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