 5852 FN(xdp_adjust_tail, 65, ##ctx) \ 5853 FN(skb_get_xfrm_state, 66, ##ctx) \ 5854 FN(get_stack, 67, ##ctx) \ 5855 FN(skb_load_bytes_relative, 68, ##ctx) \ 5856 FN(fib_lookup, 69, ##ctx) \ 5857 FN(sock_hash_update, 70, ##ctx) \ 5858 FN(msg_redirect_hash, 71, ##ctx) \ 5859 FN(sk_redirect_hash, 72, ##ctx) \ 5860 FN(lwt_push_encap, 73, ##ctx) \ 5861 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5862 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5863 FN(lwt_seg6_action, 76, ##ctx) \ 5864 FN(rc_repeat, 77, ##ctx) \ 5865 FN(rc_keydown, 78, ##ctx) \ 5866 FN(skb_cgroup_id, 79, ##ctx) \ 5867 FN(get_current_cgroup_id, 80, ##ctx) \ 5868 FN(get_local_storage, 81, ##ctx) \ 5869 FN(sk_select_reuseport, 82, ##ctx) \ 5870 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5871 FN(sk_lookup_tcp, 84, ##ctx) \ 5872 FN(sk_lookup_udp, 85, ##ctx) \ 5873 FN(sk_release, 86, ##ctx) \ 5874 FN(map_push_elem, 87, ##ctx) \ 5875 FN(map_pop_elem, 88, ##ctx) \ 5876 FN(map_peek_elem, 89, ##ctx) \ 5877 FN(msg_push_data, 90, ##ctx) \ 5878 FN(msg_pop_data, 91, ##ctx) \ 5879 FN(rc_pointer_rel, 92, ##ctx) \ 5880 FN(spin_lock, 93, ##ctx) \ 5881 FN(spin_unlock, 94, ##ctx) \ 5882 FN(sk_fullsock, 95, ##ctx) \ 5883 FN(tcp_sock, 96, ##ctx) \ 5884 FN(skb_ecn_set_ce, 97, ##ctx) \ 5885 FN(get_listener_sock, 98, ##ctx) \ 5886 FN(skc_lookup_tcp, 99, ##ctx) \ 5887 FN(tcp_check_syncookie, 100, ##ctx) \ 5888 FN(sysctl_get_name, 101, ##ctx) \ 5889 FN(sysctl_get_current_value, 102, ##ctx) \ 5890 FN(sysctl_get_new_value, 103, ##ctx) \ 5891 FN(sysctl_set_new_value, 104, ##ctx) \ 5892 FN(strtol, 105, ##ctx) \ 5893 FN(strtoul, 106, ##ctx) \ 5894 FN(sk_storage_get, 107, ##ctx) \ 5895 FN(sk_storage_delete, 108, ##ctx) \ 5896 FN(send_signal, 109, ##ctx) \ 5897 FN(tcp_gen_syncookie, 110, ##ctx) \ 5898 FN(skb_output, 111, ##ctx) \ 5899 FN(probe_read_user, 112, ##ctx) \ 5900 FN(probe_read_kernel, 113, ##ctx) \ 5901 FN(probe_read_user_str, 114, ##ctx) \ 5902 FN(probe_read_kernel_str, 115, ##ctx) \ 5903 FN(tcp_send_ack, 116, ##ctx) \ 5904 FN(send_signal_thread, 117, ##ctx) \ 5905 FN(jiffies64, 118, ##ctx) \ 5906 FN(read_branch_records, 119, ##ctx) \ 5907 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5908 FN(xdp_output, 121, ##ctx) \ 5909 FN(get_netns_cookie, 122, ##ctx) \ 5910 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5911 FN(sk_assign, 124, ##ctx) \ 5912 FN(ktime_get_boot_ns, 125, ##ctx) \ 5913 FN(seq_printf, 126, ##ctx) \ 5914 FN(seq_write, 127, ##ctx) \ 5915 FN(sk_cgroup_id, 128, ##ctx) \ 5916 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5917 FN(ringbuf_output, 130, ##ctx) \ 5918 FN(ringbuf_reserve, 131, ##ctx) \ 5919 FN(ringbuf_submit, 132, ##ctx) \ 5920 FN(ringbuf_discard, 133, ##ctx) \ 5921 FN(ringbuf_query, 134, ##ctx) \ 5922 FN(csum_level, 135, ##ctx) \ 5923 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5924 FN(skc_to_tcp_sock, 137, ##ctx) \ 5925 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5926 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5927 FN(skc_to_udp6_sock, 140, ##ctx) \ 5928 FN(get_task_stack, 141, ##ctx) \ 5929 FN(load_hdr_opt, 142, ##ctx) \ 5930 FN(store_hdr_opt, 143, ##ctx) \ 5931 FN(reserve_hdr_opt, 144, ##ctx) \ 5932 FN(inode_storage_get, 145, ##ctx) \ 5933 FN(inode_storage_delete, 146, ##ctx) \ 5934 FN(d_path, 147, ##ctx) \ 5935 FN(copy_from_user, 148, ##ctx) \ 5936 FN(snprintf_btf, 149, ##ctx) \ 5937 FN(seq_printf_btf, 150, ##ctx) \ 5938 FN(skb_cgroup_classid, 151, ##ctx) \ 5939 FN(redirect_neigh, 152, ##ctx) \ 5940 FN(per_cpu_ptr, 153, ##ctx) \ 5941 FN(this_cpu_ptr, 154, ##ctx) \ 5942 FN(redirect_peer, 155, ##ctx) \ 5943 FN(task_storage_get, 156, ##ctx) \ 5944 FN(task_storage_delete, 157, ##ctx) \ 5945 FN(get_current_task_btf, 158, ##ctx) \ 5946 FN(bprm_opts_set, 159, ##ctx) \ 5947 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5948 FN(ima_inode_hash, 161, ##ctx) \ 5949 FN(sock_from_file, 162, ##ctx) \ 5950 FN(check_mtu, 163, ##ctx) \ 5951 FN(for_each_map_elem, 164, ##ctx) \ 5952 FN(snprintf, 165, ##ctx) \ 5953 FN(sys_bpf, 166, ##ctx) \ 5954 FN(btf_find_by_name_kind, 167, ##ctx) \ 5955 FN(sys_close, 168, ##ctx) \ 5956 FN(timer_init, 169, ##ctx) \ 5957 FN(timer_set_callback, 170, ##ctx) \ 5958 FN(timer_start, 171, ##ctx) \ 5959 FN(timer_cancel, 172, ##ctx) \ 5960 FN(get_func_ip, 173, ##ctx) \ 5961 FN(get_attach_cookie, 174, ##ctx) \ 5962 FN(task_pt_regs, 175, ##ctx) \ 5963 FN(get_branch_snapshot, 176, ##ctx) \ 5964 FN(trace_vprintk, 177, ##ctx) \ 5965 FN(skc_to_unix_sock, 178, ##ctx) \ 5966 FN(kallsyms_lookup_name, 179, ##ctx) \ 5967 FN(find_vma, 180, ##ctx) \ 5968 FN(loop, 181, ##ctx) \ 5969 FN(strncmp, 182, ##ctx) \ 5970 FN(get_func_arg, 183, ##ctx) \ 5971 FN(get_func_ret, 184, ##ctx) \ 5972 FN(get_func_arg_cnt, 185, ##ctx) \ 5973 FN(get_retval, 186, ##ctx) \ 5974 FN(set_retval, 187, ##ctx) \ 5975 FN(xdp_get_buff_len, 188, ##ctx) \ 5976 FN(xdp_load_bytes, 189, ##ctx) \ 5977 FN(xdp_store_bytes, 190, ##ctx) \ 5978 FN(copy_from_user_task, 191, ##ctx) \ 5979 FN(skb_set_tstamp, 192, ##ctx) \ 5980 FN(ima_file_hash, 193, ##ctx) \ 5981 FN(kptr_xchg, 194, ##ctx) \ 5982 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5983 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5984 FN(dynptr_from_mem, 197, ##ctx) \ 5985 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5986 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5987 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5988 FN(dynptr_read, 201, ##ctx) \ 5989 FN(dynptr_write, 202, ##ctx) \ 5990 FN(dynptr_data, 203, ##ctx) \ 5991 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5992 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5993 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5994 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5995 FN(ktime_get_tai_ns, 208, ##ctx) \ 5996 FN(user_ringbuf_drain, 209, ##ctx) \ 5997 FN(cgrp_storage_get, 210, ##ctx) \ 5998 FN(cgrp_storage_delete, 211, ##ctx) \ 5999 /* */ 6000 6001/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 6002 * know or care about integer value that is now passed as second argument 6003 */ 6004#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 6005#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 6006 6007/* integer value in 'imm' field of BPF_CALL instruction selects which helper 6008 * function eBPF program intends to call 6009 */ 6010#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 6011enum bpf_func_id { 6012 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 6013 __BPF_FUNC_MAX_ID, 6014}; 6015#undef __BPF_ENUM_FN 6016 6017/* All flags used by eBPF helper functions, placed here. */ 6018 6019/* BPF_FUNC_skb_store_bytes flags. */ 6020enum { 6021 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 6022 BPF_F_INVALIDATE_HASH = (1ULL << 1), 6023}; 6024 6025/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 6026 * First 4 bits are for passing the header field size. 6027 */ 6028enum { 6029 BPF_F_HDR_FIELD_MASK = 0xfULL, 6030}; 6031 6032/* BPF_FUNC_l4_csum_replace flags. */ 6033enum { 6034 BPF_F_PSEUDO_HDR = (1ULL << 4), 6035 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 6036 BPF_F_MARK_ENFORCE = (1ULL << 6), 6037}; 6038 6039/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 6040enum { 6041 BPF_F_INGRESS = (1ULL << 0), 6042}; 6043 6044/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6045enum { 6046 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6047}; 6048 6049/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6050enum { 6051 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6052 BPF_F_USER_STACK = (1ULL << 8), 6053/* flags used by BPF_FUNC_get_stackid only. */ 6054 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6055 BPF_F_REUSE_STACKID = (1ULL << 10), 6056/* flags used by BPF_FUNC_get_stack only. */ 6057 BPF_F_USER_BUILD_ID = (1ULL << 11), 6058}; 6059 6060/* BPF_FUNC_skb_set_tunnel_key flags. */ 6061enum { 6062 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6063 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6064 BPF_F_SEQ_NUMBER = (1ULL << 3), 6065 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6066}; 6067 6068/* BPF_FUNC_skb_get_tunnel_key flags. */ 6069enum { 6070 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6071}; 6072 6073/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6074 * BPF_FUNC_perf_event_read_value flags. 6075 */ 6076enum { 6077 BPF_F_INDEX_MASK = 0xffffffffULL, 6078 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6079/* BPF_FUNC_perf_event_output for sk_buff input context. */ 6080 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6081}; 6082 6083/* Current network namespace */ 6084enum { 6085 BPF_F_CURRENT_NETNS = (-1L), 6086}; 6087 6088/* BPF_FUNC_csum_level level values. */ 6089enum { 6090 BPF_CSUM_LEVEL_QUERY, 6091 BPF_CSUM_LEVEL_INC, 6092 BPF_CSUM_LEVEL_DEC, 6093 BPF_CSUM_LEVEL_RESET, 6094}; 6095 6096/* BPF_FUNC_skb_adjust_room flags. */ 6097enum { 6098 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6099 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6100 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6101 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6102 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6103 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6104 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6105 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6106 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6107}; 6108 6109enum { 6110 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6111 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6112}; 6113 6114#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6115 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6116 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6117 6118/* BPF_FUNC_sysctl_get_name flags. */ 6119enum { 6120 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6121}; 6122 6123/* BPF_FUNC_<kernel_obj>_storage_get flags */ 6124enum { 6125 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6126 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6127 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6128 */ 6129 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6130}; 6131 6132/* BPF_FUNC_read_branch_records flags. */ 6133enum { 6134 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6135}; 6136 6137/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6138 * BPF_FUNC_bpf_ringbuf_output flags. 6139 */ 6140enum { 6141 BPF_RB_NO_WAKEUP = (1ULL << 0), 6142 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6143}; 6144 6145/* BPF_FUNC_bpf_ringbuf_query flags */ 6146enum { 6147 BPF_RB_AVAIL_DATA = 0, 6148 BPF_RB_RING_SIZE = 1, 6149 BPF_RB_CONS_POS = 2, 6150 BPF_RB_PROD_POS = 3, 6151}; 6152 6153/* BPF ring buffer constants */ 6154enum { 6155 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6156 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6157 BPF_RINGBUF_HDR_SZ = 8, 6158}; 6159 6160/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6161enum { 6162 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6163 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6164}; 6165 6166/* Mode for BPF_FUNC_skb_adjust_room helper. */ 6167enum bpf_adj_room_mode { 6168 BPF_ADJ_ROOM_NET, 6169 BPF_ADJ_ROOM_MAC, 6170}; 6171 6172/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6173enum bpf_hdr_start_off { 6174 BPF_HDR_START_MAC, 6175 BPF_HDR_START_NET, 6176}; 6177 6178/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6179enum bpf_lwt_encap_mode { 6180 BPF_LWT_ENCAP_SEG6, 6181 BPF_LWT_ENCAP_SEG6_INLINE, 6182 BPF_LWT_ENCAP_IP, 6183}; 6184 6185/* Flags for bpf_bprm_opts_set helper */ 6186enum { 6187 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6188}; 6189 6190/* Flags for bpf_redirect_map helper */ 6191enum { 6192 BPF_F_BROADCAST = (1ULL << 3), 6193 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 6194}; 6195 6196#define __bpf_md_ptr(type, name) \ 6197union { \ 6198 type name; \ 6199 __u64 :64; \ 6200} __attribute__((aligned(8))) 6201 6202enum { 6203 BPF_SKB_TSTAMP_UNSPEC, 6204 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 6205 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 6206 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 6207 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 6208 */ 6209}; 6210 6211/* user accessible mirror of in-kernel sk_buff. 6212 * new fields can only be added to the end of this structure 6213 */ 6214struct __sk_buff { 6215 __u32 len; 6216 __u32 pkt_type; 6217 __u32 mark; 6218 __u32 queue_mapping; 6219 __u32 protocol; 6220 __u32 vlan_present; 6221 __u32 vlan_tci; 6222 __u32 vlan_proto; 6223 __u32 priority; 6224 __u32 ingress_ifindex; 6225 __u32 ifindex; 6226 __u32 tc_index; 6227 __u32 cb[5]; 6228 __u32 hash; 6229 __u32 tc_classid; 6230 __u32 data; 6231 __u32 data_end; 6232 __u32 napi_id; 6233 6234 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6235 __u32 family; 6236 __u32 remote_ip4; /* Stored in network byte order */ 6237 __u32 local_ip4; /* Stored in network byte order */ 6238 __u32 remote_ip6[4]; /* Stored in network byte order */ 6239 __u32 local_ip6[4]; /* Stored in network byte order */ 6240 __u32 remote_port; /* Stored in network byte order */ 6241 __u32 local_port; /* stored in host byte order */ 6242 /* ... here. */ 6243 6244 __u32 data_meta; 6245 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6246 __u64 tstamp; 6247 __u32 wire_len; 6248 __u32 gso_segs; 6249 __bpf_md_ptr(struct bpf_sock *, sk); 6250 __u32 gso_size; 6251 __u8 tstamp_type; 6252 __u32 :24; /* Padding, future use. */ 6253 __u64 hwtstamp; 6254}; 6255 6256struct bpf_tunnel_key { 6257 __u32 tunnel_id; 6258 union { 6259 __u32 remote_ipv4; 6260 __u32 remote_ipv6[4]; 6261 }; 6262 __u8 tunnel_tos; 6263 __u8 tunnel_ttl; 6264 union { 6265 __u16 tunnel_ext; /* compat */ 6266 __be16 tunnel_flags; 6267 }; 6268 __u32 tunnel_label; 6269 union { 6270 __u32 local_ipv4; 6271 __u32 local_ipv6[4]; 6272 }; 6273}; 6274 6275/* user accessible mirror of in-kernel xfrm_state. 6276 * new fields can only be added to the end of this structure 6277 */ 6278struct bpf_xfrm_state { 6279 __u32 reqid; 6280 __u32 spi; /* Stored in network byte order */ 6281 __u16 family; 6282 __u16 ext; /* Padding, future use. */ 6283 union { 6284 __u32 remote_ipv4; /* Stored in network byte order */ 6285 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6286 }; 6287}; 6288 6289/* Generic BPF return codes which all BPF program types may support. 6290 * The values are binary compatible with their TC_ACT_* counter-part to 6291 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6292 * programs. 6293 * 6294 * XDP is handled seprately, see XDP_*. 6295 */ 6296enum bpf_ret_code { 6297 BPF_OK = 0, 6298 /* 1 reserved */ 6299 BPF_DROP = 2, 6300 /* 3-6 reserved */ 6301 BPF_REDIRECT = 7, 6302 /* >127 are reserved for prog type specific return codes. 6303 * 6304 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6305 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6306 * changed and should be routed based on its new L3 header. 6307 * (This is an L3 redirect, as opposed to L2 redirect 6308 * represented by BPF_REDIRECT above). 6309 */ 6310 BPF_LWT_REROUTE = 128, 6311 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6312 * to indicate that no custom dissection was performed, and 6313 * fallback to standard dissector is requested. 6314 */ 6315 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6316}; 6317 6318struct bpf_sock { 6319 __u32 bound_dev_if; 6320 __u32 family; 6321 __u32 type; 6322 __u32 protocol; 6323 __u32 mark; 6324 __u32 priority; 6325 /* IP address also allows 1 and 2 bytes access */ 6326 __u32 src_ip4; 6327 __u32 src_ip6[4]; 6328 __u32 src_port; /* host byte order */ 6329 __be16 dst_port; /* network byte order */ 6330 __u16 :16; /* zero padding */ 6331 __u32 dst_ip4; 6332 __u32 dst_ip6[4]; 6333 __u32 state; 6334 __s32 rx_queue_mapping; 6335}; 6336 6337struct bpf_tcp_sock { 6338 __u32 snd_cwnd; /* Sending congestion window */ 6339 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6340 __u32 rtt_min; 6341 __u32 snd_ssthresh; /* Slow start size threshold */ 6342 __u32 rcv_nxt; /* What we want to receive next */ 6343 __u32 snd_nxt; /* Next sequence we send */ 6344 __u32 snd_una; /* First byte we want an ack for */ 6345 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6346 __u32 ecn_flags; /* ECN status bits. */ 6347 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6348 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6349 __u32 packets_out; /* Packets which are "in flight" */ 6350 __u32 retrans_out; /* Retransmitted packets out */ 6351 __u32 total_retrans; /* Total retransmits for entire connection */ 6352 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6353 * total number of segments in. 6354 */ 6355 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6356 * total number of data segments in. 6357 */ 6358 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6359 * The total number of segments sent. 6360 */ 6361 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6362 * total number of data segments sent. 6363 */ 6364 __u32 lost_out; /* Lost packets */ 6365 __u32 sacked_out; /* SACK'd packets */ 6366 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6367 * sum(delta(rcv_nxt)), or how many bytes 6368 * were acked. 6369 */ 6370 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6371 * sum(delta(snd_una)), or how many bytes 6372 * were acked. 6373 */ 6374 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6375 * total number of DSACK blocks received 6376 */ 6377 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6378 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6379 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6380}; 6381 6382struct bpf_sock_tuple { 6383 union { 6384 struct { 6385 __be32 saddr; 6386 __be32 daddr; 6387 __be16 sport; 6388 __be16 dport; 6389 } ipv4; 6390 struct { 6391 __be32 saddr[4]; 6392 __be32 daddr[4]; 6393 __be16 sport; 6394 __be16 dport; 6395 } ipv6; 6396 }; 6397}; 6398 6399/* (Simplified) user return codes for tcx prog type. 6400 * A valid tcx program must return one of these defined values. All other 6401 * return codes are reserved for future use. Must remain compatible with 6402 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6403 * return codes are mapped to TCX_NEXT. 6404 */ 6405enum tcx_action_base { 6406 TCX_NEXT = -1, 6407 TCX_PASS = 0, 6408 TCX_DROP = 2, 6409 TCX_REDIRECT = 7, 6410}; 6411 6412struct bpf_xdp_sock { 6413 __u32 queue_id; 6414}; 6415 6416#define XDP_PACKET_HEADROOM 256 6417 6418/* User return codes for XDP prog type. 6419 * A valid XDP program must return one of these defined values. All other 6420 * return codes are reserved for future use. Unknown return codes will 6421 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6422 */ 6423enum xdp_action { 6424 XDP_ABORTED = 0, 6425 XDP_DROP, 6426 XDP_PASS, 6427 XDP_TX, 6428 XDP_REDIRECT, 6429}; 6430 6431/* user accessible metadata for XDP packet hook 6432 * new fields must be added to the end of this structure 6433 */ 6434struct xdp_md { 6435 __u32 data; 6436 __u32 data_end; 6437 __u32 data_meta; 6438 /* Below access go through struct xdp_rxq_info */ 6439 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6440 __u32 rx_queue_index; /* rxq->queue_index */ 6441 6442 __u32 egress_ifindex; /* txq->dev->ifindex */ 6443}; 6444 6445/* DEVMAP map-value layout 6446 * 6447 * The struct data-layout of map-value is a configuration interface. 6448 * New members can only be added to the end of this structure. 6449 */ 6450struct bpf_devmap_val { 6451 __u32 ifindex; /* device index */ 6452 union { 6453 int fd; /* prog fd on map write */ 6454 __u32 id; /* prog id on map read */ 6455 } bpf_prog; 6456}; 6457 6458/* CPUMAP map-value layout 6459 * 6460 * The struct data-layout of map-value is a configuration interface. 6461 * New members can only be added to the end of this structure. 6462 */ 6463struct bpf_cpumap_val { 6464 __u32 qsize; /* queue size to remote target CPU */ 6465 union { 6466 int fd; /* prog fd on map write */ 6467 __u32 id; /* prog id on map read */ 6468 } bpf_prog; 6469}; 6470 6471enum sk_action { 6472 SK_DROP = 0, 6473 SK_PASS, 6474}; 6475 6476/* user accessible metadata for SK_MSG packet hook, new fields must 6477 * be added to the end of this structure 6478 */ 6479struct sk_msg_md { 6480 __bpf_md_ptr(void *, data); 6481 __bpf_md_ptr(void *, data_end); 6482 6483 __u32 family; 6484 __u32 remote_ip4; /* Stored in network byte order */ 6485 __u32 local_ip4; /* Stored in network byte order */ 6486 __u32 remote_ip6[4]; /* Stored in network byte order */ 6487 __u32 local_ip6[4]; /* Stored in network byte order */ 6488 __u32 remote_port; /* Stored in network byte order */ 6489 __u32 local_port; /* stored in host byte order */ 6490 __u32 size; /* Total size of sk_msg */ 6491 6492 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6493}; 6494 6495struct sk_reuseport_md { 6496 /* 6497 * Start of directly accessible data. It begins from 6498 * the tcp/udp header. 6499 */ 6500 __bpf_md_ptr(void *, data); 6501 /* End of directly accessible data */ 6502 __bpf_md_ptr(void *, data_end); 6503 /* 6504 * Total length of packet (starting from the tcp/udp header). 6505 * Note that the directly accessible bytes (data_end - data) 6506 * could be less than this "len". Those bytes could be 6507 * indirectly read by a helper "bpf_skb_load_bytes()". 6508 */ 6509 __u32 len; 6510 /* 6511 * Eth protocol in the mac header (network byte order). e.g. 6512 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6513 */ 6514 __u32 eth_protocol; 6515 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6516 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6517 __u32 hash; /* A hash of the packet 4 tuples */ 6518 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6519 * new incoming connection request (e.g. selecting a listen sk for 6520 * the received SYN in the TCP case). reuse->sk is one of the sk 6521 * in the reuseport group. The bpf prog can use reuse->sk to learn 6522 * the local listening ip/port without looking into the skb. 6523 * 6524 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6525 * reuse->migrating_sk is the socket that needs to be migrated 6526 * to another listening socket. migrating_sk could be a fullsock 6527 * sk that is fully established or a reqsk that is in-the-middle 6528 * of 3-way handshake. 6529 */ 6530 __bpf_md_ptr(struct bpf_sock *, sk); 6531 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6532}; 6533 6534#define BPF_TAG_SIZE 8 6535 6536struct bpf_prog_info { 6537 __u32 type; 6538 __u32 id; 6539 __u8 tag[BPF_TAG_SIZE]; 6540 __u32 jited_prog_len; 6541 __u32 xlated_prog_len; 6542 __aligned_u64 jited_prog_insns; 6543 __aligned_u64 xlated_prog_insns; 6544 __u64 load_time; /* ns since boottime */ 6545 __u32 created_by_uid; 6546 __u32 nr_map_ids; 6547 __aligned_u64 map_ids; 6548 char name[BPF_OBJ_NAME_LEN]; 6549 __u32 ifindex; 6550 __u32 gpl_compatible:1; 6551 __u32 :31; /* alignment pad */ 6552 __u64 netns_dev; 6553 __u64 netns_ino; 6554 __u32 nr_jited_ksyms; 6555 __u32 nr_jited_func_lens; 6556 __aligned_u64 jited_ksyms; 6557 __aligned_u64 jited_func_lens; 6558 __u32 btf_id; 6559 __u32 func_info_rec_size; 6560 __aligned_u64 func_info; 6561 __u32 nr_func_info; 6562 __u32 nr_line_info; 6563 __aligned_u64 line_info; 6564 __aligned_u64 jited_line_info; 6565 __u32 nr_jited_line_info; 6566 __u32 line_info_rec_size; 6567 __u32 jited_line_info_rec_size; 6568 __u32 nr_prog_tags; 6569 __aligned_u64 prog_tags; 6570 __u64 run_time_ns; 6571 __u64 run_cnt; 6572 __u64 recursion_misses; 6573 __u32 verified_insns; 6574 __u32 attach_btf_obj_id; 6575 __u32 attach_btf_id; 6576} __attribute__((aligned(8))); 6577 6578struct bpf_map_info { 6579 __u32 type; 6580 __u32 id; 6581 __u32 key_size; 6582 __u32 value_size; 6583 __u32 max_entries; 6584 __u32 map_flags; 6585 char name[BPF_OBJ_NAME_LEN]; 6586 __u32 ifindex; 6587 __u32 btf_vmlinux_value_type_id; 6588 __u64 netns_dev; 6589 __u64 netns_ino; 6590 __u32 btf_id; 6591 __u32 btf_key_type_id; 6592 __u32 btf_value_type_id; 6593 __u32 btf_vmlinux_id; 6594 __u64 map_extra; 6595} __attribute__((aligned(8))); 6596 6597struct bpf_btf_info { 6598 __aligned_u64 btf; 6599 __u32 btf_size; 6600 __u32 id; 6601 __aligned_u64 name; 6602 __u32 name_len; 6603 __u32 kernel_btf; 6604} __attribute__((aligned(8))); 6605 6606struct bpf_link_info { 6607 __u32 type; 6608 __u32 id; 6609 __u32 prog_id; 6610 union { 6611 struct { 6612 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6613 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6614 } raw_tracepoint; 6615 struct { 6616 __u32 attach_type; 6617 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6618 __u32 target_btf_id; /* BTF type id inside the object */ 6619 } tracing; 6620 struct { 6621 __u64 cgroup_id; 6622 __u32 attach_type; 6623 } cgroup; 6624 struct { 6625 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6626 __u32 target_name_len; /* in/out: target_name buffer len */ 6627 6628 /* If the iter specific field is 32 bits, it can be put 6629 * in the first or second union. Otherwise it should be 6630 * put in the second union. 6631 */ 6632 union { 6633 struct { 6634 __u32 map_id; 6635 } map; 6636 }; 6637 union { 6638 struct { 6639 __u64 cgroup_id; 6640 __u32 order; 6641 } cgroup; 6642 struct { 6643 __u32 tid; 6644 __u32 pid; 6645 } task; 6646 }; 6647 } iter; 6648 struct { 6649 __u32 netns_ino; 6650 __u32 attach_type; 6651 } netns; 6652 struct { 6653 __u32 ifindex; 6654 } xdp; 6655 struct { 6656 __u32 map_id; 6657 } struct_ops; 6658 struct { 6659 __u32 pf; 6660 __u32 hooknum; 6661 __s32 priority; 6662 __u32 flags; 6663 } netfilter; 6664 struct { 6665 __aligned_u64 addrs; 6666 __u32 count; /* in/out: kprobe_multi function count */ 6667 __u32 flags; 6668 __u64 missed; 6669 __aligned_u64 cookies; 6670 } kprobe_multi; 6671 struct { 6672 __aligned_u64 path; 6673 __aligned_u64 offsets; 6674 __aligned_u64 ref_ctr_offsets; 6675 __aligned_u64 cookies; 6676 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6677 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6678 __u32 flags; 6679 __u32 pid; 6680 } uprobe_multi; 6681 struct { 6682 __u32 type; /* enum bpf_perf_event_type */ 6683 __u32 :32; 6684 union { 6685 struct { 6686 __aligned_u64 file_name; /* in/out */ 6687 __u32 name_len; 6688 __u32 offset; /* offset from file_name */ 6689 __u64 cookie; 6690 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6691 struct { 6692 __aligned_u64 func_name; /* in/out */ 6693 __u32 name_len; 6694 __u32 offset; /* offset from func_name */ 6695 __u64 addr; 6696 __u64 missed; 6697 __u64 cookie; 6698 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6699 struct { 6700 __aligned_u64 tp_name; /* in/out */ 6701 __u32 name_len; 6702 __u32 :32; 6703 __u64 cookie; 6704 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6705 struct { 6706 __u64 config; 6707 __u32 type; 6708 __u32 :32; 6709 __u64 cookie; 6710 } event; /* BPF_PERF_EVENT_EVENT */ 6711 }; 6712 } perf_event; 6713 struct { 6714 __u32 ifindex; 6715 __u32 attach_type; 6716 } tcx; 6717 struct { 6718 __u32 ifindex; 6719 __u32 attach_type; 6720 } netkit; 6721 }; 6722} __attribute__((aligned(8))); 6723 6724/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6725 * by user and intended to be used by socket (e.g. to bind to, depends on 6726 * attach type). 6727 */ 6728struct bpf_sock_addr { 6729 __u32 user_family; /* Allows 4-byte read, but no write. */ 6730 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6731 * Stored in network byte order. 6732 */ 6733 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6734 * Stored in network byte order. 6735 */ 6736 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6737 * Stored in network byte order 6738 */ 6739 __u32 family; /* Allows 4-byte read, but no write */ 6740 __u32 type; /* Allows 4-byte read, but no write */ 6741 __u32 protocol; /* Allows 4-byte read, but no write */ 6742 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6743 * Stored in network byte order. 6744 */ 6745 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6746 * Stored in network byte order. 6747 */ 6748 __bpf_md_ptr(struct bpf_sock *, sk); 6749}; 6750 6751/* User bpf_sock_ops struct to access socket values and specify request ops 6752 * and their replies. 6753 * Some of this fields are in network (bigendian) byte order and may need 6754 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6755 * New fields can only be added at the end of this structure 6756 */ 6757struct bpf_sock_ops { 6758 __u32 op; 6759 union { 6760 __u32 args[4]; /* Optionally passed to bpf program */ 6761 __u32 reply; /* Returned by bpf program */ 6762 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6763 }; 6764 __u32 family; 6765 __u32 remote_ip4; /* Stored in network byte order */ 6766 __u32 local_ip4; /* Stored in network byte order */ 6767 __u32 remote_ip6[4]; /* Stored in network byte order */ 6768 __u32 local_ip6[4]; /* Stored in network byte order */ 6769 __u32 remote_port; /* Stored in network byte order */ 6770 __u32 local_port; /* stored in host byte order */ 6771 __u32 is_fullsock; /* Some TCP fields are only valid if 6772 * there is a full socket. If not, the 6773 * fields read as zero. 6774 */ 6775 __u32 snd_cwnd; 6776 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6777 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6778 __u32 state; 6779 __u32 rtt_min; 6780 __u32 snd_ssthresh; 6781 __u32 rcv_nxt; 6782 __u32 snd_nxt; 6783 __u32 snd_una; 6784 __u32 mss_cache; 6785 __u32 ecn_flags; 6786 __u32 rate_delivered; 6787 __u32 rate_interval_us; 6788 __u32 packets_out; 6789 __u32 retrans_out; 6790 __u32 total_retrans; 6791 __u32 segs_in; 6792 __u32 data_segs_in; 6793 __u32 segs_out; 6794 __u32 data_segs_out; 6795 __u32 lost_out; 6796 __u32 sacked_out; 6797 __u32 sk_txhash; 6798 __u64 bytes_received; 6799 __u64 bytes_acked; 6800 __bpf_md_ptr(struct bpf_sock *, sk); 6801 /* [skb_data, skb_data_end) covers the whole TCP header. 6802 * 6803 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6804 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6805 * header has not been written. 6806 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6807 * been written so far. 6808 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6809 * the 3WHS. 6810 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6811 * the 3WHS. 6812 * 6813 * bpf_load_hdr_opt() can also be used to read a particular option. 6814 */ 6815 __bpf_md_ptr(void *, skb_data); 6816 __bpf_md_ptr(void *, skb_data_end); 6817 __u32 skb_len; /* The total length of a packet. 6818 * It includes the header, options, 6819 * and payload. 6820 */ 6821 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6822 * an easy way to check for tcp_flags 6823 * without parsing skb_data. 6824 * 6825 * In particular, the skb_tcp_flags 6826 * will still be available in 6827 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6828 * the outgoing header has not 6829 * been written yet. 6830 */ 6831 __u64 skb_hwtstamp; 6832}; 6833 6834/* Definitions for bpf_sock_ops_cb_flags */ 6835enum { 6836 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6837 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6838 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6839 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6840 /* Call bpf for all received TCP headers. The bpf prog will be 6841 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6842 * 6843 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6844 * for the header option related helpers that will be useful 6845 * to the bpf programs. 6846 * 6847 * It could be used at the client/active side (i.e. connect() side) 6848 * when the server told it that the server was in syncookie 6849 * mode and required the active side to resend the bpf-written 6850 * options. The active side can keep writing the bpf-options until 6851 * it received a valid packet from the server side to confirm 6852 * the earlier packet (and options) has been received. The later 6853 * example patch is using it like this at the active side when the 6854 * server is in syncookie mode. 6855 * 6856 * The bpf prog will usually turn this off in the common cases. 6857 */ 6858 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6859 /* Call bpf when kernel has received a header option that 6860 * the kernel cannot handle. The bpf prog will be called under 6861 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6862 * 6863 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6864 * for the header option related helpers that will be useful 6865 * to the bpf programs. 6866 */ 6867 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6868 /* Call bpf when the kernel is writing header options for the 6869 * outgoing packet. The bpf prog will first be called 6870 * to reserve space in a skb under 6871 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6872 * the bpf prog will be called to write the header option(s) 6873 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6874 * 6875 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6876 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6877 * related helpers that will be useful to the bpf programs. 6878 * 6879 * The kernel gets its chance to reserve space and write 6880 * options first before the BPF program does. 6881 */ 6882 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6883/* Mask of all currently supported cb flags */ 6884 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6885}; 6886 6887/* List of known BPF sock_ops operators. 6888 * New entries can only be added at the end 6889 */ 6890enum { 6891 BPF_SOCK_OPS_VOID, 6892 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6893 * -1 if default value should be used 6894 */ 6895 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6896 * window (in packets) or -1 if default 6897 * value should be used 6898 */ 6899 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6900 * active connection is initialized 6901 */ 6902 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6903 * active connection is 6904 * established 6905 */ 6906 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6907 * passive connection is 6908 * established 6909 */ 6910 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6911 * needs ECN 6912 */ 6913 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6914 * based on the path and may be 6915 * dependent on the congestion control 6916 * algorithm. In general it indicates 6917 * a congestion threshold. RTTs above 6918 * this indicate congestion 6919 */ 6920 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6921 * Arg1: value of icsk_retransmits 6922 * Arg2: value of icsk_rto 6923 * Arg3: whether RTO has expired 6924 */ 6925 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6926 * Arg1: sequence number of 1st byte 6927 * Arg2: # segments 6928 * Arg3: return value of 6929 * tcp_transmit_skb (0 => success) 6930 */ 6931 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6932 * Arg1: old_state 6933 * Arg2: new_state 6934 */ 6935 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6936 * socket transition to LISTEN state. 6937 */ 6938 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6939 */ 6940 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6941 * It will be called to handle 6942 * the packets received at 6943 * an already established 6944 * connection. 6945 * 6946 * sock_ops->skb_data: 6947 * Referring to the received skb. 6948 * It covers the TCP header only. 6949 * 6950 * bpf_load_hdr_opt() can also 6951 * be used to search for a 6952 * particular option. 6953 */ 6954 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6955 * header option later in 6956 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6957 * Arg1: bool want_cookie. (in 6958 * writing SYNACK only) 6959 * 6960 * sock_ops->skb_data: 6961 * Not available because no header has 6962 * been written yet. 6963 * 6964 * sock_ops->skb_tcp_flags: 6965 * The tcp_flags of the 6966 * outgoing skb. (e.g. SYN, ACK, FIN). 6967 * 6968 * bpf_reserve_hdr_opt() should 6969 * be used to reserve space. 6970 */ 6971 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6972 * Arg1: bool want_cookie. (in 6973 * writing SYNACK only) 6974 * 6975 * sock_ops->skb_data: 6976 * Referring to the outgoing skb. 6977 * It covers the TCP header 6978 * that has already been written 6979 * by the kernel and the 6980 * earlier bpf-progs. 6981 * 6982 * sock_ops->skb_tcp_flags: 6983 * The tcp_flags of the outgoing 6984 * skb. (e.g. SYN, ACK, FIN). 6985 * 6986 * bpf_store_hdr_opt() should 6987 * be used to write the 6988 * option. 6989 * 6990 * bpf_load_hdr_opt() can also 6991 * be used to search for a 6992 * particular option that 6993 * has already been written 6994 * by the kernel or the 6995 * earlier bpf-progs. 6996 */ 6997}; 6998 6999/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 7000 * changes between the TCP and BPF versions. Ideally this should never happen. 7001 * If it does, we need to add code to convert them before calling 7002 * the BPF sock_ops function. 7003 */ 7004enum { 7005 BPF_TCP_ESTABLISHED = 1, 7006 BPF_TCP_SYN_SENT, 7007 BPF_TCP_SYN_RECV, 7008 BPF_TCP_FIN_WAIT1, 7009 BPF_TCP_FIN_WAIT2, 7010 BPF_TCP_TIME_WAIT, 7011 BPF_TCP_CLOSE, 7012 BPF_TCP_CLOSE_WAIT, 7013 BPF_TCP_LAST_ACK, 7014 BPF_TCP_LISTEN, 7015 BPF_TCP_CLOSING, /* Now a valid state */ 7016 BPF_TCP_NEW_SYN_RECV, 7017 BPF_TCP_BOUND_INACTIVE, 7018 7019 BPF_TCP_MAX_STATES /* Leave at the end! */ 7020}; 7021 7022enum { 7023 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 7024 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 7025 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 7026 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 7027 /* Copy the SYN pkt to optval 7028 * 7029 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 7030 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 7031 * to only getting from the saved_syn. It can either get the 7032 * syn packet from: 7033 * 7034 * 1. the just-received SYN packet (only available when writing the 7035 * SYNACK). It will be useful when it is not necessary to 7036 * save the SYN packet for latter use. It is also the only way 7037 * to get the SYN during syncookie mode because the syn 7038 * packet cannot be saved during syncookie. 7039 * 7040 * OR 7041 * 7042 * 2. the earlier saved syn which was done by 7043 * bpf_setsockopt(TCP_SAVE_SYN). 7044 * 7045 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7046 * SYN packet is obtained. 7047 * 7048 * If the bpf-prog does not need the IP[46] header, the 7049 * bpf-prog can avoid parsing the IP header by using 7050 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7051 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7052 * 7053 * >0: Total number of bytes copied 7054 * -ENOSPC: Not enough space in optval. Only optlen number of 7055 * bytes is copied. 7056 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7057 * is not saved by setsockopt(TCP_SAVE_SYN). 7058 */ 7059 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7060 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7061 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7062}; 7063 7064enum { 7065 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7066}; 7067 7068/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7069 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7070 */ 7071enum { 7072 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7073 * total option spaces 7074 * required for an established 7075 * sk in order to calculate the 7076 * MSS. No skb is actually 7077 * sent. 7078 */ 7079 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7080 * when sending a SYN. 7081 */ 7082}; 7083 7084struct bpf_perf_event_value { 7085 __u64 counter; 7086 __u64 enabled; 7087 __u64 running; 7088}; 7089 7090enum { 7091 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7092 BPF_DEVCG_ACC_READ = (1ULL << 1), 7093 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7094}; 7095 7096enum { 7097 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7098 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7099}; 7100 7101struct bpf_cgroup_dev_ctx { 7102 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7103 __u32 access_type; 7104 __u32 major; 7105 __u32 minor; 7106}; 7107 7108struct bpf_raw_tracepoint_args { 7109 __u64 args[0]; 7110}; 7111 7112/* DIRECT: Skip the FIB rules and go to FIB table associated with device 7113 * OUTPUT: Do lookup from egress perspective; default is ingress 7114 */ 7115enum { 7116 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7117 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7118 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7119 BPF_FIB_LOOKUP_TBID = (1U << 3), 7120 BPF_FIB_LOOKUP_SRC = (1U << 4), 7121}; 7122 7123enum { 7124 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7125 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7126 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7127 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7128 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7129 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7130 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7131 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7132 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7133 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7134}; 7135 7136struct bpf_fib_lookup { 7137 /* input: network family for lookup (AF_INET, AF_INET6) 7138 * output: network family of egress nexthop 7139 */ 7140 __u8 family; 7141 7142 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7143 __u8 l4_protocol; 7144 __be16 sport; 7145 __be16 dport; 7146 7147 union { /* used for MTU check */ 7148 /* input to lookup */ 7149 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7150 7151 /* output: MTU value */ 7152 __u16 mtu_result; 7153 }; 7154 /* input: L3 device index for lookup 7155 * output: device index from FIB lookup 7156 */ 7157 __u32 ifindex; 7158 7159 union { 7160 /* inputs to lookup */ 7161 __u8 tos; /* AF_INET */ 7162 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7163 7164 /* output: metric of fib result (IPv4/IPv6 only) */ 7165 __u32 rt_metric; 7166 }; 7167 7168 /* input: source address to consider for lookup 7169 * output: source address result from lookup 7170 */ 7171 union { 7172 __be32 ipv4_src; 7173 __u32 ipv6_src[4]; /* in6_addr; network order */ 7174 }; 7175 7176 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7177 * network header. output: bpf_fib_lookup sets to gateway address 7178 * if FIB lookup returns gateway route 7179 */ 7180 union { 7181 __be32 ipv4_dst; 7182 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7183 }; 7184 7185 union { 7186 struct { 7187 /* output */ 7188 __be16 h_vlan_proto; 7189 __be16 h_vlan_TCI; 7190 }; 7191 /* input: when accompanied with the 7192 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7193 * specific routing table to use for the fib lookup. 7194 */ 7195 __u32 tbid; 7196 }; 7197 7198 __u8 smac[6]; /* ETH_ALEN */ 7199 __u8 dmac[6]; /* ETH_ALEN */ 7200}; 7201 7202struct bpf_redir_neigh { 7203 /* network family for lookup (AF_INET, AF_INET6) */ 7204 __u32 nh_family; 7205 /* network address of nexthop; skips fib lookup to find gateway */ 7206 union { 7207 __be32 ipv4_nh; 7208 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7209 }; 7210}; 7211 7212/* bpf_check_mtu flags*/ 7213enum bpf_check_mtu_flags { 7214 BPF_MTU_CHK_SEGS = (1U << 0), 7215}; 7216 7217enum bpf_check_mtu_ret { 7218 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7219 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7220 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7221}; 7222 7223enum bpf_task_fd_type { 7224 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7225 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7226 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7227 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7228 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7229 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7230}; 7231 7232enum { 7233 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7234 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7235 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7236}; 7237 7238struct bpf_flow_keys { 7239 __u16 nhoff; 7240 __u16 thoff; 7241 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7242 __u8 is_frag; 7243 __u8 is_first_frag; 7244 __u8 is_encap; 7245 __u8 ip_proto; 7246 __be16 n_proto; 7247 __be16 sport; 7248 __be16 dport; 7249 union { 7250 struct { 7251 __be32 ipv4_src; 7252 __be32 ipv4_dst; 7253 }; 7254 struct { 7255 __u32 ipv6_src[4]; /* in6_addr; network order */ 7256 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7257 }; 7258 }; 7259 __u32 flags; 7260 __be32 flow_label; 7261}; 7262 7263struct bpf_func_info { 7264 __u32 insn_off; 7265 __u32 type_id; 7266}; 7267 7268#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7269#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7270 7271struct bpf_line_info { 7272 __u32 insn_off; 7273 __u32 file_name_off; 7274 __u32 line_off; 7275 __u32 line_col; 7276}; 7277 7278struct bpf_spin_lock { 7279 __u32 val; 7280}; 7281 7282struct bpf_timer { 7283 __u64 __opaque[2]; 7284} __attribute__((aligned(8))); 7285 7286struct bpf_dynptr { 7287 __u64 __opaque[2]; 7288} __attribute__((aligned(8))); 7289 7290struct bpf_list_head { 7291 __u64 __opaque[2]; 7292} __attribute__((aligned(8))); 7293 7294struct bpf_list_node { 7295 __u64 __opaque[3]; 7296} __attribute__((aligned(8))); 7297 7298struct bpf_rb_root { 7299 __u64 __opaque[2]; 7300} __attribute__((aligned(8))); 7301 7302struct bpf_rb_node { 7303 __u64 __opaque[4]; 7304} __attribute__((aligned(8))); 7305 7306struct bpf_refcount { 7307 __u32 __opaque[1]; 7308} __attribute__((aligned(4))); 7309 7310struct bpf_sysctl { 7311 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7312 * Allows 1,2,4-byte read, but no write. 7313 */ 7314 __u32 file_pos; /* Sysctl file position to read from, write to. 7315 * Allows 1,2,4-byte read an 4-byte write. 7316 */ 7317}; 7318 7319struct bpf_sockopt { 7320 __bpf_md_ptr(struct bpf_sock *, sk); 7321 __bpf_md_ptr(void *, optval); 7322 __bpf_md_ptr(void *, optval_end); 7323 7324 __s32 level; 7325 __s32 optname; 7326 __s32 optlen; 7327 __s32 retval; 7328}; 7329 7330struct bpf_pidns_info { 7331 __u32 pid; 7332 __u32 tgid; 7333}; 7334 7335/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7336struct bpf_sk_lookup { 7337 union { 7338 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7339 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7340 }; 7341 7342 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7343 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7344 __u32 remote_ip4; /* Network byte order */ 7345 __u32 remote_ip6[4]; /* Network byte order */ 7346 __be16 remote_port; /* Network byte order */ 7347 __u16 :16; /* Zero padding */ 7348 __u32 local_ip4; /* Network byte order */ 7349 __u32 local_ip6[4]; /* Network byte order */ 7350 __u32 local_port; /* Host byte order */ 7351 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7352}; 7353 7354/* 7355 * struct btf_ptr is used for typed pointer representation; the 7356 * type id is used to render the pointer data as the appropriate type 7357 * via the bpf_snprintf_btf() helper described above. A flags field - 7358 * potentially to specify additional details about the BTF pointer 7359 * (rather than its mode of display) - is included for future use. 7360 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7361 */ 7362struct btf_ptr { 7363 void *ptr; 7364 __u32 type_id; 7365 __u32 flags; /* BTF ptr flags; unused at present. */ 7366}; 7367 7368/* 7369 * Flags to control bpf_snprintf_btf() behaviour. 7370 * - BTF_F_COMPACT: no formatting around type information 7371 * - BTF_F_NONAME: no struct/union member names/types 7372 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7373 * equivalent to %px. 7374 * - BTF_F_ZERO: show zero-valued struct/union members; they 7375 * are not displayed by default 7376 */ 7377enum { 7378 BTF_F_COMPACT = (1ULL << 0), 7379 BTF_F_NONAME = (1ULL << 1), 7380 BTF_F_PTR_RAW = (1ULL << 2), 7381 BTF_F_ZERO = (1ULL << 3), 7382}; 7383 7384/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7385 * has to be adjusted by relocations. It is emitted by llvm and passed to 7386 * libbpf and later to the kernel. 7387 */ 7388enum bpf_core_relo_kind { 7389 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7390 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7391 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7392 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7393 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7394 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7395 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7396 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7397 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7398 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7399 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7400 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7401 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7402}; 7403 7404/* 7405 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7406 * and from libbpf to the kernel. 7407 * 7408 * CO-RE relocation captures the following data: 7409 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7410 * its insn->imm field to be relocated with actual field info; 7411 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7412 * type or field; 7413 * - access_str_off - offset into corresponding .BTF string section. String 7414 * interpretation depends on specific relocation kind: 7415 * - for field-based relocations, string encodes an accessed field using 7416 * a sequence of field and array indices, separated by colon (:). It's 7417 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7418 * arguments for identifying offset to a field. 7419 * - for type-based relocations, strings is expected to be just "0"; 7420 * - for enum value-based relocations, string contains an index of enum 7421 * value within its enum type; 7422 * - kind - one of enum bpf_core_relo_kind; 7423 * 7424 * Example: 7425 * struct sample { 7426 * int a; 7427 * struct { 7428 * int b[10]; 7429 * }; 7430 * }; 7431 * 7432 * struct sample *s = ...; 7433 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7434 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7435 * // b is field #0 inside anon struct, accessing elem #5) 7436 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7437 * 7438 * type_id for all relocs in this example will capture BTF type id of 7439 * `struct sample`. 7440 * 7441 * Such relocation is emitted when using __builtin_preserve_access_index() 7442 * Clang built-in, passing expression that captures field address, e.g.: 7443 * 7444 * bpf_probe_read(&dst, sizeof(dst), 7445 * __builtin_preserve_access_index(&src->a.b.c)); 7446 * 7447 * In this case Clang will emit field relocation recording necessary data to 7448 * be able to find offset of embedded `a.b.c` field within `src` struct. 7449 * 7450 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7451 */ 7452struct bpf_core_relo { 7453 __u32 insn_off; 7454 __u32 type_id; 7455 __u32 access_str_off; 7456 enum bpf_core_relo_kind kind; 7457}; 7458 7459/* 7460 * Flags to control bpf_timer_start() behaviour. 7461 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7462 * relative to current time. 7463 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7464 */ 7465enum { 7466 BPF_F_TIMER_ABS = (1ULL << 0), 7467 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7468}; 7469 7470/* BPF numbers iterator state */ 7471struct bpf_iter_num { 7472 /* opaque iterator state; having __u64 here allows to preserve correct 7473 * alignment requirements in vmlinux.h, generated from BTF 7474 */ 7475 __u64 __opaque[1]; 7476} __attribute__((aligned(8))); 7477 7478#endif /* _UAPI__LINUX_BPF_H__ */ 7479