at v5.17 242 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_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 23#define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24 25/* alu/jmp fields */ 26#define BPF_MOV 0xb0 /* mov reg to reg */ 27#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28 29/* change endianness of a register */ 30#define BPF_END 0xd0 /* flags for endianness conversion: */ 31#define BPF_TO_LE 0x00 /* convert to little-endian */ 32#define BPF_TO_BE 0x08 /* convert to big-endian */ 33#define BPF_FROM_LE BPF_TO_LE 34#define BPF_FROM_BE BPF_TO_BE 35 36/* jmp encodings */ 37#define BPF_JNE 0x50 /* jump != */ 38#define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 39#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 42#define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 43#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44#define BPF_CALL 0x80 /* function call */ 45#define BPF_EXIT 0x90 /* function return */ 46 47/* atomic op type fields (stored in immediate) */ 48#define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 49#define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 50#define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 51 52/* Register numbers */ 53enum { 54 BPF_REG_0 = 0, 55 BPF_REG_1, 56 BPF_REG_2, 57 BPF_REG_3, 58 BPF_REG_4, 59 BPF_REG_5, 60 BPF_REG_6, 61 BPF_REG_7, 62 BPF_REG_8, 63 BPF_REG_9, 64 BPF_REG_10, 65 __MAX_BPF_REG, 66}; 67 68/* BPF has 10 general purpose 64-bit registers and stack frame. */ 69#define MAX_BPF_REG __MAX_BPF_REG 70 71struct bpf_insn { 72 __u8 code; /* opcode */ 73 __u8 dst_reg:4; /* dest register */ 74 __u8 src_reg:4; /* source register */ 75 __s16 off; /* signed offset */ 76 __s32 imm; /* signed immediate constant */ 77}; 78 79/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80struct bpf_lpm_trie_key { 81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82 __u8 data[0]; /* Arbitrary size */ 83}; 84 85struct bpf_cgroup_storage_key { 86 __u64 cgroup_inode_id; /* cgroup inode id */ 87 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 88}; 89 90union bpf_iter_link_info { 91 struct { 92 __u32 map_fd; 93 } map; 94}; 95 96/* BPF syscall commands, see bpf(2) man-page for more details. */ 97/** 98 * DOC: eBPF Syscall Preamble 99 * 100 * The operation to be performed by the **bpf**\ () system call is determined 101 * by the *cmd* argument. Each operation takes an accompanying argument, 102 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 103 * below). The size argument is the size of the union pointed to by *attr*. 104 */ 105/** 106 * DOC: eBPF Syscall Commands 107 * 108 * BPF_MAP_CREATE 109 * Description 110 * Create a map and return a file descriptor that refers to the 111 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 112 * is automatically enabled for the new file descriptor. 113 * 114 * Applying **close**\ (2) to the file descriptor returned by 115 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 116 * 117 * Return 118 * A new file descriptor (a nonnegative integer), or -1 if an 119 * error occurred (in which case, *errno* is set appropriately). 120 * 121 * BPF_MAP_LOOKUP_ELEM 122 * Description 123 * Look up an element with a given *key* in the map referred to 124 * by the file descriptor *map_fd*. 125 * 126 * The *flags* argument may be specified as one of the 127 * following: 128 * 129 * **BPF_F_LOCK** 130 * Look up the value of a spin-locked map without 131 * returning the lock. This must be specified if the 132 * elements contain a spinlock. 133 * 134 * Return 135 * Returns zero on success. On error, -1 is returned and *errno* 136 * is set appropriately. 137 * 138 * BPF_MAP_UPDATE_ELEM 139 * Description 140 * Create or update an element (key/value pair) in a specified map. 141 * 142 * The *flags* argument should be specified as one of the 143 * following: 144 * 145 * **BPF_ANY** 146 * Create a new element or update an existing element. 147 * **BPF_NOEXIST** 148 * Create a new element only if it did not exist. 149 * **BPF_EXIST** 150 * Update an existing element. 151 * **BPF_F_LOCK** 152 * Update a spin_lock-ed map element. 153 * 154 * Return 155 * Returns zero on success. On error, -1 is returned and *errno* 156 * is set appropriately. 157 * 158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 159 * **E2BIG**, **EEXIST**, or **ENOENT**. 160 * 161 * **E2BIG** 162 * The number of elements in the map reached the 163 * *max_entries* limit specified at map creation time. 164 * **EEXIST** 165 * If *flags* specifies **BPF_NOEXIST** and the element 166 * with *key* already exists in the map. 167 * **ENOENT** 168 * If *flags* specifies **BPF_EXIST** and the element with 169 * *key* does not exist in the map. 170 * 171 * BPF_MAP_DELETE_ELEM 172 * Description 173 * Look up and delete an element by key in a specified map. 174 * 175 * Return 176 * Returns zero on success. On error, -1 is returned and *errno* 177 * is set appropriately. 178 * 179 * BPF_MAP_GET_NEXT_KEY 180 * Description 181 * Look up an element by key in a specified map and return the key 182 * of the next element. Can be used to iterate over all elements 183 * in the map. 184 * 185 * Return 186 * Returns zero on success. On error, -1 is returned and *errno* 187 * is set appropriately. 188 * 189 * The following cases can be used to iterate over all elements of 190 * the map: 191 * 192 * * If *key* is not found, the operation returns zero and sets 193 * the *next_key* pointer to the key of the first element. 194 * * If *key* is found, the operation returns zero and sets the 195 * *next_key* pointer to the key of the next element. 196 * * If *key* is the last element, returns -1 and *errno* is set 197 * to **ENOENT**. 198 * 199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 200 * **EINVAL** on error. 201 * 202 * BPF_PROG_LOAD 203 * Description 204 * Verify and load an eBPF program, returning a new file 205 * descriptor associated with the program. 206 * 207 * Applying **close**\ (2) to the file descriptor returned by 208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 209 * 210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 211 * automatically enabled for the new file descriptor. 212 * 213 * Return 214 * A new file descriptor (a nonnegative integer), or -1 if an 215 * error occurred (in which case, *errno* is set appropriately). 216 * 217 * BPF_OBJ_PIN 218 * Description 219 * Pin an eBPF program or map referred by the specified *bpf_fd* 220 * to the provided *pathname* on the filesystem. 221 * 222 * The *pathname* argument must not contain a dot ("."). 223 * 224 * On success, *pathname* retains a reference to the eBPF object, 225 * preventing deallocation of the object when the original 226 * *bpf_fd* is closed. This allow the eBPF object to live beyond 227 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 228 * process. 229 * 230 * Applying **unlink**\ (2) or similar calls to the *pathname* 231 * unpins the object from the filesystem, removing the reference. 232 * If no other file descriptors or filesystem nodes refer to the 233 * same object, it will be deallocated (see NOTES). 234 * 235 * The filesystem type for the parent directory of *pathname* must 236 * be **BPF_FS_MAGIC**. 237 * 238 * Return 239 * Returns zero on success. On error, -1 is returned and *errno* 240 * is set appropriately. 241 * 242 * BPF_OBJ_GET 243 * Description 244 * Open a file descriptor for the eBPF object pinned to the 245 * specified *pathname*. 246 * 247 * Return 248 * A new file descriptor (a nonnegative integer), or -1 if an 249 * error occurred (in which case, *errno* is set appropriately). 250 * 251 * BPF_PROG_ATTACH 252 * Description 253 * Attach an eBPF program to a *target_fd* at the specified 254 * *attach_type* hook. 255 * 256 * The *attach_type* specifies the eBPF attachment point to 257 * attach the program to, and must be one of *bpf_attach_type* 258 * (see below). 259 * 260 * The *attach_bpf_fd* must be a valid file descriptor for a 261 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 262 * or sock_ops type corresponding to the specified *attach_type*. 263 * 264 * The *target_fd* must be a valid file descriptor for a kernel 265 * object which depends on the attach type of *attach_bpf_fd*: 266 * 267 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 268 * **BPF_PROG_TYPE_CGROUP_SKB**, 269 * **BPF_PROG_TYPE_CGROUP_SOCK**, 270 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 271 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 272 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 273 * **BPF_PROG_TYPE_SOCK_OPS** 274 * 275 * Control Group v2 hierarchy with the eBPF controller 276 * enabled. Requires the kernel to be compiled with 277 * **CONFIG_CGROUP_BPF**. 278 * 279 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 280 * 281 * Network namespace (eg /proc/self/ns/net). 282 * 283 * **BPF_PROG_TYPE_LIRC_MODE2** 284 * 285 * LIRC device path (eg /dev/lircN). Requires the kernel 286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 287 * 288 * **BPF_PROG_TYPE_SK_SKB**, 289 * **BPF_PROG_TYPE_SK_MSG** 290 * 291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 292 * 293 * Return 294 * Returns zero on success. On error, -1 is returned and *errno* 295 * is set appropriately. 296 * 297 * BPF_PROG_DETACH 298 * Description 299 * Detach the eBPF program associated with the *target_fd* at the 300 * hook specified by *attach_type*. The program must have been 301 * previously attached using **BPF_PROG_ATTACH**. 302 * 303 * Return 304 * Returns zero on success. On error, -1 is returned and *errno* 305 * is set appropriately. 306 * 307 * BPF_PROG_TEST_RUN 308 * Description 309 * Run the eBPF program associated with the *prog_fd* a *repeat* 310 * number of times against a provided program context *ctx_in* and 311 * data *data_in*, and return the modified program context 312 * *ctx_out*, *data_out* (for example, packet data), result of the 313 * execution *retval*, and *duration* of the test run. 314 * 315 * The sizes of the buffers provided as input and output 316 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 317 * be provided in the corresponding variables *ctx_size_in*, 318 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 319 * of these parameters are not provided (ie set to NULL), the 320 * corresponding size field must be zero. 321 * 322 * Some program types have particular requirements: 323 * 324 * **BPF_PROG_TYPE_SK_LOOKUP** 325 * *data_in* and *data_out* must be NULL. 326 * 327 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 328 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 329 * 330 * *ctx_out*, *data_in* and *data_out* must be NULL. 331 * *repeat* must be zero. 332 * 333 * Return 334 * Returns zero on success. On error, -1 is returned and *errno* 335 * is set appropriately. 336 * 337 * **ENOSPC** 338 * Either *data_size_out* or *ctx_size_out* is too small. 339 * **ENOTSUPP** 340 * This command is not supported by the program type of 341 * the program referred to by *prog_fd*. 342 * 343 * BPF_PROG_GET_NEXT_ID 344 * Description 345 * Fetch the next eBPF program currently loaded into the kernel. 346 * 347 * Looks for the eBPF program with an id greater than *start_id* 348 * and updates *next_id* on success. If no other eBPF programs 349 * remain with ids higher than *start_id*, returns -1 and sets 350 * *errno* to **ENOENT**. 351 * 352 * Return 353 * Returns zero on success. On error, or when no id remains, -1 354 * is returned and *errno* is set appropriately. 355 * 356 * BPF_MAP_GET_NEXT_ID 357 * Description 358 * Fetch the next eBPF map currently loaded into the kernel. 359 * 360 * Looks for the eBPF map with an id greater than *start_id* 361 * and updates *next_id* on success. If no other eBPF maps 362 * remain with ids higher than *start_id*, returns -1 and sets 363 * *errno* to **ENOENT**. 364 * 365 * Return 366 * Returns zero on success. On error, or when no id remains, -1 367 * is returned and *errno* is set appropriately. 368 * 369 * BPF_PROG_GET_FD_BY_ID 370 * Description 371 * Open a file descriptor for the eBPF program corresponding to 372 * *prog_id*. 373 * 374 * Return 375 * A new file descriptor (a nonnegative integer), or -1 if an 376 * error occurred (in which case, *errno* is set appropriately). 377 * 378 * BPF_MAP_GET_FD_BY_ID 379 * Description 380 * Open a file descriptor for the eBPF map corresponding to 381 * *map_id*. 382 * 383 * Return 384 * A new file descriptor (a nonnegative integer), or -1 if an 385 * error occurred (in which case, *errno* is set appropriately). 386 * 387 * BPF_OBJ_GET_INFO_BY_FD 388 * Description 389 * Obtain information about the eBPF object corresponding to 390 * *bpf_fd*. 391 * 392 * Populates up to *info_len* bytes of *info*, which will be in 393 * one of the following formats depending on the eBPF object type 394 * of *bpf_fd*: 395 * 396 * * **struct bpf_prog_info** 397 * * **struct bpf_map_info** 398 * * **struct bpf_btf_info** 399 * * **struct bpf_link_info** 400 * 401 * Return 402 * Returns zero on success. On error, -1 is returned and *errno* 403 * is set appropriately. 404 * 405 * BPF_PROG_QUERY 406 * Description 407 * Obtain information about eBPF programs associated with the 408 * specified *attach_type* hook. 409 * 410 * The *target_fd* must be a valid file descriptor for a kernel 411 * object which depends on the attach type of *attach_bpf_fd*: 412 * 413 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 414 * **BPF_PROG_TYPE_CGROUP_SKB**, 415 * **BPF_PROG_TYPE_CGROUP_SOCK**, 416 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 417 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 418 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 419 * **BPF_PROG_TYPE_SOCK_OPS** 420 * 421 * Control Group v2 hierarchy with the eBPF controller 422 * enabled. Requires the kernel to be compiled with 423 * **CONFIG_CGROUP_BPF**. 424 * 425 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 426 * 427 * Network namespace (eg /proc/self/ns/net). 428 * 429 * **BPF_PROG_TYPE_LIRC_MODE2** 430 * 431 * LIRC device path (eg /dev/lircN). Requires the kernel 432 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 433 * 434 * **BPF_PROG_QUERY** always fetches the number of programs 435 * attached and the *attach_flags* which were used to attach those 436 * programs. Additionally, if *prog_ids* is nonzero and the number 437 * of attached programs is less than *prog_cnt*, populates 438 * *prog_ids* with the eBPF program ids of the programs attached 439 * at *target_fd*. 440 * 441 * The following flags may alter the result: 442 * 443 * **BPF_F_QUERY_EFFECTIVE** 444 * Only return information regarding programs which are 445 * currently effective at the specified *target_fd*. 446 * 447 * Return 448 * Returns zero on success. On error, -1 is returned and *errno* 449 * is set appropriately. 450 * 451 * BPF_RAW_TRACEPOINT_OPEN 452 * Description 453 * Attach an eBPF program to a tracepoint *name* to access kernel 454 * internal arguments of the tracepoint in their raw form. 455 * 456 * The *prog_fd* must be a valid file descriptor associated with 457 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 458 * 459 * No ABI guarantees are made about the content of tracepoint 460 * arguments exposed to the corresponding eBPF program. 461 * 462 * Applying **close**\ (2) to the file descriptor returned by 463 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 464 * 465 * Return 466 * A new file descriptor (a nonnegative integer), or -1 if an 467 * error occurred (in which case, *errno* is set appropriately). 468 * 469 * BPF_BTF_LOAD 470 * Description 471 * Verify and load BPF Type Format (BTF) metadata into the kernel, 472 * returning a new file descriptor associated with the metadata. 473 * BTF is described in more detail at 474 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 475 * 476 * The *btf* parameter must point to valid memory providing 477 * *btf_size* bytes of BTF binary metadata. 478 * 479 * The returned file descriptor can be passed to other **bpf**\ () 480 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 481 * associate the BTF with those objects. 482 * 483 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 484 * parameters to specify a *btf_log_buf*, *btf_log_size* and 485 * *btf_log_level* which allow the kernel to return freeform log 486 * output regarding the BTF verification process. 487 * 488 * Return 489 * A new file descriptor (a nonnegative integer), or -1 if an 490 * error occurred (in which case, *errno* is set appropriately). 491 * 492 * BPF_BTF_GET_FD_BY_ID 493 * Description 494 * Open a file descriptor for the BPF Type Format (BTF) 495 * corresponding to *btf_id*. 496 * 497 * Return 498 * A new file descriptor (a nonnegative integer), or -1 if an 499 * error occurred (in which case, *errno* is set appropriately). 500 * 501 * BPF_TASK_FD_QUERY 502 * Description 503 * Obtain information about eBPF programs associated with the 504 * target process identified by *pid* and *fd*. 505 * 506 * If the *pid* and *fd* are associated with a tracepoint, kprobe 507 * or uprobe perf event, then the *prog_id* and *fd_type* will 508 * be populated with the eBPF program id and file descriptor type 509 * of type **bpf_task_fd_type**. If associated with a kprobe or 510 * uprobe, the *probe_offset* and *probe_addr* will also be 511 * populated. Optionally, if *buf* is provided, then up to 512 * *buf_len* bytes of *buf* will be populated with the name of 513 * the tracepoint, kprobe or uprobe. 514 * 515 * The resulting *prog_id* may be introspected in deeper detail 516 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 517 * 518 * Return 519 * Returns zero on success. On error, -1 is returned and *errno* 520 * is set appropriately. 521 * 522 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 523 * Description 524 * Look up an element with the given *key* in the map referred to 525 * by the file descriptor *fd*, and if found, delete the element. 526 * 527 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 528 * types, the *flags* argument needs to be set to 0, but for other 529 * map types, it may be specified as: 530 * 531 * **BPF_F_LOCK** 532 * Look up and delete the value of a spin-locked map 533 * without returning the lock. This must be specified if 534 * the elements contain a spinlock. 535 * 536 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 537 * implement this command as a "pop" operation, deleting the top 538 * element rather than one corresponding to *key*. 539 * The *key* and *key_len* parameters should be zeroed when 540 * issuing this operation for these map types. 541 * 542 * This command is only valid for the following map types: 543 * * **BPF_MAP_TYPE_QUEUE** 544 * * **BPF_MAP_TYPE_STACK** 545 * * **BPF_MAP_TYPE_HASH** 546 * * **BPF_MAP_TYPE_PERCPU_HASH** 547 * * **BPF_MAP_TYPE_LRU_HASH** 548 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 549 * 550 * Return 551 * Returns zero on success. On error, -1 is returned and *errno* 552 * is set appropriately. 553 * 554 * BPF_MAP_FREEZE 555 * Description 556 * Freeze the permissions of the specified map. 557 * 558 * Write permissions may be frozen by passing zero *flags*. 559 * Upon success, no future syscall invocations may alter the 560 * map state of *map_fd*. Write operations from eBPF programs 561 * are still possible for a frozen map. 562 * 563 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 564 * 565 * Return 566 * Returns zero on success. On error, -1 is returned and *errno* 567 * is set appropriately. 568 * 569 * BPF_BTF_GET_NEXT_ID 570 * Description 571 * Fetch the next BPF Type Format (BTF) object currently loaded 572 * into the kernel. 573 * 574 * Looks for the BTF object with an id greater than *start_id* 575 * and updates *next_id* on success. If no other BTF objects 576 * remain with ids higher than *start_id*, returns -1 and sets 577 * *errno* to **ENOENT**. 578 * 579 * Return 580 * Returns zero on success. On error, or when no id remains, -1 581 * is returned and *errno* is set appropriately. 582 * 583 * BPF_MAP_LOOKUP_BATCH 584 * Description 585 * Iterate and fetch multiple elements in a map. 586 * 587 * Two opaque values are used to manage batch operations, 588 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 589 * to NULL to begin the batched operation. After each subsequent 590 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 591 * *out_batch* as the *in_batch* for the next operation to 592 * continue iteration from the current point. 593 * 594 * The *keys* and *values* are output parameters which must point 595 * to memory large enough to hold *count* items based on the key 596 * and value size of the map *map_fd*. The *keys* buffer must be 597 * of *key_size* * *count*. The *values* buffer must be of 598 * *value_size* * *count*. 599 * 600 * The *elem_flags* argument may be specified as one of the 601 * following: 602 * 603 * **BPF_F_LOCK** 604 * Look up the value of a spin-locked map without 605 * returning the lock. This must be specified if the 606 * elements contain a spinlock. 607 * 608 * On success, *count* elements from the map are copied into the 609 * user buffer, with the keys copied into *keys* and the values 610 * copied into the corresponding indices in *values*. 611 * 612 * If an error is returned and *errno* is not **EFAULT**, *count* 613 * is set to the number of successfully processed elements. 614 * 615 * Return 616 * Returns zero on success. On error, -1 is returned and *errno* 617 * is set appropriately. 618 * 619 * May set *errno* to **ENOSPC** to indicate that *keys* or 620 * *values* is too small to dump an entire bucket during 621 * iteration of a hash-based map type. 622 * 623 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 624 * Description 625 * Iterate and delete all elements in a map. 626 * 627 * This operation has the same behavior as 628 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 629 * 630 * * Every element that is successfully returned is also deleted 631 * from the map. This is at least *count* elements. Note that 632 * *count* is both an input and an output parameter. 633 * * Upon returning with *errno* set to **EFAULT**, up to 634 * *count* elements may be deleted without returning the keys 635 * and values of the deleted elements. 636 * 637 * Return 638 * Returns zero on success. On error, -1 is returned and *errno* 639 * is set appropriately. 640 * 641 * BPF_MAP_UPDATE_BATCH 642 * Description 643 * Update multiple elements in a map by *key*. 644 * 645 * The *keys* and *values* are input parameters which must point 646 * to memory large enough to hold *count* items based on the key 647 * and value size of the map *map_fd*. The *keys* buffer must be 648 * of *key_size* * *count*. The *values* buffer must be of 649 * *value_size* * *count*. 650 * 651 * Each element specified in *keys* is sequentially updated to the 652 * value in the corresponding index in *values*. The *in_batch* 653 * and *out_batch* parameters are ignored and should be zeroed. 654 * 655 * The *elem_flags* argument should be specified as one of the 656 * following: 657 * 658 * **BPF_ANY** 659 * Create new elements or update a existing elements. 660 * **BPF_NOEXIST** 661 * Create new elements only if they do not exist. 662 * **BPF_EXIST** 663 * Update existing elements. 664 * **BPF_F_LOCK** 665 * Update spin_lock-ed map elements. This must be 666 * specified if the map value contains a spinlock. 667 * 668 * On success, *count* elements from the map are updated. 669 * 670 * If an error is returned and *errno* is not **EFAULT**, *count* 671 * is set to the number of successfully processed elements. 672 * 673 * Return 674 * Returns zero on success. On error, -1 is returned and *errno* 675 * is set appropriately. 676 * 677 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 678 * **E2BIG**. **E2BIG** indicates that the number of elements in 679 * the map reached the *max_entries* limit specified at map 680 * creation time. 681 * 682 * May set *errno* to one of the following error codes under 683 * specific circumstances: 684 * 685 * **EEXIST** 686 * If *flags* specifies **BPF_NOEXIST** and the element 687 * with *key* already exists in the map. 688 * **ENOENT** 689 * If *flags* specifies **BPF_EXIST** and the element with 690 * *key* does not exist in the map. 691 * 692 * BPF_MAP_DELETE_BATCH 693 * Description 694 * Delete multiple elements in a map by *key*. 695 * 696 * The *keys* parameter is an input parameter which must point 697 * to memory large enough to hold *count* items based on the key 698 * size of the map *map_fd*, that is, *key_size* * *count*. 699 * 700 * Each element specified in *keys* is sequentially deleted. The 701 * *in_batch*, *out_batch*, and *values* parameters are ignored 702 * and should be zeroed. 703 * 704 * The *elem_flags* argument may be specified as one of the 705 * following: 706 * 707 * **BPF_F_LOCK** 708 * Look up the value of a spin-locked map without 709 * returning the lock. This must be specified if the 710 * elements contain a spinlock. 711 * 712 * On success, *count* elements from the map are updated. 713 * 714 * If an error is returned and *errno* is not **EFAULT**, *count* 715 * is set to the number of successfully processed elements. If 716 * *errno* is **EFAULT**, up to *count* elements may be been 717 * deleted. 718 * 719 * Return 720 * Returns zero on success. On error, -1 is returned and *errno* 721 * is set appropriately. 722 * 723 * BPF_LINK_CREATE 724 * Description 725 * Attach an eBPF program to a *target_fd* at the specified 726 * *attach_type* hook and return a file descriptor handle for 727 * managing the link. 728 * 729 * Return 730 * A new file descriptor (a nonnegative integer), or -1 if an 731 * error occurred (in which case, *errno* is set appropriately). 732 * 733 * BPF_LINK_UPDATE 734 * Description 735 * Update the eBPF program in the specified *link_fd* to 736 * *new_prog_fd*. 737 * 738 * Return 739 * Returns zero on success. On error, -1 is returned and *errno* 740 * is set appropriately. 741 * 742 * BPF_LINK_GET_FD_BY_ID 743 * Description 744 * Open a file descriptor for the eBPF Link corresponding to 745 * *link_id*. 746 * 747 * Return 748 * A new file descriptor (a nonnegative integer), or -1 if an 749 * error occurred (in which case, *errno* is set appropriately). 750 * 751 * BPF_LINK_GET_NEXT_ID 752 * Description 753 * Fetch the next eBPF link currently loaded into the kernel. 754 * 755 * Looks for the eBPF link with an id greater than *start_id* 756 * and updates *next_id* on success. If no other eBPF links 757 * remain with ids higher than *start_id*, returns -1 and sets 758 * *errno* to **ENOENT**. 759 * 760 * Return 761 * Returns zero on success. On error, or when no id remains, -1 762 * is returned and *errno* is set appropriately. 763 * 764 * BPF_ENABLE_STATS 765 * Description 766 * Enable eBPF runtime statistics gathering. 767 * 768 * Runtime statistics gathering for the eBPF runtime is disabled 769 * by default to minimize the corresponding performance overhead. 770 * This command enables statistics globally. 771 * 772 * Multiple programs may independently enable statistics. 773 * After gathering the desired statistics, eBPF runtime statistics 774 * may be disabled again by calling **close**\ (2) for the file 775 * descriptor returned by this function. Statistics will only be 776 * disabled system-wide when all outstanding file descriptors 777 * returned by prior calls for this subcommand are closed. 778 * 779 * Return 780 * A new file descriptor (a nonnegative integer), or -1 if an 781 * error occurred (in which case, *errno* is set appropriately). 782 * 783 * BPF_ITER_CREATE 784 * Description 785 * Create an iterator on top of the specified *link_fd* (as 786 * previously created using **BPF_LINK_CREATE**) and return a 787 * file descriptor that can be used to trigger the iteration. 788 * 789 * If the resulting file descriptor is pinned to the filesystem 790 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 791 * for that path will trigger the iterator to read kernel state 792 * using the eBPF program attached to *link_fd*. 793 * 794 * Return 795 * A new file descriptor (a nonnegative integer), or -1 if an 796 * error occurred (in which case, *errno* is set appropriately). 797 * 798 * BPF_LINK_DETACH 799 * Description 800 * Forcefully detach the specified *link_fd* from its 801 * corresponding attachment point. 802 * 803 * Return 804 * Returns zero on success. On error, -1 is returned and *errno* 805 * is set appropriately. 806 * 807 * BPF_PROG_BIND_MAP 808 * Description 809 * Bind a map to the lifetime of an eBPF program. 810 * 811 * The map identified by *map_fd* is bound to the program 812 * identified by *prog_fd* and only released when *prog_fd* is 813 * released. This may be used in cases where metadata should be 814 * associated with a program which otherwise does not contain any 815 * references to the map (for example, embedded in the eBPF 816 * program instructions). 817 * 818 * Return 819 * Returns zero on success. On error, -1 is returned and *errno* 820 * is set appropriately. 821 * 822 * NOTES 823 * eBPF objects (maps and programs) can be shared between processes. 824 * 825 * * After **fork**\ (2), the child inherits file descriptors 826 * referring to the same eBPF objects. 827 * * File descriptors referring to eBPF objects can be transferred over 828 * **unix**\ (7) domain sockets. 829 * * File descriptors referring to eBPF objects can be duplicated in the 830 * usual way, using **dup**\ (2) and similar calls. 831 * * File descriptors referring to eBPF objects can be pinned to the 832 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 833 * 834 * An eBPF object is deallocated only after all file descriptors referring 835 * to the object have been closed and no references remain pinned to the 836 * filesystem or attached (for example, bound to a program or device). 837 */ 838enum bpf_cmd { 839 BPF_MAP_CREATE, 840 BPF_MAP_LOOKUP_ELEM, 841 BPF_MAP_UPDATE_ELEM, 842 BPF_MAP_DELETE_ELEM, 843 BPF_MAP_GET_NEXT_KEY, 844 BPF_PROG_LOAD, 845 BPF_OBJ_PIN, 846 BPF_OBJ_GET, 847 BPF_PROG_ATTACH, 848 BPF_PROG_DETACH, 849 BPF_PROG_TEST_RUN, 850 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 851 BPF_PROG_GET_NEXT_ID, 852 BPF_MAP_GET_NEXT_ID, 853 BPF_PROG_GET_FD_BY_ID, 854 BPF_MAP_GET_FD_BY_ID, 855 BPF_OBJ_GET_INFO_BY_FD, 856 BPF_PROG_QUERY, 857 BPF_RAW_TRACEPOINT_OPEN, 858 BPF_BTF_LOAD, 859 BPF_BTF_GET_FD_BY_ID, 860 BPF_TASK_FD_QUERY, 861 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 862 BPF_MAP_FREEZE, 863 BPF_BTF_GET_NEXT_ID, 864 BPF_MAP_LOOKUP_BATCH, 865 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 866 BPF_MAP_UPDATE_BATCH, 867 BPF_MAP_DELETE_BATCH, 868 BPF_LINK_CREATE, 869 BPF_LINK_UPDATE, 870 BPF_LINK_GET_FD_BY_ID, 871 BPF_LINK_GET_NEXT_ID, 872 BPF_ENABLE_STATS, 873 BPF_ITER_CREATE, 874 BPF_LINK_DETACH, 875 BPF_PROG_BIND_MAP, 876}; 877 878enum bpf_map_type { 879 BPF_MAP_TYPE_UNSPEC, 880 BPF_MAP_TYPE_HASH, 881 BPF_MAP_TYPE_ARRAY, 882 BPF_MAP_TYPE_PROG_ARRAY, 883 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 884 BPF_MAP_TYPE_PERCPU_HASH, 885 BPF_MAP_TYPE_PERCPU_ARRAY, 886 BPF_MAP_TYPE_STACK_TRACE, 887 BPF_MAP_TYPE_CGROUP_ARRAY, 888 BPF_MAP_TYPE_LRU_HASH, 889 BPF_MAP_TYPE_LRU_PERCPU_HASH, 890 BPF_MAP_TYPE_LPM_TRIE, 891 BPF_MAP_TYPE_ARRAY_OF_MAPS, 892 BPF_MAP_TYPE_HASH_OF_MAPS, 893 BPF_MAP_TYPE_DEVMAP, 894 BPF_MAP_TYPE_SOCKMAP, 895 BPF_MAP_TYPE_CPUMAP, 896 BPF_MAP_TYPE_XSKMAP, 897 BPF_MAP_TYPE_SOCKHASH, 898 BPF_MAP_TYPE_CGROUP_STORAGE, 899 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 900 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 901 BPF_MAP_TYPE_QUEUE, 902 BPF_MAP_TYPE_STACK, 903 BPF_MAP_TYPE_SK_STORAGE, 904 BPF_MAP_TYPE_DEVMAP_HASH, 905 BPF_MAP_TYPE_STRUCT_OPS, 906 BPF_MAP_TYPE_RINGBUF, 907 BPF_MAP_TYPE_INODE_STORAGE, 908 BPF_MAP_TYPE_TASK_STORAGE, 909 BPF_MAP_TYPE_BLOOM_FILTER, 910}; 911 912/* Note that tracing related programs such as 913 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 914 * are not subject to a stable API since kernel internal data 915 * structures can change from release to release and may 916 * therefore break existing tracing BPF programs. Tracing BPF 917 * programs correspond to /a/ specific kernel which is to be 918 * analyzed, and not /a/ specific kernel /and/ all future ones. 919 */ 920enum bpf_prog_type { 921 BPF_PROG_TYPE_UNSPEC, 922 BPF_PROG_TYPE_SOCKET_FILTER, 923 BPF_PROG_TYPE_KPROBE, 924 BPF_PROG_TYPE_SCHED_CLS, 925 BPF_PROG_TYPE_SCHED_ACT, 926 BPF_PROG_TYPE_TRACEPOINT, 927 BPF_PROG_TYPE_XDP, 928 BPF_PROG_TYPE_PERF_EVENT, 929 BPF_PROG_TYPE_CGROUP_SKB, 930 BPF_PROG_TYPE_CGROUP_SOCK, 931 BPF_PROG_TYPE_LWT_IN, 932 BPF_PROG_TYPE_LWT_OUT, 933 BPF_PROG_TYPE_LWT_XMIT, 934 BPF_PROG_TYPE_SOCK_OPS, 935 BPF_PROG_TYPE_SK_SKB, 936 BPF_PROG_TYPE_CGROUP_DEVICE, 937 BPF_PROG_TYPE_SK_MSG, 938 BPF_PROG_TYPE_RAW_TRACEPOINT, 939 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 940 BPF_PROG_TYPE_LWT_SEG6LOCAL, 941 BPF_PROG_TYPE_LIRC_MODE2, 942 BPF_PROG_TYPE_SK_REUSEPORT, 943 BPF_PROG_TYPE_FLOW_DISSECTOR, 944 BPF_PROG_TYPE_CGROUP_SYSCTL, 945 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 946 BPF_PROG_TYPE_CGROUP_SOCKOPT, 947 BPF_PROG_TYPE_TRACING, 948 BPF_PROG_TYPE_STRUCT_OPS, 949 BPF_PROG_TYPE_EXT, 950 BPF_PROG_TYPE_LSM, 951 BPF_PROG_TYPE_SK_LOOKUP, 952 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 953}; 954 955enum bpf_attach_type { 956 BPF_CGROUP_INET_INGRESS, 957 BPF_CGROUP_INET_EGRESS, 958 BPF_CGROUP_INET_SOCK_CREATE, 959 BPF_CGROUP_SOCK_OPS, 960 BPF_SK_SKB_STREAM_PARSER, 961 BPF_SK_SKB_STREAM_VERDICT, 962 BPF_CGROUP_DEVICE, 963 BPF_SK_MSG_VERDICT, 964 BPF_CGROUP_INET4_BIND, 965 BPF_CGROUP_INET6_BIND, 966 BPF_CGROUP_INET4_CONNECT, 967 BPF_CGROUP_INET6_CONNECT, 968 BPF_CGROUP_INET4_POST_BIND, 969 BPF_CGROUP_INET6_POST_BIND, 970 BPF_CGROUP_UDP4_SENDMSG, 971 BPF_CGROUP_UDP6_SENDMSG, 972 BPF_LIRC_MODE2, 973 BPF_FLOW_DISSECTOR, 974 BPF_CGROUP_SYSCTL, 975 BPF_CGROUP_UDP4_RECVMSG, 976 BPF_CGROUP_UDP6_RECVMSG, 977 BPF_CGROUP_GETSOCKOPT, 978 BPF_CGROUP_SETSOCKOPT, 979 BPF_TRACE_RAW_TP, 980 BPF_TRACE_FENTRY, 981 BPF_TRACE_FEXIT, 982 BPF_MODIFY_RETURN, 983 BPF_LSM_MAC, 984 BPF_TRACE_ITER, 985 BPF_CGROUP_INET4_GETPEERNAME, 986 BPF_CGROUP_INET6_GETPEERNAME, 987 BPF_CGROUP_INET4_GETSOCKNAME, 988 BPF_CGROUP_INET6_GETSOCKNAME, 989 BPF_XDP_DEVMAP, 990 BPF_CGROUP_INET_SOCK_RELEASE, 991 BPF_XDP_CPUMAP, 992 BPF_SK_LOOKUP, 993 BPF_XDP, 994 BPF_SK_SKB_VERDICT, 995 BPF_SK_REUSEPORT_SELECT, 996 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 997 BPF_PERF_EVENT, 998 __MAX_BPF_ATTACH_TYPE 999}; 1000 1001#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1002 1003enum bpf_link_type { 1004 BPF_LINK_TYPE_UNSPEC = 0, 1005 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1006 BPF_LINK_TYPE_TRACING = 2, 1007 BPF_LINK_TYPE_CGROUP = 3, 1008 BPF_LINK_TYPE_ITER = 4, 1009 BPF_LINK_TYPE_NETNS = 5, 1010 BPF_LINK_TYPE_XDP = 6, 1011 BPF_LINK_TYPE_PERF_EVENT = 7, 1012 1013 MAX_BPF_LINK_TYPE, 1014}; 1015 1016/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1017 * 1018 * NONE(default): No further bpf programs allowed in the subtree. 1019 * 1020 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1021 * the program in this cgroup yields to sub-cgroup program. 1022 * 1023 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1024 * that cgroup program gets run in addition to the program in this cgroup. 1025 * 1026 * Only one program is allowed to be attached to a cgroup with 1027 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1028 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1029 * release old program and attach the new one. Attach flags has to match. 1030 * 1031 * Multiple programs are allowed to be attached to a cgroup with 1032 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1033 * (those that were attached first, run first) 1034 * The programs of sub-cgroup are executed first, then programs of 1035 * this cgroup and then programs of parent cgroup. 1036 * When children program makes decision (like picking TCP CA or sock bind) 1037 * parent program has a chance to override it. 1038 * 1039 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1040 * programs for a cgroup. Though it's possible to replace an old program at 1041 * any position by also specifying BPF_F_REPLACE flag and position itself in 1042 * replace_bpf_fd attribute. Old program at this position will be released. 1043 * 1044 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1045 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1046 * Ex1: 1047 * cgrp1 (MULTI progs A, B) -> 1048 * cgrp2 (OVERRIDE prog C) -> 1049 * cgrp3 (MULTI prog D) -> 1050 * cgrp4 (OVERRIDE prog E) -> 1051 * cgrp5 (NONE prog F) 1052 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1053 * if prog F is detached, the execution is E,D,A,B 1054 * if prog F and D are detached, the execution is E,A,B 1055 * if prog F, E and D are detached, the execution is C,A,B 1056 * 1057 * All eligible programs are executed regardless of return code from 1058 * earlier programs. 1059 */ 1060#define BPF_F_ALLOW_OVERRIDE (1U << 0) 1061#define BPF_F_ALLOW_MULTI (1U << 1) 1062#define BPF_F_REPLACE (1U << 2) 1063 1064/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1065 * verifier will perform strict alignment checking as if the kernel 1066 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1067 * and NET_IP_ALIGN defined to 2. 1068 */ 1069#define BPF_F_STRICT_ALIGNMENT (1U << 0) 1070 1071/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1072 * verifier will allow any alignment whatsoever. On platforms 1073 * with strict alignment requirements for loads ands stores (such 1074 * as sparc and mips) the verifier validates that all loads and 1075 * stores provably follow this requirement. This flag turns that 1076 * checking and enforcement off. 1077 * 1078 * It is mostly used for testing when we want to validate the 1079 * context and memory access aspects of the verifier, but because 1080 * of an unaligned access the alignment check would trigger before 1081 * the one we are interested in. 1082 */ 1083#define BPF_F_ANY_ALIGNMENT (1U << 1) 1084 1085/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1086 * Verifier does sub-register def/use analysis and identifies instructions whose 1087 * def only matters for low 32-bit, high 32-bit is never referenced later 1088 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1089 * that it is safe to ignore clearing high 32-bit for these instructions. This 1090 * saves some back-ends a lot of code-gen. However such optimization is not 1091 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1092 * hence hasn't used verifier's analysis result. But, we really want to have a 1093 * way to be able to verify the correctness of the described optimization on 1094 * x86_64 on which testsuites are frequently exercised. 1095 * 1096 * So, this flag is introduced. Once it is set, verifier will randomize high 1097 * 32-bit for those instructions who has been identified as safe to ignore them. 1098 * Then, if verifier is not doing correct analysis, such randomization will 1099 * regress tests to expose bugs. 1100 */ 1101#define BPF_F_TEST_RND_HI32 (1U << 2) 1102 1103/* The verifier internal test flag. Behavior is undefined */ 1104#define BPF_F_TEST_STATE_FREQ (1U << 3) 1105 1106/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1107 * restrict map and helper usage for such programs. Sleepable BPF programs can 1108 * only be attached to hooks where kernel execution context allows sleeping. 1109 * Such programs are allowed to use helpers that may sleep like 1110 * bpf_copy_from_user(). 1111 */ 1112#define BPF_F_SLEEPABLE (1U << 4) 1113 1114/* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1115 * the following extensions: 1116 * 1117 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1118 * insn[0].imm: map fd or fd_idx 1119 * insn[1].imm: 0 1120 * insn[0].off: 0 1121 * insn[1].off: 0 1122 * ldimm64 rewrite: address of map 1123 * verifier type: CONST_PTR_TO_MAP 1124 */ 1125#define BPF_PSEUDO_MAP_FD 1 1126#define BPF_PSEUDO_MAP_IDX 5 1127 1128/* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1129 * insn[0].imm: map fd or fd_idx 1130 * insn[1].imm: offset into value 1131 * insn[0].off: 0 1132 * insn[1].off: 0 1133 * ldimm64 rewrite: address of map[0]+offset 1134 * verifier type: PTR_TO_MAP_VALUE 1135 */ 1136#define BPF_PSEUDO_MAP_VALUE 2 1137#define BPF_PSEUDO_MAP_IDX_VALUE 6 1138 1139/* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1140 * insn[0].imm: kernel btd id of VAR 1141 * insn[1].imm: 0 1142 * insn[0].off: 0 1143 * insn[1].off: 0 1144 * ldimm64 rewrite: address of the kernel variable 1145 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1146 * is struct/union. 1147 */ 1148#define BPF_PSEUDO_BTF_ID 3 1149/* insn[0].src_reg: BPF_PSEUDO_FUNC 1150 * insn[0].imm: insn offset to the func 1151 * insn[1].imm: 0 1152 * insn[0].off: 0 1153 * insn[1].off: 0 1154 * ldimm64 rewrite: address of the function 1155 * verifier type: PTR_TO_FUNC. 1156 */ 1157#define BPF_PSEUDO_FUNC 4 1158 1159/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1160 * offset to another bpf function 1161 */ 1162#define BPF_PSEUDO_CALL 1 1163/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1164 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1165 */ 1166#define BPF_PSEUDO_KFUNC_CALL 2 1167 1168/* flags for BPF_MAP_UPDATE_ELEM command */ 1169enum { 1170 BPF_ANY = 0, /* create new element or update existing */ 1171 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1172 BPF_EXIST = 2, /* update existing element */ 1173 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1174}; 1175 1176/* flags for BPF_MAP_CREATE command */ 1177enum { 1178 BPF_F_NO_PREALLOC = (1U << 0), 1179/* Instead of having one common LRU list in the 1180 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1181 * which can scale and perform better. 1182 * Note, the LRU nodes (including free nodes) cannot be moved 1183 * across different LRU lists. 1184 */ 1185 BPF_F_NO_COMMON_LRU = (1U << 1), 1186/* Specify numa node during map creation */ 1187 BPF_F_NUMA_NODE = (1U << 2), 1188 1189/* Flags for accessing BPF object from syscall side. */ 1190 BPF_F_RDONLY = (1U << 3), 1191 BPF_F_WRONLY = (1U << 4), 1192 1193/* Flag for stack_map, store build_id+offset instead of pointer */ 1194 BPF_F_STACK_BUILD_ID = (1U << 5), 1195 1196/* Zero-initialize hash function seed. This should only be used for testing. */ 1197 BPF_F_ZERO_SEED = (1U << 6), 1198 1199/* Flags for accessing BPF object from program side. */ 1200 BPF_F_RDONLY_PROG = (1U << 7), 1201 BPF_F_WRONLY_PROG = (1U << 8), 1202 1203/* Clone map from listener for newly accepted socket */ 1204 BPF_F_CLONE = (1U << 9), 1205 1206/* Enable memory-mapping BPF map */ 1207 BPF_F_MMAPABLE = (1U << 10), 1208 1209/* Share perf_event among processes */ 1210 BPF_F_PRESERVE_ELEMS = (1U << 11), 1211 1212/* Create a map that is suitable to be an inner map with dynamic max entries */ 1213 BPF_F_INNER_MAP = (1U << 12), 1214}; 1215 1216/* Flags for BPF_PROG_QUERY. */ 1217 1218/* Query effective (directly attached + inherited from ancestor cgroups) 1219 * programs that will be executed for events within a cgroup. 1220 * attach_flags with this flag are returned only for directly attached programs. 1221 */ 1222#define BPF_F_QUERY_EFFECTIVE (1U << 0) 1223 1224/* Flags for BPF_PROG_TEST_RUN */ 1225 1226/* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1227#define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1228 1229/* type for BPF_ENABLE_STATS */ 1230enum bpf_stats_type { 1231 /* enabled run_time_ns and run_cnt */ 1232 BPF_STATS_RUN_TIME = 0, 1233}; 1234 1235enum bpf_stack_build_id_status { 1236 /* user space need an empty entry to identify end of a trace */ 1237 BPF_STACK_BUILD_ID_EMPTY = 0, 1238 /* with valid build_id and offset */ 1239 BPF_STACK_BUILD_ID_VALID = 1, 1240 /* couldn't get build_id, fallback to ip */ 1241 BPF_STACK_BUILD_ID_IP = 2, 1242}; 1243 1244#define BPF_BUILD_ID_SIZE 20 1245struct bpf_stack_build_id { 1246 __s32 status; 1247 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1248 union { 1249 __u64 offset; 1250 __u64 ip; 1251 }; 1252}; 1253 1254#define BPF_OBJ_NAME_LEN 16U 1255 1256union bpf_attr { 1257 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1258 __u32 map_type; /* one of enum bpf_map_type */ 1259 __u32 key_size; /* size of key in bytes */ 1260 __u32 value_size; /* size of value in bytes */ 1261 __u32 max_entries; /* max number of entries in a map */ 1262 __u32 map_flags; /* BPF_MAP_CREATE related 1263 * flags defined above. 1264 */ 1265 __u32 inner_map_fd; /* fd pointing to the inner map */ 1266 __u32 numa_node; /* numa node (effective only if 1267 * BPF_F_NUMA_NODE is set). 1268 */ 1269 char map_name[BPF_OBJ_NAME_LEN]; 1270 __u32 map_ifindex; /* ifindex of netdev to create on */ 1271 __u32 btf_fd; /* fd pointing to a BTF type data */ 1272 __u32 btf_key_type_id; /* BTF type_id of the key */ 1273 __u32 btf_value_type_id; /* BTF type_id of the value */ 1274 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1275 * struct stored as the 1276 * map value 1277 */ 1278 /* Any per-map-type extra fields 1279 * 1280 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1281 * number of hash functions (if 0, the bloom filter will default 1282 * to using 5 hash functions). 1283 */ 1284 __u64 map_extra; 1285 }; 1286 1287 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1288 __u32 map_fd; 1289 __aligned_u64 key; 1290 union { 1291 __aligned_u64 value; 1292 __aligned_u64 next_key; 1293 }; 1294 __u64 flags; 1295 }; 1296 1297 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1298 __aligned_u64 in_batch; /* start batch, 1299 * NULL to start from beginning 1300 */ 1301 __aligned_u64 out_batch; /* output: next start batch */ 1302 __aligned_u64 keys; 1303 __aligned_u64 values; 1304 __u32 count; /* input/output: 1305 * input: # of key/value 1306 * elements 1307 * output: # of filled elements 1308 */ 1309 __u32 map_fd; 1310 __u64 elem_flags; 1311 __u64 flags; 1312 } batch; 1313 1314 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1315 __u32 prog_type; /* one of enum bpf_prog_type */ 1316 __u32 insn_cnt; 1317 __aligned_u64 insns; 1318 __aligned_u64 license; 1319 __u32 log_level; /* verbosity level of verifier */ 1320 __u32 log_size; /* size of user buffer */ 1321 __aligned_u64 log_buf; /* user supplied buffer */ 1322 __u32 kern_version; /* not used */ 1323 __u32 prog_flags; 1324 char prog_name[BPF_OBJ_NAME_LEN]; 1325 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1326 /* For some prog types expected attach type must be known at 1327 * load time to verify attach type specific parts of prog 1328 * (context accesses, allowed helpers, etc). 1329 */ 1330 __u32 expected_attach_type; 1331 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1332 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1333 __aligned_u64 func_info; /* func info */ 1334 __u32 func_info_cnt; /* number of bpf_func_info records */ 1335 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1336 __aligned_u64 line_info; /* line info */ 1337 __u32 line_info_cnt; /* number of bpf_line_info records */ 1338 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1339 union { 1340 /* valid prog_fd to attach to bpf prog */ 1341 __u32 attach_prog_fd; 1342 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1343 __u32 attach_btf_obj_fd; 1344 }; 1345 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1346 __aligned_u64 fd_array; /* array of FDs */ 1347 __aligned_u64 core_relos; 1348 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1349 }; 1350 1351 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1352 __aligned_u64 pathname; 1353 __u32 bpf_fd; 1354 __u32 file_flags; 1355 }; 1356 1357 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1358 __u32 target_fd; /* container object to attach to */ 1359 __u32 attach_bpf_fd; /* eBPF program to attach */ 1360 __u32 attach_type; 1361 __u32 attach_flags; 1362 __u32 replace_bpf_fd; /* previously attached eBPF 1363 * program to replace if 1364 * BPF_F_REPLACE is used 1365 */ 1366 }; 1367 1368 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1369 __u32 prog_fd; 1370 __u32 retval; 1371 __u32 data_size_in; /* input: len of data_in */ 1372 __u32 data_size_out; /* input/output: len of data_out 1373 * returns ENOSPC if data_out 1374 * is too small. 1375 */ 1376 __aligned_u64 data_in; 1377 __aligned_u64 data_out; 1378 __u32 repeat; 1379 __u32 duration; 1380 __u32 ctx_size_in; /* input: len of ctx_in */ 1381 __u32 ctx_size_out; /* input/output: len of ctx_out 1382 * returns ENOSPC if ctx_out 1383 * is too small. 1384 */ 1385 __aligned_u64 ctx_in; 1386 __aligned_u64 ctx_out; 1387 __u32 flags; 1388 __u32 cpu; 1389 } test; 1390 1391 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1392 union { 1393 __u32 start_id; 1394 __u32 prog_id; 1395 __u32 map_id; 1396 __u32 btf_id; 1397 __u32 link_id; 1398 }; 1399 __u32 next_id; 1400 __u32 open_flags; 1401 }; 1402 1403 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1404 __u32 bpf_fd; 1405 __u32 info_len; 1406 __aligned_u64 info; 1407 } info; 1408 1409 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1410 __u32 target_fd; /* container object to query */ 1411 __u32 attach_type; 1412 __u32 query_flags; 1413 __u32 attach_flags; 1414 __aligned_u64 prog_ids; 1415 __u32 prog_cnt; 1416 } query; 1417 1418 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1419 __u64 name; 1420 __u32 prog_fd; 1421 } raw_tracepoint; 1422 1423 struct { /* anonymous struct for BPF_BTF_LOAD */ 1424 __aligned_u64 btf; 1425 __aligned_u64 btf_log_buf; 1426 __u32 btf_size; 1427 __u32 btf_log_size; 1428 __u32 btf_log_level; 1429 }; 1430 1431 struct { 1432 __u32 pid; /* input: pid */ 1433 __u32 fd; /* input: fd */ 1434 __u32 flags; /* input: flags */ 1435 __u32 buf_len; /* input/output: buf len */ 1436 __aligned_u64 buf; /* input/output: 1437 * tp_name for tracepoint 1438 * symbol for kprobe 1439 * filename for uprobe 1440 */ 1441 __u32 prog_id; /* output: prod_id */ 1442 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1443 __u64 probe_offset; /* output: probe_offset */ 1444 __u64 probe_addr; /* output: probe_addr */ 1445 } task_fd_query; 1446 1447 struct { /* struct used by BPF_LINK_CREATE command */ 1448 __u32 prog_fd; /* eBPF program to attach */ 1449 union { 1450 __u32 target_fd; /* object to attach to */ 1451 __u32 target_ifindex; /* target ifindex */ 1452 }; 1453 __u32 attach_type; /* attach type */ 1454 __u32 flags; /* extra flags */ 1455 union { 1456 __u32 target_btf_id; /* btf_id of target to attach to */ 1457 struct { 1458 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1459 __u32 iter_info_len; /* iter_info length */ 1460 }; 1461 struct { 1462 /* black box user-provided value passed through 1463 * to BPF program at the execution time and 1464 * accessible through bpf_get_attach_cookie() BPF helper 1465 */ 1466 __u64 bpf_cookie; 1467 } perf_event; 1468 }; 1469 } link_create; 1470 1471 struct { /* struct used by BPF_LINK_UPDATE command */ 1472 __u32 link_fd; /* link fd */ 1473 /* new program fd to update link with */ 1474 __u32 new_prog_fd; 1475 __u32 flags; /* extra flags */ 1476 /* expected link's program fd; is specified only if 1477 * BPF_F_REPLACE flag is set in flags */ 1478 __u32 old_prog_fd; 1479 } link_update; 1480 1481 struct { 1482 __u32 link_fd; 1483 } link_detach; 1484 1485 struct { /* struct used by BPF_ENABLE_STATS command */ 1486 __u32 type; 1487 } enable_stats; 1488 1489 struct { /* struct used by BPF_ITER_CREATE command */ 1490 __u32 link_fd; 1491 __u32 flags; 1492 } iter_create; 1493 1494 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1495 __u32 prog_fd; 1496 __u32 map_fd; 1497 __u32 flags; /* extra flags */ 1498 } prog_bind_map; 1499 1500} __attribute__((aligned(8))); 1501 1502/* The description below is an attempt at providing documentation to eBPF 1503 * developers about the multiple available eBPF helper functions. It can be 1504 * parsed and used to produce a manual page. The workflow is the following, 1505 * and requires the rst2man utility: 1506 * 1507 * $ ./scripts/bpf_doc.py \ 1508 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1509 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1510 * $ man /tmp/bpf-helpers.7 1511 * 1512 * Note that in order to produce this external documentation, some RST 1513 * formatting is used in the descriptions to get "bold" and "italics" in 1514 * manual pages. Also note that the few trailing white spaces are 1515 * intentional, removing them would break paragraphs for rst2man. 1516 * 1517 * Start of BPF helper function descriptions: 1518 * 1519 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1520 * Description 1521 * Perform a lookup in *map* for an entry associated to *key*. 1522 * Return 1523 * Map value associated to *key*, or **NULL** if no entry was 1524 * found. 1525 * 1526 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1527 * Description 1528 * Add or update the value of the entry associated to *key* in 1529 * *map* with *value*. *flags* is one of: 1530 * 1531 * **BPF_NOEXIST** 1532 * The entry for *key* must not exist in the map. 1533 * **BPF_EXIST** 1534 * The entry for *key* must already exist in the map. 1535 * **BPF_ANY** 1536 * No condition on the existence of the entry for *key*. 1537 * 1538 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1539 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1540 * elements always exist), the helper would return an error. 1541 * Return 1542 * 0 on success, or a negative error in case of failure. 1543 * 1544 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1545 * Description 1546 * Delete entry with *key* from *map*. 1547 * Return 1548 * 0 on success, or a negative error in case of failure. 1549 * 1550 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1551 * Description 1552 * For tracing programs, safely attempt to read *size* bytes from 1553 * kernel space address *unsafe_ptr* and store the data in *dst*. 1554 * 1555 * Generally, use **bpf_probe_read_user**\ () or 1556 * **bpf_probe_read_kernel**\ () instead. 1557 * Return 1558 * 0 on success, or a negative error in case of failure. 1559 * 1560 * u64 bpf_ktime_get_ns(void) 1561 * Description 1562 * Return the time elapsed since system boot, in nanoseconds. 1563 * Does not include time the system was suspended. 1564 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1565 * Return 1566 * Current *ktime*. 1567 * 1568 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1569 * Description 1570 * This helper is a "printk()-like" facility for debugging. It 1571 * prints a message defined by format *fmt* (of size *fmt_size*) 1572 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1573 * available. It can take up to three additional **u64** 1574 * arguments (as an eBPF helpers, the total number of arguments is 1575 * limited to five). 1576 * 1577 * Each time the helper is called, it appends a line to the trace. 1578 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 1579 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1580 * The format of the trace is customizable, and the exact output 1581 * one will get depends on the options set in 1582 * *\/sys/kernel/debug/tracing/trace_options* (see also the 1583 * *README* file under the same directory). However, it usually 1584 * defaults to something like: 1585 * 1586 * :: 1587 * 1588 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1589 * 1590 * In the above: 1591 * 1592 * * ``telnet`` is the name of the current task. 1593 * * ``470`` is the PID of the current task. 1594 * * ``001`` is the CPU number on which the task is 1595 * running. 1596 * * In ``.N..``, each character refers to a set of 1597 * options (whether irqs are enabled, scheduling 1598 * options, whether hard/softirqs are running, level of 1599 * preempt_disabled respectively). **N** means that 1600 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1601 * are set. 1602 * * ``419421.045894`` is a timestamp. 1603 * * ``0x00000001`` is a fake value used by BPF for the 1604 * instruction pointer register. 1605 * * ``<formatted msg>`` is the message formatted with 1606 * *fmt*. 1607 * 1608 * The conversion specifiers supported by *fmt* are similar, but 1609 * more limited than for printk(). They are **%d**, **%i**, 1610 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1611 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1612 * of field, padding with zeroes, etc.) is available, and the 1613 * helper will return **-EINVAL** (but print nothing) if it 1614 * encounters an unknown specifier. 1615 * 1616 * Also, note that **bpf_trace_printk**\ () is slow, and should 1617 * only be used for debugging purposes. For this reason, a notice 1618 * block (spanning several lines) is printed to kernel logs and 1619 * states that the helper should not be used "for production use" 1620 * the first time this helper is used (or more precisely, when 1621 * **trace_printk**\ () buffers are allocated). For passing values 1622 * to user space, perf events should be preferred. 1623 * Return 1624 * The number of bytes written to the buffer, or a negative error 1625 * in case of failure. 1626 * 1627 * u32 bpf_get_prandom_u32(void) 1628 * Description 1629 * Get a pseudo-random number. 1630 * 1631 * From a security point of view, this helper uses its own 1632 * pseudo-random internal state, and cannot be used to infer the 1633 * seed of other random functions in the kernel. However, it is 1634 * essential to note that the generator used by the helper is not 1635 * cryptographically secure. 1636 * Return 1637 * A random 32-bit unsigned value. 1638 * 1639 * u32 bpf_get_smp_processor_id(void) 1640 * Description 1641 * Get the SMP (symmetric multiprocessing) processor id. Note that 1642 * all programs run with migration disabled, which means that the 1643 * SMP processor id is stable during all the execution of the 1644 * program. 1645 * Return 1646 * The SMP id of the processor running the program. 1647 * 1648 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1649 * Description 1650 * Store *len* bytes from address *from* into the packet 1651 * associated to *skb*, at *offset*. *flags* are a combination of 1652 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1653 * checksum for the packet after storing the bytes) and 1654 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1655 * **->swhash** and *skb*\ **->l4hash** to 0). 1656 * 1657 * A call to this helper is susceptible to change the underlying 1658 * packet buffer. Therefore, at load time, all checks on pointers 1659 * previously done by the verifier are invalidated and must be 1660 * performed again, if the helper is used in combination with 1661 * direct packet access. 1662 * Return 1663 * 0 on success, or a negative error in case of failure. 1664 * 1665 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1666 * Description 1667 * Recompute the layer 3 (e.g. IP) checksum for the packet 1668 * associated to *skb*. Computation is incremental, so the helper 1669 * must know the former value of the header field that was 1670 * modified (*from*), the new value of this field (*to*), and the 1671 * number of bytes (2 or 4) for this field, stored in *size*. 1672 * Alternatively, it is possible to store the difference between 1673 * the previous and the new values of the header field in *to*, by 1674 * setting *from* and *size* to 0. For both methods, *offset* 1675 * indicates the location of the IP checksum within the packet. 1676 * 1677 * This helper works in combination with **bpf_csum_diff**\ (), 1678 * which does not update the checksum in-place, but offers more 1679 * flexibility and can handle sizes larger than 2 or 4 for the 1680 * checksum to update. 1681 * 1682 * A call to this helper is susceptible to change the underlying 1683 * packet buffer. Therefore, at load time, all checks on pointers 1684 * previously done by the verifier are invalidated and must be 1685 * performed again, if the helper is used in combination with 1686 * direct packet access. 1687 * Return 1688 * 0 on success, or a negative error in case of failure. 1689 * 1690 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1691 * Description 1692 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1693 * packet associated to *skb*. Computation is incremental, so the 1694 * helper must know the former value of the header field that was 1695 * modified (*from*), the new value of this field (*to*), and the 1696 * number of bytes (2 or 4) for this field, stored on the lowest 1697 * four bits of *flags*. Alternatively, it is possible to store 1698 * the difference between the previous and the new values of the 1699 * header field in *to*, by setting *from* and the four lowest 1700 * bits of *flags* to 0. For both methods, *offset* indicates the 1701 * location of the IP checksum within the packet. In addition to 1702 * the size of the field, *flags* can be added (bitwise OR) actual 1703 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1704 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1705 * for updates resulting in a null checksum the value is set to 1706 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1707 * the checksum is to be computed against a pseudo-header. 1708 * 1709 * This helper works in combination with **bpf_csum_diff**\ (), 1710 * which does not update the checksum in-place, but offers more 1711 * flexibility and can handle sizes larger than 2 or 4 for the 1712 * checksum to update. 1713 * 1714 * A call to this helper is susceptible to change the underlying 1715 * packet buffer. Therefore, at load time, all checks on pointers 1716 * previously done by the verifier are invalidated and must be 1717 * performed again, if the helper is used in combination with 1718 * direct packet access. 1719 * Return 1720 * 0 on success, or a negative error in case of failure. 1721 * 1722 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1723 * Description 1724 * This special helper is used to trigger a "tail call", or in 1725 * other words, to jump into another eBPF program. The same stack 1726 * frame is used (but values on stack and in registers for the 1727 * caller are not accessible to the callee). This mechanism allows 1728 * for program chaining, either for raising the maximum number of 1729 * available eBPF instructions, or to execute given programs in 1730 * conditional blocks. For security reasons, there is an upper 1731 * limit to the number of successive tail calls that can be 1732 * performed. 1733 * 1734 * Upon call of this helper, the program attempts to jump into a 1735 * program referenced at index *index* in *prog_array_map*, a 1736 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1737 * *ctx*, a pointer to the context. 1738 * 1739 * If the call succeeds, the kernel immediately runs the first 1740 * instruction of the new program. This is not a function call, 1741 * and it never returns to the previous program. If the call 1742 * fails, then the helper has no effect, and the caller continues 1743 * to run its subsequent instructions. A call can fail if the 1744 * destination program for the jump does not exist (i.e. *index* 1745 * is superior to the number of entries in *prog_array_map*), or 1746 * if the maximum number of tail calls has been reached for this 1747 * chain of programs. This limit is defined in the kernel by the 1748 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1749 * which is currently set to 33. 1750 * Return 1751 * 0 on success, or a negative error in case of failure. 1752 * 1753 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1754 * Description 1755 * Clone and redirect the packet associated to *skb* to another 1756 * net device of index *ifindex*. Both ingress and egress 1757 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1758 * value in *flags* is used to make the distinction (ingress path 1759 * is selected if the flag is present, egress path otherwise). 1760 * This is the only flag supported for now. 1761 * 1762 * In comparison with **bpf_redirect**\ () helper, 1763 * **bpf_clone_redirect**\ () has the associated cost of 1764 * duplicating the packet buffer, but this can be executed out of 1765 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1766 * efficient, but it is handled through an action code where the 1767 * redirection happens only after the eBPF program has returned. 1768 * 1769 * A call to this helper is susceptible to change the underlying 1770 * packet buffer. Therefore, at load time, all checks on pointers 1771 * previously done by the verifier are invalidated and must be 1772 * performed again, if the helper is used in combination with 1773 * direct packet access. 1774 * Return 1775 * 0 on success, or a negative error in case of failure. 1776 * 1777 * u64 bpf_get_current_pid_tgid(void) 1778 * Return 1779 * A 64-bit integer containing the current tgid and pid, and 1780 * created as such: 1781 * *current_task*\ **->tgid << 32 \|** 1782 * *current_task*\ **->pid**. 1783 * 1784 * u64 bpf_get_current_uid_gid(void) 1785 * Return 1786 * A 64-bit integer containing the current GID and UID, and 1787 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1788 * 1789 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1790 * Description 1791 * Copy the **comm** attribute of the current task into *buf* of 1792 * *size_of_buf*. The **comm** attribute contains the name of 1793 * the executable (excluding the path) for the current task. The 1794 * *size_of_buf* must be strictly positive. On success, the 1795 * helper makes sure that the *buf* is NUL-terminated. On failure, 1796 * it is filled with zeroes. 1797 * Return 1798 * 0 on success, or a negative error in case of failure. 1799 * 1800 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1801 * Description 1802 * Retrieve the classid for the current task, i.e. for the net_cls 1803 * cgroup to which *skb* belongs. 1804 * 1805 * This helper can be used on TC egress path, but not on ingress. 1806 * 1807 * The net_cls cgroup provides an interface to tag network packets 1808 * based on a user-provided identifier for all traffic coming from 1809 * the tasks belonging to the related cgroup. See also the related 1810 * kernel documentation, available from the Linux sources in file 1811 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1812 * 1813 * The Linux kernel has two versions for cgroups: there are 1814 * cgroups v1 and cgroups v2. Both are available to users, who can 1815 * use a mixture of them, but note that the net_cls cgroup is for 1816 * cgroup v1 only. This makes it incompatible with BPF programs 1817 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1818 * only hold data for one version of cgroups at a time). 1819 * 1820 * This helper is only available is the kernel was compiled with 1821 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1822 * "**y**" or to "**m**". 1823 * Return 1824 * The classid, or 0 for the default unconfigured classid. 1825 * 1826 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1827 * Description 1828 * Push a *vlan_tci* (VLAN tag control information) of protocol 1829 * *vlan_proto* to the packet associated to *skb*, then update 1830 * the checksum. Note that if *vlan_proto* is different from 1831 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1832 * be **ETH_P_8021Q**. 1833 * 1834 * A call to this helper is susceptible to change the underlying 1835 * packet buffer. Therefore, at load time, all checks on pointers 1836 * previously done by the verifier are invalidated and must be 1837 * performed again, if the helper is used in combination with 1838 * direct packet access. 1839 * Return 1840 * 0 on success, or a negative error in case of failure. 1841 * 1842 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1843 * Description 1844 * Pop a VLAN header from the packet associated to *skb*. 1845 * 1846 * A call to this helper is susceptible to change the underlying 1847 * packet buffer. Therefore, at load time, all checks on pointers 1848 * previously done by the verifier are invalidated and must be 1849 * performed again, if the helper is used in combination with 1850 * direct packet access. 1851 * Return 1852 * 0 on success, or a negative error in case of failure. 1853 * 1854 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1855 * Description 1856 * Get tunnel metadata. This helper takes a pointer *key* to an 1857 * empty **struct bpf_tunnel_key** of **size**, that will be 1858 * filled with tunnel metadata for the packet associated to *skb*. 1859 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1860 * indicates that the tunnel is based on IPv6 protocol instead of 1861 * IPv4. 1862 * 1863 * The **struct bpf_tunnel_key** is an object that generalizes the 1864 * principal parameters used by various tunneling protocols into a 1865 * single struct. This way, it can be used to easily make a 1866 * decision based on the contents of the encapsulation header, 1867 * "summarized" in this struct. In particular, it holds the IP 1868 * address of the remote end (IPv4 or IPv6, depending on the case) 1869 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1870 * this struct exposes the *key*\ **->tunnel_id**, which is 1871 * generally mapped to a VNI (Virtual Network Identifier), making 1872 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1873 * () helper. 1874 * 1875 * Let's imagine that the following code is part of a program 1876 * attached to the TC ingress interface, on one end of a GRE 1877 * tunnel, and is supposed to filter out all messages coming from 1878 * remote ends with IPv4 address other than 10.0.0.1: 1879 * 1880 * :: 1881 * 1882 * int ret; 1883 * struct bpf_tunnel_key key = {}; 1884 * 1885 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1886 * if (ret < 0) 1887 * return TC_ACT_SHOT; // drop packet 1888 * 1889 * if (key.remote_ipv4 != 0x0a000001) 1890 * return TC_ACT_SHOT; // drop packet 1891 * 1892 * return TC_ACT_OK; // accept packet 1893 * 1894 * This interface can also be used with all encapsulation devices 1895 * that can operate in "collect metadata" mode: instead of having 1896 * one network device per specific configuration, the "collect 1897 * metadata" mode only requires a single device where the 1898 * configuration can be extracted from this helper. 1899 * 1900 * This can be used together with various tunnels such as VXLan, 1901 * Geneve, GRE or IP in IP (IPIP). 1902 * Return 1903 * 0 on success, or a negative error in case of failure. 1904 * 1905 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1906 * Description 1907 * Populate tunnel metadata for packet associated to *skb.* The 1908 * tunnel metadata is set to the contents of *key*, of *size*. The 1909 * *flags* can be set to a combination of the following values: 1910 * 1911 * **BPF_F_TUNINFO_IPV6** 1912 * Indicate that the tunnel is based on IPv6 protocol 1913 * instead of IPv4. 1914 * **BPF_F_ZERO_CSUM_TX** 1915 * For IPv4 packets, add a flag to tunnel metadata 1916 * indicating that checksum computation should be skipped 1917 * and checksum set to zeroes. 1918 * **BPF_F_DONT_FRAGMENT** 1919 * Add a flag to tunnel metadata indicating that the 1920 * packet should not be fragmented. 1921 * **BPF_F_SEQ_NUMBER** 1922 * Add a flag to tunnel metadata indicating that a 1923 * sequence number should be added to tunnel header before 1924 * sending the packet. This flag was added for GRE 1925 * encapsulation, but might be used with other protocols 1926 * as well in the future. 1927 * 1928 * Here is a typical usage on the transmit path: 1929 * 1930 * :: 1931 * 1932 * struct bpf_tunnel_key key; 1933 * populate key ... 1934 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1935 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1936 * 1937 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1938 * helper for additional information. 1939 * Return 1940 * 0 on success, or a negative error in case of failure. 1941 * 1942 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1943 * Description 1944 * Read the value of a perf event counter. This helper relies on a 1945 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1946 * the perf event counter is selected when *map* is updated with 1947 * perf event file descriptors. The *map* is an array whose size 1948 * is the number of available CPUs, and each cell contains a value 1949 * relative to one CPU. The value to retrieve is indicated by 1950 * *flags*, that contains the index of the CPU to look up, masked 1951 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1952 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1953 * current CPU should be retrieved. 1954 * 1955 * Note that before Linux 4.13, only hardware perf event can be 1956 * retrieved. 1957 * 1958 * Also, be aware that the newer helper 1959 * **bpf_perf_event_read_value**\ () is recommended over 1960 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1961 * quirks where error and counter value are used as a return code 1962 * (which is wrong to do since ranges may overlap). This issue is 1963 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1964 * time provides more features over the **bpf_perf_event_read**\ 1965 * () interface. Please refer to the description of 1966 * **bpf_perf_event_read_value**\ () for details. 1967 * Return 1968 * The value of the perf event counter read from the map, or a 1969 * negative error code in case of failure. 1970 * 1971 * long bpf_redirect(u32 ifindex, u64 flags) 1972 * Description 1973 * Redirect the packet to another net device of index *ifindex*. 1974 * This helper is somewhat similar to **bpf_clone_redirect**\ 1975 * (), except that the packet is not cloned, which provides 1976 * increased performance. 1977 * 1978 * Except for XDP, both ingress and egress interfaces can be used 1979 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1980 * to make the distinction (ingress path is selected if the flag 1981 * is present, egress path otherwise). Currently, XDP only 1982 * supports redirection to the egress interface, and accepts no 1983 * flag at all. 1984 * 1985 * The same effect can also be attained with the more generic 1986 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1987 * redirect target instead of providing it directly to the helper. 1988 * Return 1989 * For XDP, the helper returns **XDP_REDIRECT** on success or 1990 * **XDP_ABORTED** on error. For other program types, the values 1991 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1992 * error. 1993 * 1994 * u32 bpf_get_route_realm(struct sk_buff *skb) 1995 * Description 1996 * Retrieve the realm or the route, that is to say the 1997 * **tclassid** field of the destination for the *skb*. The 1998 * identifier retrieved is a user-provided tag, similar to the 1999 * one used with the net_cls cgroup (see description for 2000 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2001 * held by a route (a destination entry), not by a task. 2002 * 2003 * Retrieving this identifier works with the clsact TC egress hook 2004 * (see also **tc-bpf(8)**), or alternatively on conventional 2005 * classful egress qdiscs, but not on TC ingress path. In case of 2006 * clsact TC egress hook, this has the advantage that, internally, 2007 * the destination entry has not been dropped yet in the transmit 2008 * path. Therefore, the destination entry does not need to be 2009 * artificially held via **netif_keep_dst**\ () for a classful 2010 * qdisc until the *skb* is freed. 2011 * 2012 * This helper is available only if the kernel was compiled with 2013 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2014 * Return 2015 * The realm of the route for the packet associated to *skb*, or 0 2016 * if none was found. 2017 * 2018 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2019 * Description 2020 * Write raw *data* blob into a special BPF perf event held by 2021 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2022 * event must have the following attributes: **PERF_SAMPLE_RAW** 2023 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2024 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2025 * 2026 * The *flags* are used to indicate the index in *map* for which 2027 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2028 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2029 * to indicate that the index of the current CPU core should be 2030 * used. 2031 * 2032 * The value to write, of *size*, is passed through eBPF stack and 2033 * pointed by *data*. 2034 * 2035 * The context of the program *ctx* needs also be passed to the 2036 * helper. 2037 * 2038 * On user space, a program willing to read the values needs to 2039 * call **perf_event_open**\ () on the perf event (either for 2040 * one or for all CPUs) and to store the file descriptor into the 2041 * *map*. This must be done before the eBPF program can send data 2042 * into it. An example is available in file 2043 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2044 * tree (the eBPF program counterpart is in 2045 * *samples/bpf/trace_output_kern.c*). 2046 * 2047 * **bpf_perf_event_output**\ () achieves better performance 2048 * than **bpf_trace_printk**\ () for sharing data with user 2049 * space, and is much better suitable for streaming data from eBPF 2050 * programs. 2051 * 2052 * Note that this helper is not restricted to tracing use cases 2053 * and can be used with programs attached to TC or XDP as well, 2054 * where it allows for passing data to user space listeners. Data 2055 * can be: 2056 * 2057 * * Only custom structs, 2058 * * Only the packet payload, or 2059 * * A combination of both. 2060 * Return 2061 * 0 on success, or a negative error in case of failure. 2062 * 2063 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2064 * Description 2065 * This helper was provided as an easy way to load data from a 2066 * packet. It can be used to load *len* bytes from *offset* from 2067 * the packet associated to *skb*, into the buffer pointed by 2068 * *to*. 2069 * 2070 * Since Linux 4.7, usage of this helper has mostly been replaced 2071 * by "direct packet access", enabling packet data to be 2072 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2073 * pointing respectively to the first byte of packet data and to 2074 * the byte after the last byte of packet data. However, it 2075 * remains useful if one wishes to read large quantities of data 2076 * at once from a packet into the eBPF stack. 2077 * Return 2078 * 0 on success, or a negative error in case of failure. 2079 * 2080 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2081 * Description 2082 * Walk a user or a kernel stack and return its id. To achieve 2083 * this, the helper needs *ctx*, which is a pointer to the context 2084 * on which the tracing program is executed, and a pointer to a 2085 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2086 * 2087 * The last argument, *flags*, holds the number of stack frames to 2088 * skip (from 0 to 255), masked with 2089 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2090 * a combination of the following flags: 2091 * 2092 * **BPF_F_USER_STACK** 2093 * Collect a user space stack instead of a kernel stack. 2094 * **BPF_F_FAST_STACK_CMP** 2095 * Compare stacks by hash only. 2096 * **BPF_F_REUSE_STACKID** 2097 * If two different stacks hash into the same *stackid*, 2098 * discard the old one. 2099 * 2100 * The stack id retrieved is a 32 bit long integer handle which 2101 * can be further combined with other data (including other stack 2102 * ids) and used as a key into maps. This can be useful for 2103 * generating a variety of graphs (such as flame graphs or off-cpu 2104 * graphs). 2105 * 2106 * For walking a stack, this helper is an improvement over 2107 * **bpf_probe_read**\ (), which can be used with unrolled loops 2108 * but is not efficient and consumes a lot of eBPF instructions. 2109 * Instead, **bpf_get_stackid**\ () can collect up to 2110 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2111 * this limit can be controlled with the **sysctl** program, and 2112 * that it should be manually increased in order to profile long 2113 * user stacks (such as stacks for Java programs). To do so, use: 2114 * 2115 * :: 2116 * 2117 * # sysctl kernel.perf_event_max_stack=<new value> 2118 * Return 2119 * The positive or null stack id on success, or a negative error 2120 * in case of failure. 2121 * 2122 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2123 * Description 2124 * Compute a checksum difference, from the raw buffer pointed by 2125 * *from*, of length *from_size* (that must be a multiple of 4), 2126 * towards the raw buffer pointed by *to*, of size *to_size* 2127 * (same remark). An optional *seed* can be added to the value 2128 * (this can be cascaded, the seed may come from a previous call 2129 * to the helper). 2130 * 2131 * This is flexible enough to be used in several ways: 2132 * 2133 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2134 * checksum, it can be used when pushing new data. 2135 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2136 * checksum, it can be used when removing data from a packet. 2137 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2138 * can be used to compute a diff. Note that *from_size* and 2139 * *to_size* do not need to be equal. 2140 * 2141 * This helper can be used in combination with 2142 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2143 * which one can feed in the difference computed with 2144 * **bpf_csum_diff**\ (). 2145 * Return 2146 * The checksum result, or a negative error code in case of 2147 * failure. 2148 * 2149 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2150 * Description 2151 * Retrieve tunnel options metadata for the packet associated to 2152 * *skb*, and store the raw tunnel option data to the buffer *opt* 2153 * of *size*. 2154 * 2155 * This helper can be used with encapsulation devices that can 2156 * operate in "collect metadata" mode (please refer to the related 2157 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2158 * more details). A particular example where this can be used is 2159 * in combination with the Geneve encapsulation protocol, where it 2160 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2161 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2162 * the eBPF program. This allows for full customization of these 2163 * headers. 2164 * Return 2165 * The size of the option data retrieved. 2166 * 2167 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2168 * Description 2169 * Set tunnel options metadata for the packet associated to *skb* 2170 * to the option data contained in the raw buffer *opt* of *size*. 2171 * 2172 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2173 * helper for additional information. 2174 * Return 2175 * 0 on success, or a negative error in case of failure. 2176 * 2177 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2178 * Description 2179 * Change the protocol of the *skb* to *proto*. Currently 2180 * supported are transition from IPv4 to IPv6, and from IPv6 to 2181 * IPv4. The helper takes care of the groundwork for the 2182 * transition, including resizing the socket buffer. The eBPF 2183 * program is expected to fill the new headers, if any, via 2184 * **skb_store_bytes**\ () and to recompute the checksums with 2185 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2186 * (). The main case for this helper is to perform NAT64 2187 * operations out of an eBPF program. 2188 * 2189 * Internally, the GSO type is marked as dodgy so that headers are 2190 * checked and segments are recalculated by the GSO/GRO engine. 2191 * The size for GSO target is adapted as well. 2192 * 2193 * All values for *flags* are reserved for future usage, and must 2194 * be left at zero. 2195 * 2196 * A call to this helper is susceptible to change the underlying 2197 * packet buffer. Therefore, at load time, all checks on pointers 2198 * previously done by the verifier are invalidated and must be 2199 * performed again, if the helper is used in combination with 2200 * direct packet access. 2201 * Return 2202 * 0 on success, or a negative error in case of failure. 2203 * 2204 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2205 * Description 2206 * Change the packet type for the packet associated to *skb*. This 2207 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2208 * the eBPF program does not have a write access to *skb*\ 2209 * **->pkt_type** beside this helper. Using a helper here allows 2210 * for graceful handling of errors. 2211 * 2212 * The major use case is to change incoming *skb*s to 2213 * **PACKET_HOST** in a programmatic way instead of having to 2214 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2215 * example. 2216 * 2217 * Note that *type* only allows certain values. At this time, they 2218 * are: 2219 * 2220 * **PACKET_HOST** 2221 * Packet is for us. 2222 * **PACKET_BROADCAST** 2223 * Send packet to all. 2224 * **PACKET_MULTICAST** 2225 * Send packet to group. 2226 * **PACKET_OTHERHOST** 2227 * Send packet to someone else. 2228 * Return 2229 * 0 on success, or a negative error in case of failure. 2230 * 2231 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2232 * Description 2233 * Check whether *skb* is a descendant of the cgroup2 held by 2234 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2235 * Return 2236 * The return value depends on the result of the test, and can be: 2237 * 2238 * * 0, if the *skb* failed the cgroup2 descendant test. 2239 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2240 * * A negative error code, if an error occurred. 2241 * 2242 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2243 * Description 2244 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2245 * not set, in particular if the hash was cleared due to mangling, 2246 * recompute this hash. Later accesses to the hash can be done 2247 * directly with *skb*\ **->hash**. 2248 * 2249 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2250 * prototype with **bpf_skb_change_proto**\ (), or calling 2251 * **bpf_skb_store_bytes**\ () with the 2252 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2253 * the hash and to trigger a new computation for the next call to 2254 * **bpf_get_hash_recalc**\ (). 2255 * Return 2256 * The 32-bit hash. 2257 * 2258 * u64 bpf_get_current_task(void) 2259 * Return 2260 * A pointer to the current task struct. 2261 * 2262 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2263 * Description 2264 * Attempt in a safe way to write *len* bytes from the buffer 2265 * *src* to *dst* in memory. It only works for threads that are in 2266 * user context, and *dst* must be a valid user space address. 2267 * 2268 * This helper should not be used to implement any kind of 2269 * security mechanism because of TOC-TOU attacks, but rather to 2270 * debug, divert, and manipulate execution of semi-cooperative 2271 * processes. 2272 * 2273 * Keep in mind that this feature is meant for experiments, and it 2274 * has a risk of crashing the system and running programs. 2275 * Therefore, when an eBPF program using this helper is attached, 2276 * a warning including PID and process name is printed to kernel 2277 * logs. 2278 * Return 2279 * 0 on success, or a negative error in case of failure. 2280 * 2281 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2282 * Description 2283 * Check whether the probe is being run is the context of a given 2284 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2285 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2286 * Return 2287 * The return value depends on the result of the test, and can be: 2288 * 2289 * * 0, if current task belongs to the cgroup2. 2290 * * 1, if current task does not belong to the cgroup2. 2291 * * A negative error code, if an error occurred. 2292 * 2293 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2294 * Description 2295 * Resize (trim or grow) the packet associated to *skb* to the 2296 * new *len*. The *flags* are reserved for future usage, and must 2297 * be left at zero. 2298 * 2299 * The basic idea is that the helper performs the needed work to 2300 * change the size of the packet, then the eBPF program rewrites 2301 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2302 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2303 * and others. This helper is a slow path utility intended for 2304 * replies with control messages. And because it is targeted for 2305 * slow path, the helper itself can afford to be slow: it 2306 * implicitly linearizes, unclones and drops offloads from the 2307 * *skb*. 2308 * 2309 * A call to this helper is susceptible to change the underlying 2310 * packet buffer. Therefore, at load time, all checks on pointers 2311 * previously done by the verifier are invalidated and must be 2312 * performed again, if the helper is used in combination with 2313 * direct packet access. 2314 * Return 2315 * 0 on success, or a negative error in case of failure. 2316 * 2317 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2318 * Description 2319 * Pull in non-linear data in case the *skb* is non-linear and not 2320 * all of *len* are part of the linear section. Make *len* bytes 2321 * from *skb* readable and writable. If a zero value is passed for 2322 * *len*, then the whole length of the *skb* is pulled. 2323 * 2324 * This helper is only needed for reading and writing with direct 2325 * packet access. 2326 * 2327 * For direct packet access, testing that offsets to access 2328 * are within packet boundaries (test on *skb*\ **->data_end**) is 2329 * susceptible to fail if offsets are invalid, or if the requested 2330 * data is in non-linear parts of the *skb*. On failure the 2331 * program can just bail out, or in the case of a non-linear 2332 * buffer, use a helper to make the data available. The 2333 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2334 * the data. Another one consists in using **bpf_skb_pull_data** 2335 * to pull in once the non-linear parts, then retesting and 2336 * eventually access the data. 2337 * 2338 * At the same time, this also makes sure the *skb* is uncloned, 2339 * which is a necessary condition for direct write. As this needs 2340 * to be an invariant for the write part only, the verifier 2341 * detects writes and adds a prologue that is calling 2342 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2343 * the very beginning in case it is indeed cloned. 2344 * 2345 * A call to this helper is susceptible to change the underlying 2346 * packet buffer. Therefore, at load time, all checks on pointers 2347 * previously done by the verifier are invalidated and must be 2348 * performed again, if the helper is used in combination with 2349 * direct packet access. 2350 * Return 2351 * 0 on success, or a negative error in case of failure. 2352 * 2353 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2354 * Description 2355 * Add the checksum *csum* into *skb*\ **->csum** in case the 2356 * driver has supplied a checksum for the entire packet into that 2357 * field. Return an error otherwise. This helper is intended to be 2358 * used in combination with **bpf_csum_diff**\ (), in particular 2359 * when the checksum needs to be updated after data has been 2360 * written into the packet through direct packet access. 2361 * Return 2362 * The checksum on success, or a negative error code in case of 2363 * failure. 2364 * 2365 * void bpf_set_hash_invalid(struct sk_buff *skb) 2366 * Description 2367 * Invalidate the current *skb*\ **->hash**. It can be used after 2368 * mangling on headers through direct packet access, in order to 2369 * indicate that the hash is outdated and to trigger a 2370 * recalculation the next time the kernel tries to access this 2371 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2372 * 2373 * long bpf_get_numa_node_id(void) 2374 * Description 2375 * Return the id of the current NUMA node. The primary use case 2376 * for this helper is the selection of sockets for the local NUMA 2377 * node, when the program is attached to sockets using the 2378 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2379 * but the helper is also available to other eBPF program types, 2380 * similarly to **bpf_get_smp_processor_id**\ (). 2381 * Return 2382 * The id of current NUMA node. 2383 * 2384 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2385 * Description 2386 * Grows headroom of packet associated to *skb* and adjusts the 2387 * offset of the MAC header accordingly, adding *len* bytes of 2388 * space. It automatically extends and reallocates memory as 2389 * required. 2390 * 2391 * This helper can be used on a layer 3 *skb* to push a MAC header 2392 * for redirection into a layer 2 device. 2393 * 2394 * All values for *flags* are reserved for future usage, and must 2395 * be left at zero. 2396 * 2397 * A call to this helper is susceptible to change the underlying 2398 * packet buffer. Therefore, at load time, all checks on pointers 2399 * previously done by the verifier are invalidated and must be 2400 * performed again, if the helper is used in combination with 2401 * direct packet access. 2402 * Return 2403 * 0 on success, or a negative error in case of failure. 2404 * 2405 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2406 * Description 2407 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2408 * it is possible to use a negative value for *delta*. This helper 2409 * can be used to prepare the packet for pushing or popping 2410 * headers. 2411 * 2412 * A call to this helper is susceptible to change the underlying 2413 * packet buffer. Therefore, at load time, all checks on pointers 2414 * previously done by the verifier are invalidated and must be 2415 * performed again, if the helper is used in combination with 2416 * direct packet access. 2417 * Return 2418 * 0 on success, or a negative error in case of failure. 2419 * 2420 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2421 * Description 2422 * Copy a NUL terminated string from an unsafe kernel address 2423 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2424 * more details. 2425 * 2426 * Generally, use **bpf_probe_read_user_str**\ () or 2427 * **bpf_probe_read_kernel_str**\ () instead. 2428 * Return 2429 * On success, the strictly positive length of the string, 2430 * including the trailing NUL character. On error, a negative 2431 * value. 2432 * 2433 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2434 * Description 2435 * If the **struct sk_buff** pointed by *skb* has a known socket, 2436 * retrieve the cookie (generated by the kernel) of this socket. 2437 * If no cookie has been set yet, generate a new cookie. Once 2438 * generated, the socket cookie remains stable for the life of the 2439 * socket. This helper can be useful for monitoring per socket 2440 * networking traffic statistics as it provides a global socket 2441 * identifier that can be assumed unique. 2442 * Return 2443 * A 8-byte long unique number on success, or 0 if the socket 2444 * field is missing inside *skb*. 2445 * 2446 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2447 * Description 2448 * Equivalent to bpf_get_socket_cookie() helper that accepts 2449 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2450 * Return 2451 * A 8-byte long unique number. 2452 * 2453 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2454 * Description 2455 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2456 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2457 * Return 2458 * A 8-byte long unique number. 2459 * 2460 * u64 bpf_get_socket_cookie(struct sock *sk) 2461 * Description 2462 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2463 * *sk*, but gets socket from a BTF **struct sock**. This helper 2464 * also works for sleepable programs. 2465 * Return 2466 * A 8-byte long unique number or 0 if *sk* is NULL. 2467 * 2468 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2469 * Return 2470 * The owner UID of the socket associated to *skb*. If the socket 2471 * is **NULL**, or if it is not a full socket (i.e. if it is a 2472 * time-wait or a request socket instead), **overflowuid** value 2473 * is returned (note that **overflowuid** might also be the actual 2474 * UID value for the socket). 2475 * 2476 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2477 * Description 2478 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2479 * to value *hash*. 2480 * Return 2481 * 0 2482 * 2483 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2484 * Description 2485 * Emulate a call to **setsockopt()** on the socket associated to 2486 * *bpf_socket*, which must be a full socket. The *level* at 2487 * which the option resides and the name *optname* of the option 2488 * must be specified, see **setsockopt(2)** for more information. 2489 * The option value of length *optlen* is pointed by *optval*. 2490 * 2491 * *bpf_socket* should be one of the following: 2492 * 2493 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2494 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2495 * and **BPF_CGROUP_INET6_CONNECT**. 2496 * 2497 * This helper actually implements a subset of **setsockopt()**. 2498 * It supports the following *level*\ s: 2499 * 2500 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2501 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2502 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2503 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 2504 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2505 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2506 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2507 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2508 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 2509 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2510 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2511 * Return 2512 * 0 on success, or a negative error in case of failure. 2513 * 2514 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2515 * Description 2516 * Grow or shrink the room for data in the packet associated to 2517 * *skb* by *len_diff*, and according to the selected *mode*. 2518 * 2519 * By default, the helper will reset any offloaded checksum 2520 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2521 * by the following flag: 2522 * 2523 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2524 * checksum data of the skb to CHECKSUM_NONE. 2525 * 2526 * There are two supported modes at this time: 2527 * 2528 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2529 * (room space is added or removed below the layer 2 header). 2530 * 2531 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2532 * (room space is added or removed below the layer 3 header). 2533 * 2534 * The following flags are supported at this time: 2535 * 2536 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2537 * Adjusting mss in this way is not allowed for datagrams. 2538 * 2539 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2540 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2541 * Any new space is reserved to hold a tunnel header. 2542 * Configure skb offsets and other fields accordingly. 2543 * 2544 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2545 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2546 * Use with ENCAP_L3 flags to further specify the tunnel type. 2547 * 2548 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2549 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2550 * type; *len* is the length of the inner MAC header. 2551 * 2552 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2553 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2554 * L2 type as Ethernet. 2555 * 2556 * A call to this helper is susceptible to change the underlying 2557 * packet buffer. Therefore, at load time, all checks on pointers 2558 * previously done by the verifier are invalidated and must be 2559 * performed again, if the helper is used in combination with 2560 * direct packet access. 2561 * Return 2562 * 0 on success, or a negative error in case of failure. 2563 * 2564 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2565 * Description 2566 * Redirect the packet to the endpoint referenced by *map* at 2567 * index *key*. Depending on its type, this *map* can contain 2568 * references to net devices (for forwarding packets through other 2569 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2570 * but this is only implemented for native XDP (with driver 2571 * support) as of this writing). 2572 * 2573 * The lower two bits of *flags* are used as the return code if 2574 * the map lookup fails. This is so that the return value can be 2575 * one of the XDP program return codes up to **XDP_TX**, as chosen 2576 * by the caller. The higher bits of *flags* can be set to 2577 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2578 * 2579 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2580 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2581 * interface will be excluded when do broadcasting. 2582 * 2583 * See also **bpf_redirect**\ (), which only supports redirecting 2584 * to an ifindex, but doesn't require a map to do so. 2585 * Return 2586 * **XDP_REDIRECT** on success, or the value of the two lower bits 2587 * of the *flags* argument on error. 2588 * 2589 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2590 * Description 2591 * Redirect the packet to the socket referenced by *map* (of type 2592 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2593 * egress interfaces can be used for redirection. The 2594 * **BPF_F_INGRESS** value in *flags* is used to make the 2595 * distinction (ingress path is selected if the flag is present, 2596 * egress path otherwise). This is the only flag supported for now. 2597 * Return 2598 * **SK_PASS** on success, or **SK_DROP** on error. 2599 * 2600 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2601 * Description 2602 * Add an entry to, or update a *map* referencing sockets. The 2603 * *skops* is used as a new value for the entry associated to 2604 * *key*. *flags* is one of: 2605 * 2606 * **BPF_NOEXIST** 2607 * The entry for *key* must not exist in the map. 2608 * **BPF_EXIST** 2609 * The entry for *key* must already exist in the map. 2610 * **BPF_ANY** 2611 * No condition on the existence of the entry for *key*. 2612 * 2613 * If the *map* has eBPF programs (parser and verdict), those will 2614 * be inherited by the socket being added. If the socket is 2615 * already attached to eBPF programs, this results in an error. 2616 * Return 2617 * 0 on success, or a negative error in case of failure. 2618 * 2619 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2620 * Description 2621 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2622 * *delta* (which can be positive or negative). Note that this 2623 * operation modifies the address stored in *xdp_md*\ **->data**, 2624 * so the latter must be loaded only after the helper has been 2625 * called. 2626 * 2627 * The use of *xdp_md*\ **->data_meta** is optional and programs 2628 * are not required to use it. The rationale is that when the 2629 * packet is processed with XDP (e.g. as DoS filter), it is 2630 * possible to push further meta data along with it before passing 2631 * to the stack, and to give the guarantee that an ingress eBPF 2632 * program attached as a TC classifier on the same device can pick 2633 * this up for further post-processing. Since TC works with socket 2634 * buffers, it remains possible to set from XDP the **mark** or 2635 * **priority** pointers, or other pointers for the socket buffer. 2636 * Having this scratch space generic and programmable allows for 2637 * more flexibility as the user is free to store whatever meta 2638 * data they need. 2639 * 2640 * A call to this helper is susceptible to change the underlying 2641 * packet buffer. Therefore, at load time, all checks on pointers 2642 * previously done by the verifier are invalidated and must be 2643 * performed again, if the helper is used in combination with 2644 * direct packet access. 2645 * Return 2646 * 0 on success, or a negative error in case of failure. 2647 * 2648 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2649 * Description 2650 * Read the value of a perf event counter, and store it into *buf* 2651 * of size *buf_size*. This helper relies on a *map* of type 2652 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2653 * counter is selected when *map* is updated with perf event file 2654 * descriptors. The *map* is an array whose size is the number of 2655 * available CPUs, and each cell contains a value relative to one 2656 * CPU. The value to retrieve is indicated by *flags*, that 2657 * contains the index of the CPU to look up, masked with 2658 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2659 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2660 * current CPU should be retrieved. 2661 * 2662 * This helper behaves in a way close to 2663 * **bpf_perf_event_read**\ () helper, save that instead of 2664 * just returning the value observed, it fills the *buf* 2665 * structure. This allows for additional data to be retrieved: in 2666 * particular, the enabled and running times (in *buf*\ 2667 * **->enabled** and *buf*\ **->running**, respectively) are 2668 * copied. In general, **bpf_perf_event_read_value**\ () is 2669 * recommended over **bpf_perf_event_read**\ (), which has some 2670 * ABI issues and provides fewer functionalities. 2671 * 2672 * These values are interesting, because hardware PMU (Performance 2673 * Monitoring Unit) counters are limited resources. When there are 2674 * more PMU based perf events opened than available counters, 2675 * kernel will multiplex these events so each event gets certain 2676 * percentage (but not all) of the PMU time. In case that 2677 * multiplexing happens, the number of samples or counter value 2678 * will not reflect the case compared to when no multiplexing 2679 * occurs. This makes comparison between different runs difficult. 2680 * Typically, the counter value should be normalized before 2681 * comparing to other experiments. The usual normalization is done 2682 * as follows. 2683 * 2684 * :: 2685 * 2686 * normalized_counter = counter * t_enabled / t_running 2687 * 2688 * Where t_enabled is the time enabled for event and t_running is 2689 * the time running for event since last normalization. The 2690 * enabled and running times are accumulated since the perf event 2691 * open. To achieve scaling factor between two invocations of an 2692 * eBPF program, users can use CPU id as the key (which is 2693 * typical for perf array usage model) to remember the previous 2694 * value and do the calculation inside the eBPF program. 2695 * Return 2696 * 0 on success, or a negative error in case of failure. 2697 * 2698 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2699 * Description 2700 * For en eBPF program attached to a perf event, retrieve the 2701 * value of the event counter associated to *ctx* and store it in 2702 * the structure pointed by *buf* and of size *buf_size*. Enabled 2703 * and running times are also stored in the structure (see 2704 * description of helper **bpf_perf_event_read_value**\ () for 2705 * more details). 2706 * Return 2707 * 0 on success, or a negative error in case of failure. 2708 * 2709 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2710 * Description 2711 * Emulate a call to **getsockopt()** on the socket associated to 2712 * *bpf_socket*, which must be a full socket. The *level* at 2713 * which the option resides and the name *optname* of the option 2714 * must be specified, see **getsockopt(2)** for more information. 2715 * The retrieved value is stored in the structure pointed by 2716 * *opval* and of length *optlen*. 2717 * 2718 * *bpf_socket* should be one of the following: 2719 * 2720 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2721 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2722 * and **BPF_CGROUP_INET6_CONNECT**. 2723 * 2724 * This helper actually implements a subset of **getsockopt()**. 2725 * It supports the following *level*\ s: 2726 * 2727 * * **IPPROTO_TCP**, which supports *optname* 2728 * **TCP_CONGESTION**. 2729 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2730 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2731 * Return 2732 * 0 on success, or a negative error in case of failure. 2733 * 2734 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2735 * Description 2736 * Used for error injection, this helper uses kprobes to override 2737 * the return value of the probed function, and to set it to *rc*. 2738 * The first argument is the context *regs* on which the kprobe 2739 * works. 2740 * 2741 * This helper works by setting the PC (program counter) 2742 * to an override function which is run in place of the original 2743 * probed function. This means the probed function is not run at 2744 * all. The replacement function just returns with the required 2745 * value. 2746 * 2747 * This helper has security implications, and thus is subject to 2748 * restrictions. It is only available if the kernel was compiled 2749 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2750 * option, and in this case it only works on functions tagged with 2751 * **ALLOW_ERROR_INJECTION** in the kernel code. 2752 * 2753 * Also, the helper is only available for the architectures having 2754 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2755 * x86 architecture is the only one to support this feature. 2756 * Return 2757 * 0 2758 * 2759 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2760 * Description 2761 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2762 * for the full TCP socket associated to *bpf_sock_ops* to 2763 * *argval*. 2764 * 2765 * The primary use of this field is to determine if there should 2766 * be calls to eBPF programs of type 2767 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2768 * code. A program of the same type can change its value, per 2769 * connection and as necessary, when the connection is 2770 * established. This field is directly accessible for reading, but 2771 * this helper must be used for updates in order to return an 2772 * error if an eBPF program tries to set a callback that is not 2773 * supported in the current kernel. 2774 * 2775 * *argval* is a flag array which can combine these flags: 2776 * 2777 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2778 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2779 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2780 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2781 * 2782 * Therefore, this function can be used to clear a callback flag by 2783 * setting the appropriate bit to zero. e.g. to disable the RTO 2784 * callback: 2785 * 2786 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2787 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2788 * 2789 * Here are some examples of where one could call such eBPF 2790 * program: 2791 * 2792 * * When RTO fires. 2793 * * When a packet is retransmitted. 2794 * * When the connection terminates. 2795 * * When a packet is sent. 2796 * * When a packet is received. 2797 * Return 2798 * Code **-EINVAL** if the socket is not a full TCP socket; 2799 * otherwise, a positive number containing the bits that could not 2800 * be set is returned (which comes down to 0 if all bits were set 2801 * as required). 2802 * 2803 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2804 * Description 2805 * This helper is used in programs implementing policies at the 2806 * socket level. If the message *msg* is allowed to pass (i.e. if 2807 * the verdict eBPF program returns **SK_PASS**), redirect it to 2808 * the socket referenced by *map* (of type 2809 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2810 * egress interfaces can be used for redirection. The 2811 * **BPF_F_INGRESS** value in *flags* is used to make the 2812 * distinction (ingress path is selected if the flag is present, 2813 * egress path otherwise). This is the only flag supported for now. 2814 * Return 2815 * **SK_PASS** on success, or **SK_DROP** on error. 2816 * 2817 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2818 * Description 2819 * For socket policies, apply the verdict of the eBPF program to 2820 * the next *bytes* (number of bytes) of message *msg*. 2821 * 2822 * For example, this helper can be used in the following cases: 2823 * 2824 * * A single **sendmsg**\ () or **sendfile**\ () system call 2825 * contains multiple logical messages that the eBPF program is 2826 * supposed to read and for which it should apply a verdict. 2827 * * An eBPF program only cares to read the first *bytes* of a 2828 * *msg*. If the message has a large payload, then setting up 2829 * and calling the eBPF program repeatedly for all bytes, even 2830 * though the verdict is already known, would create unnecessary 2831 * overhead. 2832 * 2833 * When called from within an eBPF program, the helper sets a 2834 * counter internal to the BPF infrastructure, that is used to 2835 * apply the last verdict to the next *bytes*. If *bytes* is 2836 * smaller than the current data being processed from a 2837 * **sendmsg**\ () or **sendfile**\ () system call, the first 2838 * *bytes* will be sent and the eBPF program will be re-run with 2839 * the pointer for start of data pointing to byte number *bytes* 2840 * **+ 1**. If *bytes* is larger than the current data being 2841 * processed, then the eBPF verdict will be applied to multiple 2842 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2843 * consumed. 2844 * 2845 * Note that if a socket closes with the internal counter holding 2846 * a non-zero value, this is not a problem because data is not 2847 * being buffered for *bytes* and is sent as it is received. 2848 * Return 2849 * 0 2850 * 2851 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2852 * Description 2853 * For socket policies, prevent the execution of the verdict eBPF 2854 * program for message *msg* until *bytes* (byte number) have been 2855 * accumulated. 2856 * 2857 * This can be used when one needs a specific number of bytes 2858 * before a verdict can be assigned, even if the data spans 2859 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2860 * case would be a user calling **sendmsg**\ () repeatedly with 2861 * 1-byte long message segments. Obviously, this is bad for 2862 * performance, but it is still valid. If the eBPF program needs 2863 * *bytes* bytes to validate a header, this helper can be used to 2864 * prevent the eBPF program to be called again until *bytes* have 2865 * been accumulated. 2866 * Return 2867 * 0 2868 * 2869 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2870 * Description 2871 * For socket policies, pull in non-linear data from user space 2872 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2873 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2874 * respectively. 2875 * 2876 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2877 * *msg* it can only parse data that the (**data**, **data_end**) 2878 * pointers have already consumed. For **sendmsg**\ () hooks this 2879 * is likely the first scatterlist element. But for calls relying 2880 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2881 * be the range (**0**, **0**) because the data is shared with 2882 * user space and by default the objective is to avoid allowing 2883 * user space to modify data while (or after) eBPF verdict is 2884 * being decided. This helper can be used to pull in data and to 2885 * set the start and end pointer to given values. Data will be 2886 * copied if necessary (i.e. if data was not linear and if start 2887 * and end pointers do not point to the same chunk). 2888 * 2889 * A call to this helper is susceptible to change the underlying 2890 * packet buffer. Therefore, at load time, all checks on pointers 2891 * previously done by the verifier are invalidated and must be 2892 * performed again, if the helper is used in combination with 2893 * direct packet access. 2894 * 2895 * All values for *flags* are reserved for future usage, and must 2896 * be left at zero. 2897 * Return 2898 * 0 on success, or a negative error in case of failure. 2899 * 2900 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2901 * Description 2902 * Bind the socket associated to *ctx* to the address pointed by 2903 * *addr*, of length *addr_len*. This allows for making outgoing 2904 * connection from the desired IP address, which can be useful for 2905 * example when all processes inside a cgroup should use one 2906 * single IP address on a host that has multiple IP configured. 2907 * 2908 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2909 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2910 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2911 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2912 * behavior and lets the kernel efficiently pick up an unused 2913 * port as long as 4-tuple is unique. Passing non-zero port might 2914 * lead to degraded performance. 2915 * Return 2916 * 0 on success, or a negative error in case of failure. 2917 * 2918 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2919 * Description 2920 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2921 * possible to both shrink and grow the packet tail. 2922 * Shrink done via *delta* being a negative integer. 2923 * 2924 * A call to this helper is susceptible to change the underlying 2925 * packet buffer. Therefore, at load time, all checks on pointers 2926 * previously done by the verifier are invalidated and must be 2927 * performed again, if the helper is used in combination with 2928 * direct packet access. 2929 * Return 2930 * 0 on success, or a negative error in case of failure. 2931 * 2932 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2933 * Description 2934 * Retrieve the XFRM state (IP transform framework, see also 2935 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2936 * 2937 * The retrieved value is stored in the **struct bpf_xfrm_state** 2938 * pointed by *xfrm_state* and of length *size*. 2939 * 2940 * All values for *flags* are reserved for future usage, and must 2941 * be left at zero. 2942 * 2943 * This helper is available only if the kernel was compiled with 2944 * **CONFIG_XFRM** configuration option. 2945 * Return 2946 * 0 on success, or a negative error in case of failure. 2947 * 2948 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2949 * Description 2950 * Return a user or a kernel stack in bpf program provided buffer. 2951 * To achieve this, the helper needs *ctx*, which is a pointer 2952 * to the context on which the tracing program is executed. 2953 * To store the stacktrace, the bpf program provides *buf* with 2954 * a nonnegative *size*. 2955 * 2956 * The last argument, *flags*, holds the number of stack frames to 2957 * skip (from 0 to 255), masked with 2958 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2959 * the following flags: 2960 * 2961 * **BPF_F_USER_STACK** 2962 * Collect a user space stack instead of a kernel stack. 2963 * **BPF_F_USER_BUILD_ID** 2964 * Collect buildid+offset instead of ips for user stack, 2965 * only valid if **BPF_F_USER_STACK** is also specified. 2966 * 2967 * **bpf_get_stack**\ () can collect up to 2968 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2969 * to sufficient large buffer size. Note that 2970 * this limit can be controlled with the **sysctl** program, and 2971 * that it should be manually increased in order to profile long 2972 * user stacks (such as stacks for Java programs). To do so, use: 2973 * 2974 * :: 2975 * 2976 * # sysctl kernel.perf_event_max_stack=<new value> 2977 * Return 2978 * A non-negative value equal to or less than *size* on success, 2979 * or a negative error in case of failure. 2980 * 2981 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2982 * Description 2983 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2984 * it provides an easy way to load *len* bytes from *offset* 2985 * from the packet associated to *skb*, into the buffer pointed 2986 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2987 * a fifth argument *start_header* exists in order to select a 2988 * base offset to start from. *start_header* can be one of: 2989 * 2990 * **BPF_HDR_START_MAC** 2991 * Base offset to load data from is *skb*'s mac header. 2992 * **BPF_HDR_START_NET** 2993 * Base offset to load data from is *skb*'s network header. 2994 * 2995 * In general, "direct packet access" is the preferred method to 2996 * access packet data, however, this helper is in particular useful 2997 * in socket filters where *skb*\ **->data** does not always point 2998 * to the start of the mac header and where "direct packet access" 2999 * is not available. 3000 * Return 3001 * 0 on success, or a negative error in case of failure. 3002 * 3003 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3004 * Description 3005 * Do FIB lookup in kernel tables using parameters in *params*. 3006 * If lookup is successful and result shows packet is to be 3007 * forwarded, the neighbor tables are searched for the nexthop. 3008 * If successful (ie., FIB lookup shows forwarding and nexthop 3009 * is resolved), the nexthop address is returned in ipv4_dst 3010 * or ipv6_dst based on family, smac is set to mac address of 3011 * egress device, dmac is set to nexthop mac address, rt_metric 3012 * is set to metric from route (IPv4/IPv6 only), and ifindex 3013 * is set to the device index of the nexthop from the FIB lookup. 3014 * 3015 * *plen* argument is the size of the passed in struct. 3016 * *flags* argument can be a combination of one or more of the 3017 * following values: 3018 * 3019 * **BPF_FIB_LOOKUP_DIRECT** 3020 * Do a direct table lookup vs full lookup using FIB 3021 * rules. 3022 * **BPF_FIB_LOOKUP_OUTPUT** 3023 * Perform lookup from an egress perspective (default is 3024 * ingress). 3025 * 3026 * *ctx* is either **struct xdp_md** for XDP programs or 3027 * **struct sk_buff** tc cls_act programs. 3028 * Return 3029 * * < 0 if any input argument is invalid 3030 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3031 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3032 * packet is not forwarded or needs assist from full stack 3033 * 3034 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3035 * was exceeded and output params->mtu_result contains the MTU. 3036 * 3037 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3038 * Description 3039 * Add an entry to, or update a sockhash *map* referencing sockets. 3040 * The *skops* is used as a new value for the entry associated to 3041 * *key*. *flags* is one of: 3042 * 3043 * **BPF_NOEXIST** 3044 * The entry for *key* must not exist in the map. 3045 * **BPF_EXIST** 3046 * The entry for *key* must already exist in the map. 3047 * **BPF_ANY** 3048 * No condition on the existence of the entry for *key*. 3049 * 3050 * If the *map* has eBPF programs (parser and verdict), those will 3051 * be inherited by the socket being added. If the socket is 3052 * already attached to eBPF programs, this results in an error. 3053 * Return 3054 * 0 on success, or a negative error in case of failure. 3055 * 3056 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3057 * Description 3058 * This helper is used in programs implementing policies at the 3059 * socket level. If the message *msg* is allowed to pass (i.e. if 3060 * the verdict eBPF program returns **SK_PASS**), redirect it to 3061 * the socket referenced by *map* (of type 3062 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3063 * egress interfaces can be used for redirection. The 3064 * **BPF_F_INGRESS** value in *flags* is used to make the 3065 * distinction (ingress path is selected if the flag is present, 3066 * egress path otherwise). This is the only flag supported for now. 3067 * Return 3068 * **SK_PASS** on success, or **SK_DROP** on error. 3069 * 3070 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3071 * Description 3072 * This helper is used in programs implementing policies at the 3073 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3074 * if the verdict eBPF program returns **SK_PASS**), redirect it 3075 * to the socket referenced by *map* (of type 3076 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3077 * egress interfaces can be used for redirection. The 3078 * **BPF_F_INGRESS** value in *flags* is used to make the 3079 * distinction (ingress path is selected if the flag is present, 3080 * egress otherwise). This is the only flag supported for now. 3081 * Return 3082 * **SK_PASS** on success, or **SK_DROP** on error. 3083 * 3084 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3085 * Description 3086 * Encapsulate the packet associated to *skb* within a Layer 3 3087 * protocol header. This header is provided in the buffer at 3088 * address *hdr*, with *len* its size in bytes. *type* indicates 3089 * the protocol of the header and can be one of: 3090 * 3091 * **BPF_LWT_ENCAP_SEG6** 3092 * IPv6 encapsulation with Segment Routing Header 3093 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3094 * the IPv6 header is computed by the kernel. 3095 * **BPF_LWT_ENCAP_SEG6_INLINE** 3096 * Only works if *skb* contains an IPv6 packet. Insert a 3097 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3098 * the IPv6 header. 3099 * **BPF_LWT_ENCAP_IP** 3100 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3101 * must be IPv4 or IPv6, followed by zero or more 3102 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3103 * total bytes in all prepended headers. Please note that 3104 * if **skb_is_gso**\ (*skb*) is true, no more than two 3105 * headers can be prepended, and the inner header, if 3106 * present, should be either GRE or UDP/GUE. 3107 * 3108 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3109 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3110 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3111 * **BPF_PROG_TYPE_LWT_XMIT**. 3112 * 3113 * A call to this helper is susceptible to change the underlying 3114 * packet buffer. Therefore, at load time, all checks on pointers 3115 * previously done by the verifier are invalidated and must be 3116 * performed again, if the helper is used in combination with 3117 * direct packet access. 3118 * Return 3119 * 0 on success, or a negative error in case of failure. 3120 * 3121 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3122 * Description 3123 * Store *len* bytes from address *from* into the packet 3124 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3125 * inside the outermost IPv6 Segment Routing Header can be 3126 * modified through this helper. 3127 * 3128 * A call to this helper is susceptible to change the underlying 3129 * packet buffer. Therefore, at load time, all checks on pointers 3130 * previously done by the verifier are invalidated and must be 3131 * performed again, if the helper is used in combination with 3132 * direct packet access. 3133 * Return 3134 * 0 on success, or a negative error in case of failure. 3135 * 3136 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3137 * Description 3138 * Adjust the size allocated to TLVs in the outermost IPv6 3139 * Segment Routing Header contained in the packet associated to 3140 * *skb*, at position *offset* by *delta* bytes. Only offsets 3141 * after the segments are accepted. *delta* can be as well 3142 * positive (growing) as negative (shrinking). 3143 * 3144 * A call to this helper is susceptible to change the underlying 3145 * packet buffer. Therefore, at load time, all checks on pointers 3146 * previously done by the verifier are invalidated and must be 3147 * performed again, if the helper is used in combination with 3148 * direct packet access. 3149 * Return 3150 * 0 on success, or a negative error in case of failure. 3151 * 3152 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3153 * Description 3154 * Apply an IPv6 Segment Routing action of type *action* to the 3155 * packet associated to *skb*. Each action takes a parameter 3156 * contained at address *param*, and of length *param_len* bytes. 3157 * *action* can be one of: 3158 * 3159 * **SEG6_LOCAL_ACTION_END_X** 3160 * End.X action: Endpoint with Layer-3 cross-connect. 3161 * Type of *param*: **struct in6_addr**. 3162 * **SEG6_LOCAL_ACTION_END_T** 3163 * End.T action: Endpoint with specific IPv6 table lookup. 3164 * Type of *param*: **int**. 3165 * **SEG6_LOCAL_ACTION_END_B6** 3166 * End.B6 action: Endpoint bound to an SRv6 policy. 3167 * Type of *param*: **struct ipv6_sr_hdr**. 3168 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3169 * End.B6.Encap action: Endpoint bound to an SRv6 3170 * encapsulation policy. 3171 * Type of *param*: **struct ipv6_sr_hdr**. 3172 * 3173 * A call to this helper is susceptible to change the underlying 3174 * packet buffer. Therefore, at load time, all checks on pointers 3175 * previously done by the verifier are invalidated and must be 3176 * performed again, if the helper is used in combination with 3177 * direct packet access. 3178 * Return 3179 * 0 on success, or a negative error in case of failure. 3180 * 3181 * long bpf_rc_repeat(void *ctx) 3182 * Description 3183 * This helper is used in programs implementing IR decoding, to 3184 * report a successfully decoded repeat key message. This delays 3185 * the generation of a key up event for previously generated 3186 * key down event. 3187 * 3188 * Some IR protocols like NEC have a special IR message for 3189 * repeating last button, for when a button is held down. 3190 * 3191 * The *ctx* should point to the lirc sample as passed into 3192 * the program. 3193 * 3194 * This helper is only available is the kernel was compiled with 3195 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3196 * "**y**". 3197 * Return 3198 * 0 3199 * 3200 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3201 * Description 3202 * This helper is used in programs implementing IR decoding, to 3203 * report a successfully decoded key press with *scancode*, 3204 * *toggle* value in the given *protocol*. The scancode will be 3205 * translated to a keycode using the rc keymap, and reported as 3206 * an input key down event. After a period a key up event is 3207 * generated. This period can be extended by calling either 3208 * **bpf_rc_keydown**\ () again with the same values, or calling 3209 * **bpf_rc_repeat**\ (). 3210 * 3211 * Some protocols include a toggle bit, in case the button was 3212 * released and pressed again between consecutive scancodes. 3213 * 3214 * The *ctx* should point to the lirc sample as passed into 3215 * the program. 3216 * 3217 * The *protocol* is the decoded protocol number (see 3218 * **enum rc_proto** for some predefined values). 3219 * 3220 * This helper is only available is the kernel was compiled with 3221 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3222 * "**y**". 3223 * Return 3224 * 0 3225 * 3226 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3227 * Description 3228 * Return the cgroup v2 id of the socket associated with the *skb*. 3229 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3230 * helper for cgroup v1 by providing a tag resp. identifier that 3231 * can be matched on or used for map lookups e.g. to implement 3232 * policy. The cgroup v2 id of a given path in the hierarchy is 3233 * exposed in user space through the f_handle API in order to get 3234 * to the same 64-bit id. 3235 * 3236 * This helper can be used on TC egress path, but not on ingress, 3237 * and is available only if the kernel was compiled with the 3238 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3239 * Return 3240 * The id is returned or 0 in case the id could not be retrieved. 3241 * 3242 * u64 bpf_get_current_cgroup_id(void) 3243 * Return 3244 * A 64-bit integer containing the current cgroup id based 3245 * on the cgroup within which the current task is running. 3246 * 3247 * void *bpf_get_local_storage(void *map, u64 flags) 3248 * Description 3249 * Get the pointer to the local storage area. 3250 * The type and the size of the local storage is defined 3251 * by the *map* argument. 3252 * The *flags* meaning is specific for each map type, 3253 * and has to be 0 for cgroup local storage. 3254 * 3255 * Depending on the BPF program type, a local storage area 3256 * can be shared between multiple instances of the BPF program, 3257 * running simultaneously. 3258 * 3259 * A user should care about the synchronization by himself. 3260 * For example, by using the **BPF_ATOMIC** instructions to alter 3261 * the shared data. 3262 * Return 3263 * A pointer to the local storage area. 3264 * 3265 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3266 * Description 3267 * Select a **SO_REUSEPORT** socket from a 3268 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3269 * It checks the selected socket is matching the incoming 3270 * request in the socket buffer. 3271 * Return 3272 * 0 on success, or a negative error in case of failure. 3273 * 3274 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3275 * Description 3276 * Return id of cgroup v2 that is ancestor of cgroup associated 3277 * with the *skb* at the *ancestor_level*. The root cgroup is at 3278 * *ancestor_level* zero and each step down the hierarchy 3279 * increments the level. If *ancestor_level* == level of cgroup 3280 * associated with *skb*, then return value will be same as that 3281 * of **bpf_skb_cgroup_id**\ (). 3282 * 3283 * The helper is useful to implement policies based on cgroups 3284 * that are upper in hierarchy than immediate cgroup associated 3285 * with *skb*. 3286 * 3287 * The format of returned id and helper limitations are same as in 3288 * **bpf_skb_cgroup_id**\ (). 3289 * Return 3290 * The id is returned or 0 in case the id could not be retrieved. 3291 * 3292 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3293 * Description 3294 * Look for TCP socket matching *tuple*, optionally in a child 3295 * network namespace *netns*. The return value must be checked, 3296 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3297 * 3298 * The *ctx* should point to the context of the program, such as 3299 * the skb or socket (depending on the hook in use). This is used 3300 * to determine the base network namespace for the lookup. 3301 * 3302 * *tuple_size* must be one of: 3303 * 3304 * **sizeof**\ (*tuple*\ **->ipv4**) 3305 * Look for an IPv4 socket. 3306 * **sizeof**\ (*tuple*\ **->ipv6**) 3307 * Look for an IPv6 socket. 3308 * 3309 * If the *netns* is a negative signed 32-bit integer, then the 3310 * socket lookup table in the netns associated with the *ctx* 3311 * will be used. For the TC hooks, this is the netns of the device 3312 * in the skb. For socket hooks, this is the netns of the socket. 3313 * If *netns* is any other signed 32-bit value greater than or 3314 * equal to zero then it specifies the ID of the netns relative to 3315 * the netns associated with the *ctx*. *netns* values beyond the 3316 * range of 32-bit integers are reserved for future use. 3317 * 3318 * All values for *flags* are reserved for future usage, and must 3319 * be left at zero. 3320 * 3321 * This helper is available only if the kernel was compiled with 3322 * **CONFIG_NET** configuration option. 3323 * Return 3324 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3325 * For sockets with reuseport option, the **struct bpf_sock** 3326 * result is from *reuse*\ **->socks**\ [] using the hash of the 3327 * tuple. 3328 * 3329 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3330 * Description 3331 * Look for UDP socket matching *tuple*, optionally in a child 3332 * network namespace *netns*. The return value must be checked, 3333 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3334 * 3335 * The *ctx* should point to the context of the program, such as 3336 * the skb or socket (depending on the hook in use). This is used 3337 * to determine the base network namespace for the lookup. 3338 * 3339 * *tuple_size* must be one of: 3340 * 3341 * **sizeof**\ (*tuple*\ **->ipv4**) 3342 * Look for an IPv4 socket. 3343 * **sizeof**\ (*tuple*\ **->ipv6**) 3344 * Look for an IPv6 socket. 3345 * 3346 * If the *netns* is a negative signed 32-bit integer, then the 3347 * socket lookup table in the netns associated with the *ctx* 3348 * will be used. For the TC hooks, this is the netns of the device 3349 * in the skb. For socket hooks, this is the netns of the socket. 3350 * If *netns* is any other signed 32-bit value greater than or 3351 * equal to zero then it specifies the ID of the netns relative to 3352 * the netns associated with the *ctx*. *netns* values beyond the 3353 * range of 32-bit integers are reserved for future use. 3354 * 3355 * All values for *flags* are reserved for future usage, and must 3356 * be left at zero. 3357 * 3358 * This helper is available only if the kernel was compiled with 3359 * **CONFIG_NET** configuration option. 3360 * Return 3361 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3362 * For sockets with reuseport option, the **struct bpf_sock** 3363 * result is from *reuse*\ **->socks**\ [] using the hash of the 3364 * tuple. 3365 * 3366 * long bpf_sk_release(void *sock) 3367 * Description 3368 * Release the reference held by *sock*. *sock* must be a 3369 * non-**NULL** pointer that was returned from 3370 * **bpf_sk_lookup_xxx**\ (). 3371 * Return 3372 * 0 on success, or a negative error in case of failure. 3373 * 3374 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3375 * Description 3376 * Push an element *value* in *map*. *flags* is one of: 3377 * 3378 * **BPF_EXIST** 3379 * If the queue/stack is full, the oldest element is 3380 * removed to make room for this. 3381 * Return 3382 * 0 on success, or a negative error in case of failure. 3383 * 3384 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3385 * Description 3386 * Pop an element from *map*. 3387 * Return 3388 * 0 on success, or a negative error in case of failure. 3389 * 3390 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3391 * Description 3392 * Get an element from *map* without removing it. 3393 * Return 3394 * 0 on success, or a negative error in case of failure. 3395 * 3396 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3397 * Description 3398 * For socket policies, insert *len* bytes into *msg* at offset 3399 * *start*. 3400 * 3401 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3402 * *msg* it may want to insert metadata or options into the *msg*. 3403 * This can later be read and used by any of the lower layer BPF 3404 * hooks. 3405 * 3406 * This helper may fail if under memory pressure (a malloc 3407 * fails) in these cases BPF programs will get an appropriate 3408 * error and BPF programs will need to handle them. 3409 * Return 3410 * 0 on success, or a negative error in case of failure. 3411 * 3412 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3413 * Description 3414 * Will remove *len* bytes from a *msg* starting at byte *start*. 3415 * This may result in **ENOMEM** errors under certain situations if 3416 * an allocation and copy are required due to a full ring buffer. 3417 * However, the helper will try to avoid doing the allocation 3418 * if possible. Other errors can occur if input parameters are 3419 * invalid either due to *start* byte not being valid part of *msg* 3420 * payload and/or *pop* value being to large. 3421 * Return 3422 * 0 on success, or a negative error in case of failure. 3423 * 3424 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3425 * Description 3426 * This helper is used in programs implementing IR decoding, to 3427 * report a successfully decoded pointer movement. 3428 * 3429 * The *ctx* should point to the lirc sample as passed into 3430 * the program. 3431 * 3432 * This helper is only available is the kernel was compiled with 3433 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3434 * "**y**". 3435 * Return 3436 * 0 3437 * 3438 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3439 * Description 3440 * Acquire a spinlock represented by the pointer *lock*, which is 3441 * stored as part of a value of a map. Taking the lock allows to 3442 * safely update the rest of the fields in that value. The 3443 * spinlock can (and must) later be released with a call to 3444 * **bpf_spin_unlock**\ (\ *lock*\ ). 3445 * 3446 * Spinlocks in BPF programs come with a number of restrictions 3447 * and constraints: 3448 * 3449 * * **bpf_spin_lock** objects are only allowed inside maps of 3450 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3451 * list could be extended in the future). 3452 * * BTF description of the map is mandatory. 3453 * * The BPF program can take ONE lock at a time, since taking two 3454 * or more could cause dead locks. 3455 * * Only one **struct bpf_spin_lock** is allowed per map element. 3456 * * When the lock is taken, calls (either BPF to BPF or helpers) 3457 * are not allowed. 3458 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3459 * allowed inside a spinlock-ed region. 3460 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3461 * the lock, on all execution paths, before it returns. 3462 * * The BPF program can access **struct bpf_spin_lock** only via 3463 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3464 * helpers. Loading or storing data into the **struct 3465 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3466 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3467 * of the map value must be a struct and have **struct 3468 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3469 * Nested lock inside another struct is not allowed. 3470 * * The **struct bpf_spin_lock** *lock* field in a map value must 3471 * be aligned on a multiple of 4 bytes in that value. 3472 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3473 * the **bpf_spin_lock** field to user space. 3474 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3475 * a BPF program, do not update the **bpf_spin_lock** field. 3476 * * **bpf_spin_lock** cannot be on the stack or inside a 3477 * networking packet (it can only be inside of a map values). 3478 * * **bpf_spin_lock** is available to root only. 3479 * * Tracing programs and socket filter programs cannot use 3480 * **bpf_spin_lock**\ () due to insufficient preemption checks 3481 * (but this may change in the future). 3482 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3483 * Return 3484 * 0 3485 * 3486 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3487 * Description 3488 * Release the *lock* previously locked by a call to 3489 * **bpf_spin_lock**\ (\ *lock*\ ). 3490 * Return 3491 * 0 3492 * 3493 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3494 * Description 3495 * This helper gets a **struct bpf_sock** pointer such 3496 * that all the fields in this **bpf_sock** can be accessed. 3497 * Return 3498 * A **struct bpf_sock** pointer on success, or **NULL** in 3499 * case of failure. 3500 * 3501 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3502 * Description 3503 * This helper gets a **struct bpf_tcp_sock** pointer from a 3504 * **struct bpf_sock** pointer. 3505 * Return 3506 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3507 * case of failure. 3508 * 3509 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3510 * Description 3511 * Set ECN (Explicit Congestion Notification) field of IP header 3512 * to **CE** (Congestion Encountered) if current value is **ECT** 3513 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3514 * and IPv4. 3515 * Return 3516 * 1 if the **CE** flag is set (either by the current helper call 3517 * or because it was already present), 0 if it is not set. 3518 * 3519 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3520 * Description 3521 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3522 * **bpf_sk_release**\ () is unnecessary and not allowed. 3523 * Return 3524 * A **struct bpf_sock** pointer on success, or **NULL** in 3525 * case of failure. 3526 * 3527 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3528 * Description 3529 * Look for TCP socket matching *tuple*, optionally in a child 3530 * network namespace *netns*. The return value must be checked, 3531 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3532 * 3533 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3534 * that it also returns timewait or request sockets. Use 3535 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3536 * full structure. 3537 * 3538 * This helper is available only if the kernel was compiled with 3539 * **CONFIG_NET** configuration option. 3540 * Return 3541 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3542 * For sockets with reuseport option, the **struct bpf_sock** 3543 * result is from *reuse*\ **->socks**\ [] using the hash of the 3544 * tuple. 3545 * 3546 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3547 * Description 3548 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3549 * the listening socket in *sk*. 3550 * 3551 * *iph* points to the start of the IPv4 or IPv6 header, while 3552 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3553 * **sizeof**\ (**struct ip6hdr**). 3554 * 3555 * *th* points to the start of the TCP header, while *th_len* 3556 * contains **sizeof**\ (**struct tcphdr**). 3557 * Return 3558 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3559 * error otherwise. 3560 * 3561 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3562 * Description 3563 * Get name of sysctl in /proc/sys/ and copy it into provided by 3564 * program buffer *buf* of size *buf_len*. 3565 * 3566 * The buffer is always NUL terminated, unless it's zero-sized. 3567 * 3568 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3569 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3570 * only (e.g. "tcp_mem"). 3571 * Return 3572 * Number of character copied (not including the trailing NUL). 3573 * 3574 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3575 * truncated name in this case). 3576 * 3577 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3578 * Description 3579 * Get current value of sysctl as it is presented in /proc/sys 3580 * (incl. newline, etc), and copy it as a string into provided 3581 * by program buffer *buf* of size *buf_len*. 3582 * 3583 * The whole value is copied, no matter what file position user 3584 * space issued e.g. sys_read at. 3585 * 3586 * The buffer is always NUL terminated, unless it's zero-sized. 3587 * Return 3588 * Number of character copied (not including the trailing NUL). 3589 * 3590 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3591 * truncated name in this case). 3592 * 3593 * **-EINVAL** if current value was unavailable, e.g. because 3594 * sysctl is uninitialized and read returns -EIO for it. 3595 * 3596 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3597 * Description 3598 * Get new value being written by user space to sysctl (before 3599 * the actual write happens) and copy it as a string into 3600 * provided by program buffer *buf* of size *buf_len*. 3601 * 3602 * User space may write new value at file position > 0. 3603 * 3604 * The buffer is always NUL terminated, unless it's zero-sized. 3605 * Return 3606 * Number of character copied (not including the trailing NUL). 3607 * 3608 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3609 * truncated name in this case). 3610 * 3611 * **-EINVAL** if sysctl is being read. 3612 * 3613 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3614 * Description 3615 * Override new value being written by user space to sysctl with 3616 * value provided by program in buffer *buf* of size *buf_len*. 3617 * 3618 * *buf* should contain a string in same form as provided by user 3619 * space on sysctl write. 3620 * 3621 * User space may write new value at file position > 0. To override 3622 * the whole sysctl value file position should be set to zero. 3623 * Return 3624 * 0 on success. 3625 * 3626 * **-E2BIG** if the *buf_len* is too big. 3627 * 3628 * **-EINVAL** if sysctl is being read. 3629 * 3630 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3631 * Description 3632 * Convert the initial part of the string from buffer *buf* of 3633 * size *buf_len* to a long integer according to the given base 3634 * and save the result in *res*. 3635 * 3636 * The string may begin with an arbitrary amount of white space 3637 * (as determined by **isspace**\ (3)) followed by a single 3638 * optional '**-**' sign. 3639 * 3640 * Five least significant bits of *flags* encode base, other bits 3641 * are currently unused. 3642 * 3643 * Base must be either 8, 10, 16 or 0 to detect it automatically 3644 * similar to user space **strtol**\ (3). 3645 * Return 3646 * Number of characters consumed on success. Must be positive but 3647 * no more than *buf_len*. 3648 * 3649 * **-EINVAL** if no valid digits were found or unsupported base 3650 * was provided. 3651 * 3652 * **-ERANGE** if resulting value was out of range. 3653 * 3654 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3655 * Description 3656 * Convert the initial part of the string from buffer *buf* of 3657 * size *buf_len* to an unsigned long integer according to the 3658 * given base and save the result in *res*. 3659 * 3660 * The string may begin with an arbitrary amount of white space 3661 * (as determined by **isspace**\ (3)). 3662 * 3663 * Five least significant bits of *flags* encode base, other bits 3664 * are currently unused. 3665 * 3666 * Base must be either 8, 10, 16 or 0 to detect it automatically 3667 * similar to user space **strtoul**\ (3). 3668 * Return 3669 * Number of characters consumed on success. Must be positive but 3670 * no more than *buf_len*. 3671 * 3672 * **-EINVAL** if no valid digits were found or unsupported base 3673 * was provided. 3674 * 3675 * **-ERANGE** if resulting value was out of range. 3676 * 3677 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3678 * Description 3679 * Get a bpf-local-storage from a *sk*. 3680 * 3681 * Logically, it could be thought of getting the value from 3682 * a *map* with *sk* as the **key**. From this 3683 * perspective, the usage is not much different from 3684 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3685 * helper enforces the key must be a full socket and the map must 3686 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3687 * 3688 * Underneath, the value is stored locally at *sk* instead of 3689 * the *map*. The *map* is used as the bpf-local-storage 3690 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3691 * searched against all bpf-local-storages residing at *sk*. 3692 * 3693 * *sk* is a kernel **struct sock** pointer for LSM program. 3694 * *sk* is a **struct bpf_sock** pointer for other program types. 3695 * 3696 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3697 * used such that a new bpf-local-storage will be 3698 * created if one does not exist. *value* can be used 3699 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3700 * the initial value of a bpf-local-storage. If *value* is 3701 * **NULL**, the new bpf-local-storage will be zero initialized. 3702 * Return 3703 * A bpf-local-storage pointer is returned on success. 3704 * 3705 * **NULL** if not found or there was an error in adding 3706 * a new bpf-local-storage. 3707 * 3708 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3709 * Description 3710 * Delete a bpf-local-storage from a *sk*. 3711 * Return 3712 * 0 on success. 3713 * 3714 * **-ENOENT** if the bpf-local-storage cannot be found. 3715 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3716 * 3717 * long bpf_send_signal(u32 sig) 3718 * Description 3719 * Send signal *sig* to the process of the current task. 3720 * The signal may be delivered to any of this process's threads. 3721 * Return 3722 * 0 on success or successfully queued. 3723 * 3724 * **-EBUSY** if work queue under nmi is full. 3725 * 3726 * **-EINVAL** if *sig* is invalid. 3727 * 3728 * **-EPERM** if no permission to send the *sig*. 3729 * 3730 * **-EAGAIN** if bpf program can try again. 3731 * 3732 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3733 * Description 3734 * Try to issue a SYN cookie for the packet with corresponding 3735 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3736 * 3737 * *iph* points to the start of the IPv4 or IPv6 header, while 3738 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3739 * **sizeof**\ (**struct ip6hdr**). 3740 * 3741 * *th* points to the start of the TCP header, while *th_len* 3742 * contains the length of the TCP header. 3743 * Return 3744 * On success, lower 32 bits hold the generated SYN cookie in 3745 * followed by 16 bits which hold the MSS value for that cookie, 3746 * and the top 16 bits are unused. 3747 * 3748 * On failure, the returned value is one of the following: 3749 * 3750 * **-EINVAL** SYN cookie cannot be issued due to error 3751 * 3752 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3753 * 3754 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3755 * 3756 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3757 * 3758 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3759 * Description 3760 * Write raw *data* blob into a special BPF perf event held by 3761 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3762 * event must have the following attributes: **PERF_SAMPLE_RAW** 3763 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3764 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3765 * 3766 * The *flags* are used to indicate the index in *map* for which 3767 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3768 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3769 * to indicate that the index of the current CPU core should be 3770 * used. 3771 * 3772 * The value to write, of *size*, is passed through eBPF stack and 3773 * pointed by *data*. 3774 * 3775 * *ctx* is a pointer to in-kernel struct sk_buff. 3776 * 3777 * This helper is similar to **bpf_perf_event_output**\ () but 3778 * restricted to raw_tracepoint bpf programs. 3779 * Return 3780 * 0 on success, or a negative error in case of failure. 3781 * 3782 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 3783 * Description 3784 * Safely attempt to read *size* bytes from user space address 3785 * *unsafe_ptr* and store the data in *dst*. 3786 * Return 3787 * 0 on success, or a negative error in case of failure. 3788 * 3789 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 3790 * Description 3791 * Safely attempt to read *size* bytes from kernel space address 3792 * *unsafe_ptr* and store the data in *dst*. 3793 * Return 3794 * 0 on success, or a negative error in case of failure. 3795 * 3796 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 3797 * Description 3798 * Copy a NUL terminated string from an unsafe user address 3799 * *unsafe_ptr* to *dst*. The *size* should include the 3800 * terminating NUL byte. In case the string length is smaller than 3801 * *size*, the target is not padded with further NUL bytes. If the 3802 * string length is larger than *size*, just *size*-1 bytes are 3803 * copied and the last byte is set to NUL. 3804 * 3805 * On success, returns the number of bytes that were written, 3806 * including the terminal NUL. This makes this helper useful in 3807 * tracing programs for reading strings, and more importantly to 3808 * get its length at runtime. See the following snippet: 3809 * 3810 * :: 3811 * 3812 * SEC("kprobe/sys_open") 3813 * void bpf_sys_open(struct pt_regs *ctx) 3814 * { 3815 * char buf[PATHLEN]; // PATHLEN is defined to 256 3816 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3817 * ctx->di); 3818 * 3819 * // Consume buf, for example push it to 3820 * // userspace via bpf_perf_event_output(); we 3821 * // can use res (the string length) as event 3822 * // size, after checking its boundaries. 3823 * } 3824 * 3825 * In comparison, using **bpf_probe_read_user**\ () helper here 3826 * instead to read the string would require to estimate the length 3827 * at compile time, and would often result in copying more memory 3828 * than necessary. 3829 * 3830 * Another useful use case is when parsing individual process 3831 * arguments or individual environment variables navigating 3832 * *current*\ **->mm->arg_start** and *current*\ 3833 * **->mm->env_start**: using this helper and the return value, 3834 * one can quickly iterate at the right offset of the memory area. 3835 * Return 3836 * On success, the strictly positive length of the output string, 3837 * including the trailing NUL character. On error, a negative 3838 * value. 3839 * 3840 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3841 * Description 3842 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3843 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3844 * Return 3845 * On success, the strictly positive length of the string, including 3846 * the trailing NUL character. On error, a negative value. 3847 * 3848 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3849 * Description 3850 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3851 * *rcv_nxt* is the ack_seq to be sent out. 3852 * Return 3853 * 0 on success, or a negative error in case of failure. 3854 * 3855 * long bpf_send_signal_thread(u32 sig) 3856 * Description 3857 * Send signal *sig* to the thread corresponding to the current task. 3858 * Return 3859 * 0 on success or successfully queued. 3860 * 3861 * **-EBUSY** if work queue under nmi is full. 3862 * 3863 * **-EINVAL** if *sig* is invalid. 3864 * 3865 * **-EPERM** if no permission to send the *sig*. 3866 * 3867 * **-EAGAIN** if bpf program can try again. 3868 * 3869 * u64 bpf_jiffies64(void) 3870 * Description 3871 * Obtain the 64bit jiffies 3872 * Return 3873 * The 64 bit jiffies 3874 * 3875 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3876 * Description 3877 * For an eBPF program attached to a perf event, retrieve the 3878 * branch records (**struct perf_branch_entry**) associated to *ctx* 3879 * and store it in the buffer pointed by *buf* up to size 3880 * *size* bytes. 3881 * Return 3882 * On success, number of bytes written to *buf*. On error, a 3883 * negative value. 3884 * 3885 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3886 * instead return the number of bytes required to store all the 3887 * branch entries. If this flag is set, *buf* may be NULL. 3888 * 3889 * **-EINVAL** if arguments invalid or **size** not a multiple 3890 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3891 * 3892 * **-ENOENT** if architecture does not support branch records. 3893 * 3894 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3895 * Description 3896 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3897 * *namespace* will be returned in *nsdata*. 3898 * Return 3899 * 0 on success, or one of the following in case of failure: 3900 * 3901 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3902 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3903 * 3904 * **-ENOENT** if pidns does not exists for the current task. 3905 * 3906 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3907 * Description 3908 * Write raw *data* blob into a special BPF perf event held by 3909 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3910 * event must have the following attributes: **PERF_SAMPLE_RAW** 3911 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3912 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3913 * 3914 * The *flags* are used to indicate the index in *map* for which 3915 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3916 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3917 * to indicate that the index of the current CPU core should be 3918 * used. 3919 * 3920 * The value to write, of *size*, is passed through eBPF stack and 3921 * pointed by *data*. 3922 * 3923 * *ctx* is a pointer to in-kernel struct xdp_buff. 3924 * 3925 * This helper is similar to **bpf_perf_eventoutput**\ () but 3926 * restricted to raw_tracepoint bpf programs. 3927 * Return 3928 * 0 on success, or a negative error in case of failure. 3929 * 3930 * u64 bpf_get_netns_cookie(void *ctx) 3931 * Description 3932 * Retrieve the cookie (generated by the kernel) of the network 3933 * namespace the input *ctx* is associated with. The network 3934 * namespace cookie remains stable for its lifetime and provides 3935 * a global identifier that can be assumed unique. If *ctx* is 3936 * NULL, then the helper returns the cookie for the initial 3937 * network namespace. The cookie itself is very similar to that 3938 * of **bpf_get_socket_cookie**\ () helper, but for network 3939 * namespaces instead of sockets. 3940 * Return 3941 * A 8-byte long opaque number. 3942 * 3943 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3944 * Description 3945 * Return id of cgroup v2 that is ancestor of the cgroup associated 3946 * with the current task at the *ancestor_level*. The root cgroup 3947 * is at *ancestor_level* zero and each step down the hierarchy 3948 * increments the level. If *ancestor_level* == level of cgroup 3949 * associated with the current task, then return value will be the 3950 * same as that of **bpf_get_current_cgroup_id**\ (). 3951 * 3952 * The helper is useful to implement policies based on cgroups 3953 * that are upper in hierarchy than immediate cgroup associated 3954 * with the current task. 3955 * 3956 * The format of returned id and helper limitations are same as in 3957 * **bpf_get_current_cgroup_id**\ (). 3958 * Return 3959 * The id is returned or 0 in case the id could not be retrieved. 3960 * 3961 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3962 * Description 3963 * Helper is overloaded depending on BPF program type. This 3964 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3965 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3966 * 3967 * Assign the *sk* to the *skb*. When combined with appropriate 3968 * routing configuration to receive the packet towards the socket, 3969 * will cause *skb* to be delivered to the specified socket. 3970 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3971 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3972 * interfere with successful delivery to the socket. 3973 * 3974 * This operation is only valid from TC ingress path. 3975 * 3976 * The *flags* argument must be zero. 3977 * Return 3978 * 0 on success, or a negative error in case of failure: 3979 * 3980 * **-EINVAL** if specified *flags* are not supported. 3981 * 3982 * **-ENOENT** if the socket is unavailable for assignment. 3983 * 3984 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3985 * 3986 * **-EOPNOTSUPP** if the operation is not supported, for example 3987 * a call from outside of TC ingress. 3988 * 3989 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3990 * (reuseport). 3991 * 3992 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3993 * Description 3994 * Helper is overloaded depending on BPF program type. This 3995 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3996 * 3997 * Select the *sk* as a result of a socket lookup. 3998 * 3999 * For the operation to succeed passed socket must be compatible 4000 * with the packet description provided by the *ctx* object. 4001 * 4002 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4003 * be an exact match. While IP family (**AF_INET** or 4004 * **AF_INET6**) must be compatible, that is IPv6 sockets 4005 * that are not v6-only can be selected for IPv4 packets. 4006 * 4007 * Only TCP listeners and UDP unconnected sockets can be 4008 * selected. *sk* can also be NULL to reset any previous 4009 * selection. 4010 * 4011 * *flags* argument can combination of following values: 4012 * 4013 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4014 * socket selection, potentially done by a BPF program 4015 * that ran before us. 4016 * 4017 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4018 * load-balancing within reuseport group for the socket 4019 * being selected. 4020 * 4021 * On success *ctx->sk* will point to the selected socket. 4022 * 4023 * Return 4024 * 0 on success, or a negative errno in case of failure. 4025 * 4026 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4027 * not compatible with packet family (*ctx->family*). 4028 * 4029 * * **-EEXIST** if socket has been already selected, 4030 * potentially by another program, and 4031 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4032 * 4033 * * **-EINVAL** if unsupported flags were specified. 4034 * 4035 * * **-EPROTOTYPE** if socket L4 protocol 4036 * (*sk->protocol*) doesn't match packet protocol 4037 * (*ctx->protocol*). 4038 * 4039 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4040 * state (TCP listening or UDP unconnected). 4041 * 4042 * u64 bpf_ktime_get_boot_ns(void) 4043 * Description 4044 * Return the time elapsed since system boot, in nanoseconds. 4045 * Does include the time the system was suspended. 4046 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4047 * Return 4048 * Current *ktime*. 4049 * 4050 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4051 * Description 4052 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4053 * out the format string. 4054 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4055 * the format string itself. The *data* and *data_len* are format string 4056 * arguments. The *data* are a **u64** array and corresponding format string 4057 * values are stored in the array. For strings and pointers where pointees 4058 * are accessed, only the pointer values are stored in the *data* array. 4059 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4060 * 4061 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4062 * Reading kernel memory may fail due to either invalid address or 4063 * valid address but requiring a major memory fault. If reading kernel memory 4064 * fails, the string for **%s** will be an empty string, and the ip 4065 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4066 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4067 * Return 4068 * 0 on success, or a negative error in case of failure: 4069 * 4070 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4071 * by returning 1 from bpf program. 4072 * 4073 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4074 * 4075 * **-E2BIG** if *fmt* contains too many format specifiers. 4076 * 4077 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4078 * 4079 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4080 * Description 4081 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4082 * The *m* represents the seq_file. The *data* and *len* represent the 4083 * data to write in bytes. 4084 * Return 4085 * 0 on success, or a negative error in case of failure: 4086 * 4087 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4088 * 4089 * u64 bpf_sk_cgroup_id(void *sk) 4090 * Description 4091 * Return the cgroup v2 id of the socket *sk*. 4092 * 4093 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4094 * returned from **bpf_sk_lookup_xxx**\ (), 4095 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4096 * same as in **bpf_skb_cgroup_id**\ (). 4097 * 4098 * This helper is available only if the kernel was compiled with 4099 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4100 * Return 4101 * The id is returned or 0 in case the id could not be retrieved. 4102 * 4103 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4104 * Description 4105 * Return id of cgroup v2 that is ancestor of cgroup associated 4106 * with the *sk* at the *ancestor_level*. The root cgroup is at 4107 * *ancestor_level* zero and each step down the hierarchy 4108 * increments the level. If *ancestor_level* == level of cgroup 4109 * associated with *sk*, then return value will be same as that 4110 * of **bpf_sk_cgroup_id**\ (). 4111 * 4112 * The helper is useful to implement policies based on cgroups 4113 * that are upper in hierarchy than immediate cgroup associated 4114 * with *sk*. 4115 * 4116 * The format of returned id and helper limitations are same as in 4117 * **bpf_sk_cgroup_id**\ (). 4118 * Return 4119 * The id is returned or 0 in case the id could not be retrieved. 4120 * 4121 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4122 * Description 4123 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4124 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4125 * of new data availability is sent. 4126 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4127 * of new data availability is sent unconditionally. 4128 * If **0** is specified in *flags*, an adaptive notification 4129 * of new data availability is sent. 4130 * 4131 * An adaptive notification is a notification sent whenever the user-space 4132 * process has caught up and consumed all available payloads. In case the user-space 4133 * process is still processing a previous payload, then no notification is needed 4134 * as it will process the newly added payload automatically. 4135 * Return 4136 * 0 on success, or a negative error in case of failure. 4137 * 4138 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4139 * Description 4140 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4141 * *flags* must be 0. 4142 * Return 4143 * Valid pointer with *size* bytes of memory available; NULL, 4144 * otherwise. 4145 * 4146 * void bpf_ringbuf_submit(void *data, u64 flags) 4147 * Description 4148 * Submit reserved ring buffer sample, pointed to by *data*. 4149 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4150 * of new data availability is sent. 4151 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4152 * of new data availability is sent unconditionally. 4153 * If **0** is specified in *flags*, an adaptive notification 4154 * of new data availability is sent. 4155 * 4156 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4157 * Return 4158 * Nothing. Always succeeds. 4159 * 4160 * void bpf_ringbuf_discard(void *data, u64 flags) 4161 * Description 4162 * Discard reserved ring buffer sample, pointed to by *data*. 4163 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4164 * of new data availability is sent. 4165 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4166 * of new data availability is sent unconditionally. 4167 * If **0** is specified in *flags*, an adaptive notification 4168 * of new data availability is sent. 4169 * 4170 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4171 * Return 4172 * Nothing. Always succeeds. 4173 * 4174 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4175 * Description 4176 * Query various characteristics of provided ring buffer. What 4177 * exactly is queries is determined by *flags*: 4178 * 4179 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4180 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4181 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4182 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4183 * 4184 * Data returned is just a momentary snapshot of actual values 4185 * and could be inaccurate, so this facility should be used to 4186 * power heuristics and for reporting, not to make 100% correct 4187 * calculation. 4188 * Return 4189 * Requested value, or 0, if *flags* are not recognized. 4190 * 4191 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4192 * Description 4193 * Change the skbs checksum level by one layer up or down, or 4194 * reset it entirely to none in order to have the stack perform 4195 * checksum validation. The level is applicable to the following 4196 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4197 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4198 * through **bpf_skb_adjust_room**\ () helper with passing in 4199 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4200 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4201 * the UDP header is removed. Similarly, an encap of the latter 4202 * into the former could be accompanied by a helper call to 4203 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4204 * skb is still intended to be processed in higher layers of the 4205 * stack instead of just egressing at tc. 4206 * 4207 * There are three supported level settings at this time: 4208 * 4209 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4210 * with CHECKSUM_UNNECESSARY. 4211 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4212 * with CHECKSUM_UNNECESSARY. 4213 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4214 * sets CHECKSUM_NONE to force checksum validation by the stack. 4215 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4216 * skb->csum_level. 4217 * Return 4218 * 0 on success, or a negative error in case of failure. In the 4219 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4220 * is returned or the error code -EACCES in case the skb is not 4221 * subject to CHECKSUM_UNNECESSARY. 4222 * 4223 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4224 * Description 4225 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4226 * Return 4227 * *sk* if casting is valid, or **NULL** otherwise. 4228 * 4229 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4230 * Description 4231 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4232 * Return 4233 * *sk* if casting is valid, or **NULL** otherwise. 4234 * 4235 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4236 * Description 4237 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4238 * Return 4239 * *sk* if casting is valid, or **NULL** otherwise. 4240 * 4241 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4242 * Description 4243 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4244 * Return 4245 * *sk* if casting is valid, or **NULL** otherwise. 4246 * 4247 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4248 * Description 4249 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4250 * Return 4251 * *sk* if casting is valid, or **NULL** otherwise. 4252 * 4253 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4254 * Description 4255 * Return a user or a kernel stack in bpf program provided buffer. 4256 * To achieve this, the helper needs *task*, which is a valid 4257 * pointer to **struct task_struct**. To store the stacktrace, the 4258 * bpf program provides *buf* with a nonnegative *size*. 4259 * 4260 * The last argument, *flags*, holds the number of stack frames to 4261 * skip (from 0 to 255), masked with 4262 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4263 * the following flags: 4264 * 4265 * **BPF_F_USER_STACK** 4266 * Collect a user space stack instead of a kernel stack. 4267 * **BPF_F_USER_BUILD_ID** 4268 * Collect buildid+offset instead of ips for user stack, 4269 * only valid if **BPF_F_USER_STACK** is also specified. 4270 * 4271 * **bpf_get_task_stack**\ () can collect up to 4272 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4273 * to sufficient large buffer size. Note that 4274 * this limit can be controlled with the **sysctl** program, and 4275 * that it should be manually increased in order to profile long 4276 * user stacks (such as stacks for Java programs). To do so, use: 4277 * 4278 * :: 4279 * 4280 * # sysctl kernel.perf_event_max_stack=<new value> 4281 * Return 4282 * A non-negative value equal to or less than *size* on success, 4283 * or a negative error in case of failure. 4284 * 4285 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4286 * Description 4287 * Load header option. Support reading a particular TCP header 4288 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4289 * 4290 * If *flags* is 0, it will search the option from the 4291 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4292 * has details on what skb_data contains under different 4293 * *skops*\ **->op**. 4294 * 4295 * The first byte of the *searchby_res* specifies the 4296 * kind that it wants to search. 4297 * 4298 * If the searching kind is an experimental kind 4299 * (i.e. 253 or 254 according to RFC6994). It also 4300 * needs to specify the "magic" which is either 4301 * 2 bytes or 4 bytes. It then also needs to 4302 * specify the size of the magic by using 4303 * the 2nd byte which is "kind-length" of a TCP 4304 * header option and the "kind-length" also 4305 * includes the first 2 bytes "kind" and "kind-length" 4306 * itself as a normal TCP header option also does. 4307 * 4308 * For example, to search experimental kind 254 with 4309 * 2 byte magic 0xeB9F, the searchby_res should be 4310 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4311 * 4312 * To search for the standard window scale option (3), 4313 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4314 * Note, kind-length must be 0 for regular option. 4315 * 4316 * Searching for No-Op (0) and End-of-Option-List (1) are 4317 * not supported. 4318 * 4319 * *len* must be at least 2 bytes which is the minimal size 4320 * of a header option. 4321 * 4322 * Supported flags: 4323 * 4324 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4325 * saved_syn packet or the just-received syn packet. 4326 * 4327 * Return 4328 * > 0 when found, the header option is copied to *searchby_res*. 4329 * The return value is the total length copied. On failure, a 4330 * negative error code is returned: 4331 * 4332 * **-EINVAL** if a parameter is invalid. 4333 * 4334 * **-ENOMSG** if the option is not found. 4335 * 4336 * **-ENOENT** if no syn packet is available when 4337 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4338 * 4339 * **-ENOSPC** if there is not enough space. Only *len* number of 4340 * bytes are copied. 4341 * 4342 * **-EFAULT** on failure to parse the header options in the 4343 * packet. 4344 * 4345 * **-EPERM** if the helper cannot be used under the current 4346 * *skops*\ **->op**. 4347 * 4348 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4349 * Description 4350 * Store header option. The data will be copied 4351 * from buffer *from* with length *len* to the TCP header. 4352 * 4353 * The buffer *from* should have the whole option that 4354 * includes the kind, kind-length, and the actual 4355 * option data. The *len* must be at least kind-length 4356 * long. The kind-length does not have to be 4 byte 4357 * aligned. The kernel will take care of the padding 4358 * and setting the 4 bytes aligned value to th->doff. 4359 * 4360 * This helper will check for duplicated option 4361 * by searching the same option in the outgoing skb. 4362 * 4363 * This helper can only be called during 4364 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4365 * 4366 * Return 4367 * 0 on success, or negative error in case of failure: 4368 * 4369 * **-EINVAL** If param is invalid. 4370 * 4371 * **-ENOSPC** if there is not enough space in the header. 4372 * Nothing has been written 4373 * 4374 * **-EEXIST** if the option already exists. 4375 * 4376 * **-EFAULT** on failrue to parse the existing header options. 4377 * 4378 * **-EPERM** if the helper cannot be used under the current 4379 * *skops*\ **->op**. 4380 * 4381 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4382 * Description 4383 * Reserve *len* bytes for the bpf header option. The 4384 * space will be used by **bpf_store_hdr_opt**\ () later in 4385 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4386 * 4387 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4388 * the total number of bytes will be reserved. 4389 * 4390 * This helper can only be called during 4391 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4392 * 4393 * Return 4394 * 0 on success, or negative error in case of failure: 4395 * 4396 * **-EINVAL** if a parameter is invalid. 4397 * 4398 * **-ENOSPC** if there is not enough space in the header. 4399 * 4400 * **-EPERM** if the helper cannot be used under the current 4401 * *skops*\ **->op**. 4402 * 4403 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4404 * Description 4405 * Get a bpf_local_storage from an *inode*. 4406 * 4407 * Logically, it could be thought of as getting the value from 4408 * a *map* with *inode* as the **key**. From this 4409 * perspective, the usage is not much different from 4410 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4411 * helper enforces the key must be an inode and the map must also 4412 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4413 * 4414 * Underneath, the value is stored locally at *inode* instead of 4415 * the *map*. The *map* is used as the bpf-local-storage 4416 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4417 * searched against all bpf_local_storage residing at *inode*. 4418 * 4419 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4420 * used such that a new bpf_local_storage will be 4421 * created if one does not exist. *value* can be used 4422 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4423 * the initial value of a bpf_local_storage. If *value* is 4424 * **NULL**, the new bpf_local_storage will be zero initialized. 4425 * Return 4426 * A bpf_local_storage pointer is returned on success. 4427 * 4428 * **NULL** if not found or there was an error in adding 4429 * a new bpf_local_storage. 4430 * 4431 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4432 * Description 4433 * Delete a bpf_local_storage from an *inode*. 4434 * Return 4435 * 0 on success. 4436 * 4437 * **-ENOENT** if the bpf_local_storage cannot be found. 4438 * 4439 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4440 * Description 4441 * Return full path for given **struct path** object, which 4442 * needs to be the kernel BTF *path* object. The path is 4443 * returned in the provided buffer *buf* of size *sz* and 4444 * is zero terminated. 4445 * 4446 * Return 4447 * On success, the strictly positive length of the string, 4448 * including the trailing NUL character. On error, a negative 4449 * value. 4450 * 4451 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4452 * Description 4453 * Read *size* bytes from user space address *user_ptr* and store 4454 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4455 * Return 4456 * 0 on success, or a negative error in case of failure. 4457 * 4458 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4459 * Description 4460 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4461 * using *ptr*->type_id. This value should specify the type 4462 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4463 * can be used to look up vmlinux BTF type ids. Traversing the 4464 * data structure using BTF, the type information and values are 4465 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4466 * the pointer data is carried out to avoid kernel crashes during 4467 * operation. Smaller types can use string space on the stack; 4468 * larger programs can use map data to store the string 4469 * representation. 4470 * 4471 * The string can be subsequently shared with userspace via 4472 * bpf_perf_event_output() or ring buffer interfaces. 4473 * bpf_trace_printk() is to be avoided as it places too small 4474 * a limit on string size to be useful. 4475 * 4476 * *flags* is a combination of 4477 * 4478 * **BTF_F_COMPACT** 4479 * no formatting around type information 4480 * **BTF_F_NONAME** 4481 * no struct/union member names/types 4482 * **BTF_F_PTR_RAW** 4483 * show raw (unobfuscated) pointer values; 4484 * equivalent to printk specifier %px. 4485 * **BTF_F_ZERO** 4486 * show zero-valued struct/union members; they 4487 * are not displayed by default 4488 * 4489 * Return 4490 * The number of bytes that were written (or would have been 4491 * written if output had to be truncated due to string size), 4492 * or a negative error in cases of failure. 4493 * 4494 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4495 * Description 4496 * Use BTF to write to seq_write a string representation of 4497 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4498 * *flags* are identical to those used for bpf_snprintf_btf. 4499 * Return 4500 * 0 on success or a negative error in case of failure. 4501 * 4502 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4503 * Description 4504 * See **bpf_get_cgroup_classid**\ () for the main description. 4505 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4506 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4507 * associated socket instead of the current process. 4508 * Return 4509 * The id is returned or 0 in case the id could not be retrieved. 4510 * 4511 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4512 * Description 4513 * Redirect the packet to another net device of index *ifindex* 4514 * and fill in L2 addresses from neighboring subsystem. This helper 4515 * is somewhat similar to **bpf_redirect**\ (), except that it 4516 * populates L2 addresses as well, meaning, internally, the helper 4517 * relies on the neighbor lookup for the L2 address of the nexthop. 4518 * 4519 * The helper will perform a FIB lookup based on the skb's 4520 * networking header to get the address of the next hop, unless 4521 * this is supplied by the caller in the *params* argument. The 4522 * *plen* argument indicates the len of *params* and should be set 4523 * to 0 if *params* is NULL. 4524 * 4525 * The *flags* argument is reserved and must be 0. The helper is 4526 * currently only supported for tc BPF program types, and enabled 4527 * for IPv4 and IPv6 protocols. 4528 * Return 4529 * The helper returns **TC_ACT_REDIRECT** on success or 4530 * **TC_ACT_SHOT** on error. 4531 * 4532 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4533 * Description 4534 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4535 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4536 * extern variable decorated with '__ksym'. For ksym, there is a 4537 * global var (either static or global) defined of the same name 4538 * in the kernel. The ksym is percpu if the global var is percpu. 4539 * The returned pointer points to the global percpu var on *cpu*. 4540 * 4541 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4542 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4543 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4544 * bpf_per_cpu_ptr() must check the returned value. 4545 * Return 4546 * A pointer pointing to the kernel percpu variable on *cpu*, or 4547 * NULL, if *cpu* is invalid. 4548 * 4549 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4550 * Description 4551 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4552 * pointer to the percpu kernel variable on this cpu. See the 4553 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4554 * 4555 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4556 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4557 * never return NULL. 4558 * Return 4559 * A pointer pointing to the kernel percpu variable on this cpu. 4560 * 4561 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4562 * Description 4563 * Redirect the packet to another net device of index *ifindex*. 4564 * This helper is somewhat similar to **bpf_redirect**\ (), except 4565 * that the redirection happens to the *ifindex*' peer device and 4566 * the netns switch takes place from ingress to ingress without 4567 * going through the CPU's backlog queue. 4568 * 4569 * The *flags* argument is reserved and must be 0. The helper is 4570 * currently only supported for tc BPF program types at the ingress 4571 * hook and for veth device types. The peer device must reside in a 4572 * different network namespace. 4573 * Return 4574 * The helper returns **TC_ACT_REDIRECT** on success or 4575 * **TC_ACT_SHOT** on error. 4576 * 4577 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4578 * Description 4579 * Get a bpf_local_storage from the *task*. 4580 * 4581 * Logically, it could be thought of as getting the value from 4582 * a *map* with *task* as the **key**. From this 4583 * perspective, the usage is not much different from 4584 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4585 * helper enforces the key must be an task_struct and the map must also 4586 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4587 * 4588 * Underneath, the value is stored locally at *task* instead of 4589 * the *map*. The *map* is used as the bpf-local-storage 4590 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4591 * searched against all bpf_local_storage residing at *task*. 4592 * 4593 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4594 * used such that a new bpf_local_storage will be 4595 * created if one does not exist. *value* can be used 4596 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4597 * the initial value of a bpf_local_storage. If *value* is 4598 * **NULL**, the new bpf_local_storage will be zero initialized. 4599 * Return 4600 * A bpf_local_storage pointer is returned on success. 4601 * 4602 * **NULL** if not found or there was an error in adding 4603 * a new bpf_local_storage. 4604 * 4605 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4606 * Description 4607 * Delete a bpf_local_storage from a *task*. 4608 * Return 4609 * 0 on success. 4610 * 4611 * **-ENOENT** if the bpf_local_storage cannot be found. 4612 * 4613 * struct task_struct *bpf_get_current_task_btf(void) 4614 * Description 4615 * Return a BTF pointer to the "current" task. 4616 * This pointer can also be used in helpers that accept an 4617 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4618 * Return 4619 * Pointer to the current task. 4620 * 4621 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4622 * Description 4623 * Set or clear certain options on *bprm*: 4624 * 4625 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4626 * which sets the **AT_SECURE** auxv for glibc. The bit 4627 * is cleared if the flag is not specified. 4628 * Return 4629 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4630 * 4631 * u64 bpf_ktime_get_coarse_ns(void) 4632 * Description 4633 * Return a coarse-grained version of the time elapsed since 4634 * system boot, in nanoseconds. Does not include time the system 4635 * was suspended. 4636 * 4637 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4638 * Return 4639 * Current *ktime*. 4640 * 4641 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4642 * Description 4643 * Returns the stored IMA hash of the *inode* (if it's avaialable). 4644 * If the hash is larger than *size*, then only *size* 4645 * bytes will be copied to *dst* 4646 * Return 4647 * The **hash_algo** is returned on success, 4648 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4649 * invalid arguments are passed. 4650 * 4651 * struct socket *bpf_sock_from_file(struct file *file) 4652 * Description 4653 * If the given file represents a socket, returns the associated 4654 * socket. 4655 * Return 4656 * A pointer to a struct socket on success or NULL if the file is 4657 * not a socket. 4658 * 4659 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4660 * Description 4661 * Check packet size against exceeding MTU of net device (based 4662 * on *ifindex*). This helper will likely be used in combination 4663 * with helpers that adjust/change the packet size. 4664 * 4665 * The argument *len_diff* can be used for querying with a planned 4666 * size change. This allows to check MTU prior to changing packet 4667 * ctx. Providing an *len_diff* adjustment that is larger than the 4668 * actual packet size (resulting in negative packet size) will in 4669 * principle not exceed the MTU, why it is not considered a 4670 * failure. Other BPF-helpers are needed for performing the 4671 * planned size change, why the responsability for catch a negative 4672 * packet size belong in those helpers. 4673 * 4674 * Specifying *ifindex* zero means the MTU check is performed 4675 * against the current net device. This is practical if this isn't 4676 * used prior to redirect. 4677 * 4678 * On input *mtu_len* must be a valid pointer, else verifier will 4679 * reject BPF program. If the value *mtu_len* is initialized to 4680 * zero then the ctx packet size is use. When value *mtu_len* is 4681 * provided as input this specify the L3 length that the MTU check 4682 * is done against. Remember XDP and TC length operate at L2, but 4683 * this value is L3 as this correlate to MTU and IP-header tot_len 4684 * values which are L3 (similar behavior as bpf_fib_lookup). 4685 * 4686 * The Linux kernel route table can configure MTUs on a more 4687 * specific per route level, which is not provided by this helper. 4688 * For route level MTU checks use the **bpf_fib_lookup**\ () 4689 * helper. 4690 * 4691 * *ctx* is either **struct xdp_md** for XDP programs or 4692 * **struct sk_buff** for tc cls_act programs. 4693 * 4694 * The *flags* argument can be a combination of one or more of the 4695 * following values: 4696 * 4697 * **BPF_MTU_CHK_SEGS** 4698 * This flag will only works for *ctx* **struct sk_buff**. 4699 * If packet context contains extra packet segment buffers 4700 * (often knows as GSO skb), then MTU check is harder to 4701 * check at this point, because in transmit path it is 4702 * possible for the skb packet to get re-segmented 4703 * (depending on net device features). This could still be 4704 * a MTU violation, so this flag enables performing MTU 4705 * check against segments, with a different violation 4706 * return code to tell it apart. Check cannot use len_diff. 4707 * 4708 * On return *mtu_len* pointer contains the MTU value of the net 4709 * device. Remember the net device configured MTU is the L3 size, 4710 * which is returned here and XDP and TC length operate at L2. 4711 * Helper take this into account for you, but remember when using 4712 * MTU value in your BPF-code. 4713 * 4714 * Return 4715 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4716 * 4717 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4718 * 4719 * MTU violations return positive values, but also populate MTU 4720 * value in *mtu_len* pointer, as this can be needed for 4721 * implementing PMTU handing: 4722 * 4723 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4724 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4725 * 4726 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4727 * Description 4728 * For each element in **map**, call **callback_fn** function with 4729 * **map**, **callback_ctx** and other map-specific parameters. 4730 * The **callback_fn** should be a static function and 4731 * the **callback_ctx** should be a pointer to the stack. 4732 * The **flags** is used to control certain aspects of the helper. 4733 * Currently, the **flags** must be 0. 4734 * 4735 * The following are a list of supported map types and their 4736 * respective expected callback signatures: 4737 * 4738 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4739 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4740 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4741 * 4742 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4743 * 4744 * For per_cpu maps, the map_value is the value on the cpu where the 4745 * bpf_prog is running. 4746 * 4747 * If **callback_fn** return 0, the helper will continue to the next 4748 * element. If return value is 1, the helper will skip the rest of 4749 * elements and return. Other return values are not used now. 4750 * 4751 * Return 4752 * The number of traversed map elements for success, **-EINVAL** for 4753 * invalid **flags**. 4754 * 4755 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4756 * Description 4757 * Outputs a string into the **str** buffer of size **str_size** 4758 * based on a format string stored in a read-only map pointed by 4759 * **fmt**. 4760 * 4761 * Each format specifier in **fmt** corresponds to one u64 element 4762 * in the **data** array. For strings and pointers where pointees 4763 * are accessed, only the pointer values are stored in the *data* 4764 * array. The *data_len* is the size of *data* in bytes - must be 4765 * a multiple of 8. 4766 * 4767 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4768 * memory. Reading kernel memory may fail due to either invalid 4769 * address or valid address but requiring a major memory fault. If 4770 * reading kernel memory fails, the string for **%s** will be an 4771 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4772 * Not returning error to bpf program is consistent with what 4773 * **bpf_trace_printk**\ () does for now. 4774 * 4775 * Return 4776 * The strictly positive length of the formatted string, including 4777 * the trailing zero character. If the return value is greater than 4778 * **str_size**, **str** contains a truncated string, guaranteed to 4779 * be zero-terminated except when **str_size** is 0. 4780 * 4781 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 4782 * 4783 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 4784 * Description 4785 * Execute bpf syscall with given arguments. 4786 * Return 4787 * A syscall result. 4788 * 4789 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 4790 * Description 4791 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 4792 * Return 4793 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 4794 * 4795 * long bpf_sys_close(u32 fd) 4796 * Description 4797 * Execute close syscall for given FD. 4798 * Return 4799 * A syscall result. 4800 * 4801 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4802 * Description 4803 * Initialize the timer. 4804 * First 4 bits of *flags* specify clockid. 4805 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4806 * All other bits of *flags* are reserved. 4807 * The verifier will reject the program if *timer* is not from 4808 * the same *map*. 4809 * Return 4810 * 0 on success. 4811 * **-EBUSY** if *timer* is already initialized. 4812 * **-EINVAL** if invalid *flags* are passed. 4813 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4814 * The user space should either hold a file descriptor to a map with timers 4815 * or pin such map in bpffs. When map is unpinned or file descriptor is 4816 * closed all timers in the map will be cancelled and freed. 4817 * 4818 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4819 * Description 4820 * Configure the timer to call *callback_fn* static function. 4821 * Return 4822 * 0 on success. 4823 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4824 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4825 * The user space should either hold a file descriptor to a map with timers 4826 * or pin such map in bpffs. When map is unpinned or file descriptor is 4827 * closed all timers in the map will be cancelled and freed. 4828 * 4829 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4830 * Description 4831 * Set timer expiration N nanoseconds from the current time. The 4832 * configured callback will be invoked in soft irq context on some cpu 4833 * and will not repeat unless another bpf_timer_start() is made. 4834 * In such case the next invocation can migrate to a different cpu. 4835 * Since struct bpf_timer is a field inside map element the map 4836 * owns the timer. The bpf_timer_set_callback() will increment refcnt 4837 * of BPF program to make sure that callback_fn code stays valid. 4838 * When user space reference to a map reaches zero all timers 4839 * in a map are cancelled and corresponding program's refcnts are 4840 * decremented. This is done to make sure that Ctrl-C of a user 4841 * process doesn't leave any timers running. If map is pinned in 4842 * bpffs the callback_fn can re-arm itself indefinitely. 4843 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4844 * cancel and free the timer in the given map element. 4845 * The map can contain timers that invoke callback_fn-s from different 4846 * programs. The same callback_fn can serve different timers from 4847 * different maps if key/value layout matches across maps. 4848 * Every bpf_timer_set_callback() can have different callback_fn. 4849 * 4850 * Return 4851 * 0 on success. 4852 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4853 * or invalid *flags* are passed. 4854 * 4855 * long bpf_timer_cancel(struct bpf_timer *timer) 4856 * Description 4857 * Cancel the timer and wait for callback_fn to finish if it was running. 4858 * Return 4859 * 0 if the timer was not active. 4860 * 1 if the timer was active. 4861 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4862 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4863 * own timer which would have led to a deadlock otherwise. 4864 * 4865 * u64 bpf_get_func_ip(void *ctx) 4866 * Description 4867 * Get address of the traced function (for tracing and kprobe programs). 4868 * Return 4869 * Address of the traced function. 4870 * 4871 * u64 bpf_get_attach_cookie(void *ctx) 4872 * Description 4873 * Get bpf_cookie value provided (optionally) during the program 4874 * attachment. It might be different for each individual 4875 * attachment, even if BPF program itself is the same. 4876 * Expects BPF program context *ctx* as a first argument. 4877 * 4878 * Supported for the following program types: 4879 * - kprobe/uprobe; 4880 * - tracepoint; 4881 * - perf_event. 4882 * Return 4883 * Value specified by user at BPF link creation/attachment time 4884 * or 0, if it was not specified. 4885 * 4886 * long bpf_task_pt_regs(struct task_struct *task) 4887 * Description 4888 * Get the struct pt_regs associated with **task**. 4889 * Return 4890 * A pointer to struct pt_regs. 4891 * 4892 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 4893 * Description 4894 * Get branch trace from hardware engines like Intel LBR. The 4895 * hardware engine is stopped shortly after the helper is 4896 * called. Therefore, the user need to filter branch entries 4897 * based on the actual use case. To capture branch trace 4898 * before the trigger point of the BPF program, the helper 4899 * should be called at the beginning of the BPF program. 4900 * 4901 * The data is stored as struct perf_branch_entry into output 4902 * buffer *entries*. *size* is the size of *entries* in bytes. 4903 * *flags* is reserved for now and must be zero. 4904 * 4905 * Return 4906 * On success, number of bytes written to *buf*. On error, a 4907 * negative value. 4908 * 4909 * **-EINVAL** if *flags* is not zero. 4910 * 4911 * **-ENOENT** if architecture does not support branch records. 4912 * 4913 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4914 * Description 4915 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 4916 * to format and can handle more format args as a result. 4917 * 4918 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 4919 * Return 4920 * The number of bytes written to the buffer, or a negative error 4921 * in case of failure. 4922 * 4923 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 4924 * Description 4925 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 4926 * Return 4927 * *sk* if casting is valid, or **NULL** otherwise. 4928 * 4929 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 4930 * Description 4931 * Get the address of a kernel symbol, returned in *res*. *res* is 4932 * set to 0 if the symbol is not found. 4933 * Return 4934 * On success, zero. On error, a negative value. 4935 * 4936 * **-EINVAL** if *flags* is not zero. 4937 * 4938 * **-EINVAL** if string *name* is not the same size as *name_sz*. 4939 * 4940 * **-ENOENT** if symbol is not found. 4941 * 4942 * **-EPERM** if caller does not have permission to obtain kernel address. 4943 * 4944 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 4945 * Description 4946 * Find vma of *task* that contains *addr*, call *callback_fn* 4947 * function with *task*, *vma*, and *callback_ctx*. 4948 * The *callback_fn* should be a static function and 4949 * the *callback_ctx* should be a pointer to the stack. 4950 * The *flags* is used to control certain aspects of the helper. 4951 * Currently, the *flags* must be 0. 4952 * 4953 * The expected callback signature is 4954 * 4955 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 4956 * 4957 * Return 4958 * 0 on success. 4959 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 4960 * **-EBUSY** if failed to try lock mmap_lock. 4961 * **-EINVAL** for invalid **flags**. 4962 * 4963 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 4964 * Description 4965 * For **nr_loops**, call **callback_fn** function 4966 * with **callback_ctx** as the context parameter. 4967 * The **callback_fn** should be a static function and 4968 * the **callback_ctx** should be a pointer to the stack. 4969 * The **flags** is used to control certain aspects of the helper. 4970 * Currently, the **flags** must be 0. Currently, nr_loops is 4971 * limited to 1 << 23 (~8 million) loops. 4972 * 4973 * long (\*callback_fn)(u32 index, void \*ctx); 4974 * 4975 * where **index** is the current index in the loop. The index 4976 * is zero-indexed. 4977 * 4978 * If **callback_fn** returns 0, the helper will continue to the next 4979 * loop. If return value is 1, the helper will skip the rest of 4980 * the loops and return. Other return values are not used now, 4981 * and will be rejected by the verifier. 4982 * 4983 * Return 4984 * The number of loops performed, **-EINVAL** for invalid **flags**, 4985 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 4986 * 4987 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 4988 * Description 4989 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 4990 * to be null-terminated and **s1_sz** is the maximum storage 4991 * size of **s1**. **s2** must be a read-only string. 4992 * Return 4993 * An integer less than, equal to, or greater than zero 4994 * if the first **s1_sz** bytes of **s1** is found to be 4995 * less than, to match, or be greater than **s2**. 4996 * 4997 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 4998 * Description 4999 * Get **n**-th argument (zero based) of the traced function (for tracing programs) 5000 * returned in **value**. 5001 * 5002 * Return 5003 * 0 on success. 5004 * **-EINVAL** if n >= arguments count of traced function. 5005 * 5006 * long bpf_get_func_ret(void *ctx, u64 *value) 5007 * Description 5008 * Get return value of the traced function (for tracing programs) 5009 * in **value**. 5010 * 5011 * Return 5012 * 0 on success. 5013 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5014 * 5015 * long bpf_get_func_arg_cnt(void *ctx) 5016 * Description 5017 * Get number of arguments of the traced function (for tracing programs). 5018 * 5019 * Return 5020 * The number of arguments of the traced function. 5021 */ 5022#define __BPF_FUNC_MAPPER(FN) \ 5023 FN(unspec), \ 5024 FN(map_lookup_elem), \ 5025 FN(map_update_elem), \ 5026 FN(map_delete_elem), \ 5027 FN(probe_read), \ 5028 FN(ktime_get_ns), \ 5029 FN(trace_printk), \ 5030 FN(get_prandom_u32), \ 5031 FN(get_smp_processor_id), \ 5032 FN(skb_store_bytes), \ 5033 FN(l3_csum_replace), \ 5034 FN(l4_csum_replace), \ 5035 FN(tail_call), \ 5036 FN(clone_redirect), \ 5037 FN(get_current_pid_tgid), \ 5038 FN(get_current_uid_gid), \ 5039 FN(get_current_comm), \ 5040 FN(get_cgroup_classid), \ 5041 FN(skb_vlan_push), \ 5042 FN(skb_vlan_pop), \ 5043 FN(skb_get_tunnel_key), \ 5044 FN(skb_set_tunnel_key), \ 5045 FN(perf_event_read), \ 5046 FN(redirect), \ 5047 FN(get_route_realm), \ 5048 FN(perf_event_output), \ 5049 FN(skb_load_bytes), \ 5050 FN(get_stackid), \ 5051 FN(csum_diff), \ 5052 FN(skb_get_tunnel_opt), \ 5053 FN(skb_set_tunnel_opt), \ 5054 FN(skb_change_proto), \ 5055 FN(skb_change_type), \ 5056 FN(skb_under_cgroup), \ 5057 FN(get_hash_recalc), \ 5058 FN(get_current_task), \ 5059 FN(probe_write_user), \ 5060 FN(current_task_under_cgroup), \ 5061 FN(skb_change_tail), \ 5062 FN(skb_pull_data), \ 5063 FN(csum_update), \ 5064 FN(set_hash_invalid), \ 5065 FN(get_numa_node_id), \ 5066 FN(skb_change_head), \ 5067 FN(xdp_adjust_head), \ 5068 FN(probe_read_str), \ 5069 FN(get_socket_cookie), \ 5070 FN(get_socket_uid), \ 5071 FN(set_hash), \ 5072 FN(setsockopt), \ 5073 FN(skb_adjust_room), \ 5074 FN(redirect_map), \ 5075 FN(sk_redirect_map), \ 5076 FN(sock_map_update), \ 5077 FN(xdp_adjust_meta), \ 5078 FN(perf_event_read_value), \ 5079 FN(perf_prog_read_value), \ 5080 FN(getsockopt), \ 5081 FN(override_return), \ 5082 FN(sock_ops_cb_flags_set), \ 5083 FN(msg_redirect_map), \ 5084 FN(msg_apply_bytes), \ 5085 FN(msg_cork_bytes), \ 5086 FN(msg_pull_data), \ 5087 FN(bind), \ 5088 FN(xdp_adjust_tail), \ 5089 FN(skb_get_xfrm_state), \ 5090 FN(get_stack), \ 5091 FN(skb_load_bytes_relative), \ 5092 FN(fib_lookup), \ 5093 FN(sock_hash_update), \ 5094 FN(msg_redirect_hash), \ 5095 FN(sk_redirect_hash), \ 5096 FN(lwt_push_encap), \ 5097 FN(lwt_seg6_store_bytes), \ 5098 FN(lwt_seg6_adjust_srh), \ 5099 FN(lwt_seg6_action), \ 5100 FN(rc_repeat), \ 5101 FN(rc_keydown), \ 5102 FN(skb_cgroup_id), \ 5103 FN(get_current_cgroup_id), \ 5104 FN(get_local_storage), \ 5105 FN(sk_select_reuseport), \ 5106 FN(skb_ancestor_cgroup_id), \ 5107 FN(sk_lookup_tcp), \ 5108 FN(sk_lookup_udp), \ 5109 FN(sk_release), \ 5110 FN(map_push_elem), \ 5111 FN(map_pop_elem), \ 5112 FN(map_peek_elem), \ 5113 FN(msg_push_data), \ 5114 FN(msg_pop_data), \ 5115 FN(rc_pointer_rel), \ 5116 FN(spin_lock), \ 5117 FN(spin_unlock), \ 5118 FN(sk_fullsock), \ 5119 FN(tcp_sock), \ 5120 FN(skb_ecn_set_ce), \ 5121 FN(get_listener_sock), \ 5122 FN(skc_lookup_tcp), \ 5123 FN(tcp_check_syncookie), \ 5124 FN(sysctl_get_name), \ 5125 FN(sysctl_get_current_value), \ 5126 FN(sysctl_get_new_value), \ 5127 FN(sysctl_set_new_value), \ 5128 FN(strtol), \ 5129 FN(strtoul), \ 5130 FN(sk_storage_get), \ 5131 FN(sk_storage_delete), \ 5132 FN(send_signal), \ 5133 FN(tcp_gen_syncookie), \ 5134 FN(skb_output), \ 5135 FN(probe_read_user), \ 5136 FN(probe_read_kernel), \ 5137 FN(probe_read_user_str), \ 5138 FN(probe_read_kernel_str), \ 5139 FN(tcp_send_ack), \ 5140 FN(send_signal_thread), \ 5141 FN(jiffies64), \ 5142 FN(read_branch_records), \ 5143 FN(get_ns_current_pid_tgid), \ 5144 FN(xdp_output), \ 5145 FN(get_netns_cookie), \ 5146 FN(get_current_ancestor_cgroup_id), \ 5147 FN(sk_assign), \ 5148 FN(ktime_get_boot_ns), \ 5149 FN(seq_printf), \ 5150 FN(seq_write), \ 5151 FN(sk_cgroup_id), \ 5152 FN(sk_ancestor_cgroup_id), \ 5153 FN(ringbuf_output), \ 5154 FN(ringbuf_reserve), \ 5155 FN(ringbuf_submit), \ 5156 FN(ringbuf_discard), \ 5157 FN(ringbuf_query), \ 5158 FN(csum_level), \ 5159 FN(skc_to_tcp6_sock), \ 5160 FN(skc_to_tcp_sock), \ 5161 FN(skc_to_tcp_timewait_sock), \ 5162 FN(skc_to_tcp_request_sock), \ 5163 FN(skc_to_udp6_sock), \ 5164 FN(get_task_stack), \ 5165 FN(load_hdr_opt), \ 5166 FN(store_hdr_opt), \ 5167 FN(reserve_hdr_opt), \ 5168 FN(inode_storage_get), \ 5169 FN(inode_storage_delete), \ 5170 FN(d_path), \ 5171 FN(copy_from_user), \ 5172 FN(snprintf_btf), \ 5173 FN(seq_printf_btf), \ 5174 FN(skb_cgroup_classid), \ 5175 FN(redirect_neigh), \ 5176 FN(per_cpu_ptr), \ 5177 FN(this_cpu_ptr), \ 5178 FN(redirect_peer), \ 5179 FN(task_storage_get), \ 5180 FN(task_storage_delete), \ 5181 FN(get_current_task_btf), \ 5182 FN(bprm_opts_set), \ 5183 FN(ktime_get_coarse_ns), \ 5184 FN(ima_inode_hash), \ 5185 FN(sock_from_file), \ 5186 FN(check_mtu), \ 5187 FN(for_each_map_elem), \ 5188 FN(snprintf), \ 5189 FN(sys_bpf), \ 5190 FN(btf_find_by_name_kind), \ 5191 FN(sys_close), \ 5192 FN(timer_init), \ 5193 FN(timer_set_callback), \ 5194 FN(timer_start), \ 5195 FN(timer_cancel), \ 5196 FN(get_func_ip), \ 5197 FN(get_attach_cookie), \ 5198 FN(task_pt_regs), \ 5199 FN(get_branch_snapshot), \ 5200 FN(trace_vprintk), \ 5201 FN(skc_to_unix_sock), \ 5202 FN(kallsyms_lookup_name), \ 5203 FN(find_vma), \ 5204 FN(loop), \ 5205 FN(strncmp), \ 5206 FN(get_func_arg), \ 5207 FN(get_func_ret), \ 5208 FN(get_func_arg_cnt), \ 5209 /* */ 5210 5211/* integer value in 'imm' field of BPF_CALL instruction selects which helper 5212 * function eBPF program intends to call 5213 */ 5214#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5215enum bpf_func_id { 5216 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5217 __BPF_FUNC_MAX_ID, 5218}; 5219#undef __BPF_ENUM_FN 5220 5221/* All flags used by eBPF helper functions, placed here. */ 5222 5223/* BPF_FUNC_skb_store_bytes flags. */ 5224enum { 5225 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5226 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5227}; 5228 5229/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5230 * First 4 bits are for passing the header field size. 5231 */ 5232enum { 5233 BPF_F_HDR_FIELD_MASK = 0xfULL, 5234}; 5235 5236/* BPF_FUNC_l4_csum_replace flags. */ 5237enum { 5238 BPF_F_PSEUDO_HDR = (1ULL << 4), 5239 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5240 BPF_F_MARK_ENFORCE = (1ULL << 6), 5241}; 5242 5243/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5244enum { 5245 BPF_F_INGRESS = (1ULL << 0), 5246}; 5247 5248/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5249enum { 5250 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5251}; 5252 5253/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5254enum { 5255 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5256 BPF_F_USER_STACK = (1ULL << 8), 5257/* flags used by BPF_FUNC_get_stackid only. */ 5258 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5259 BPF_F_REUSE_STACKID = (1ULL << 10), 5260/* flags used by BPF_FUNC_get_stack only. */ 5261 BPF_F_USER_BUILD_ID = (1ULL << 11), 5262}; 5263 5264/* BPF_FUNC_skb_set_tunnel_key flags. */ 5265enum { 5266 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5267 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5268 BPF_F_SEQ_NUMBER = (1ULL << 3), 5269}; 5270 5271/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5272 * BPF_FUNC_perf_event_read_value flags. 5273 */ 5274enum { 5275 BPF_F_INDEX_MASK = 0xffffffffULL, 5276 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5277/* BPF_FUNC_perf_event_output for sk_buff input context. */ 5278 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5279}; 5280 5281/* Current network namespace */ 5282enum { 5283 BPF_F_CURRENT_NETNS = (-1L), 5284}; 5285 5286/* BPF_FUNC_csum_level level values. */ 5287enum { 5288 BPF_CSUM_LEVEL_QUERY, 5289 BPF_CSUM_LEVEL_INC, 5290 BPF_CSUM_LEVEL_DEC, 5291 BPF_CSUM_LEVEL_RESET, 5292}; 5293 5294/* BPF_FUNC_skb_adjust_room flags. */ 5295enum { 5296 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5297 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5298 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5299 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5300 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5301 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5302 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5303}; 5304 5305enum { 5306 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5307 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5308}; 5309 5310#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5311 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5312 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5313 5314/* BPF_FUNC_sysctl_get_name flags. */ 5315enum { 5316 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5317}; 5318 5319/* BPF_FUNC_<kernel_obj>_storage_get flags */ 5320enum { 5321 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5322 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5323 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5324 */ 5325 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5326}; 5327 5328/* BPF_FUNC_read_branch_records flags. */ 5329enum { 5330 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5331}; 5332 5333/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5334 * BPF_FUNC_bpf_ringbuf_output flags. 5335 */ 5336enum { 5337 BPF_RB_NO_WAKEUP = (1ULL << 0), 5338 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5339}; 5340 5341/* BPF_FUNC_bpf_ringbuf_query flags */ 5342enum { 5343 BPF_RB_AVAIL_DATA = 0, 5344 BPF_RB_RING_SIZE = 1, 5345 BPF_RB_CONS_POS = 2, 5346 BPF_RB_PROD_POS = 3, 5347}; 5348 5349/* BPF ring buffer constants */ 5350enum { 5351 BPF_RINGBUF_BUSY_BIT = (1U << 31), 5352 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5353 BPF_RINGBUF_HDR_SZ = 8, 5354}; 5355 5356/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5357enum { 5358 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5359 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5360}; 5361 5362/* Mode for BPF_FUNC_skb_adjust_room helper. */ 5363enum bpf_adj_room_mode { 5364 BPF_ADJ_ROOM_NET, 5365 BPF_ADJ_ROOM_MAC, 5366}; 5367 5368/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 5369enum bpf_hdr_start_off { 5370 BPF_HDR_START_MAC, 5371 BPF_HDR_START_NET, 5372}; 5373 5374/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5375enum bpf_lwt_encap_mode { 5376 BPF_LWT_ENCAP_SEG6, 5377 BPF_LWT_ENCAP_SEG6_INLINE, 5378 BPF_LWT_ENCAP_IP, 5379}; 5380 5381/* Flags for bpf_bprm_opts_set helper */ 5382enum { 5383 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 5384}; 5385 5386/* Flags for bpf_redirect_map helper */ 5387enum { 5388 BPF_F_BROADCAST = (1ULL << 3), 5389 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5390}; 5391 5392#define __bpf_md_ptr(type, name) \ 5393union { \ 5394 type name; \ 5395 __u64 :64; \ 5396} __attribute__((aligned(8))) 5397 5398/* user accessible mirror of in-kernel sk_buff. 5399 * new fields can only be added to the end of this structure 5400 */ 5401struct __sk_buff { 5402 __u32 len; 5403 __u32 pkt_type; 5404 __u32 mark; 5405 __u32 queue_mapping; 5406 __u32 protocol; 5407 __u32 vlan_present; 5408 __u32 vlan_tci; 5409 __u32 vlan_proto; 5410 __u32 priority; 5411 __u32 ingress_ifindex; 5412 __u32 ifindex; 5413 __u32 tc_index; 5414 __u32 cb[5]; 5415 __u32 hash; 5416 __u32 tc_classid; 5417 __u32 data; 5418 __u32 data_end; 5419 __u32 napi_id; 5420 5421 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 5422 __u32 family; 5423 __u32 remote_ip4; /* Stored in network byte order */ 5424 __u32 local_ip4; /* Stored in network byte order */ 5425 __u32 remote_ip6[4]; /* Stored in network byte order */ 5426 __u32 local_ip6[4]; /* Stored in network byte order */ 5427 __u32 remote_port; /* Stored in network byte order */ 5428 __u32 local_port; /* stored in host byte order */ 5429 /* ... here. */ 5430 5431 __u32 data_meta; 5432 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5433 __u64 tstamp; 5434 __u32 wire_len; 5435 __u32 gso_segs; 5436 __bpf_md_ptr(struct bpf_sock *, sk); 5437 __u32 gso_size; 5438 __u32 :32; /* Padding, future use. */ 5439 __u64 hwtstamp; 5440}; 5441 5442struct bpf_tunnel_key { 5443 __u32 tunnel_id; 5444 union { 5445 __u32 remote_ipv4; 5446 __u32 remote_ipv6[4]; 5447 }; 5448 __u8 tunnel_tos; 5449 __u8 tunnel_ttl; 5450 __u16 tunnel_ext; /* Padding, future use. */ 5451 __u32 tunnel_label; 5452}; 5453 5454/* user accessible mirror of in-kernel xfrm_state. 5455 * new fields can only be added to the end of this structure 5456 */ 5457struct bpf_xfrm_state { 5458 __u32 reqid; 5459 __u32 spi; /* Stored in network byte order */ 5460 __u16 family; 5461 __u16 ext; /* Padding, future use. */ 5462 union { 5463 __u32 remote_ipv4; /* Stored in network byte order */ 5464 __u32 remote_ipv6[4]; /* Stored in network byte order */ 5465 }; 5466}; 5467 5468/* Generic BPF return codes which all BPF program types may support. 5469 * The values are binary compatible with their TC_ACT_* counter-part to 5470 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 5471 * programs. 5472 * 5473 * XDP is handled seprately, see XDP_*. 5474 */ 5475enum bpf_ret_code { 5476 BPF_OK = 0, 5477 /* 1 reserved */ 5478 BPF_DROP = 2, 5479 /* 3-6 reserved */ 5480 BPF_REDIRECT = 7, 5481 /* >127 are reserved for prog type specific return codes. 5482 * 5483 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 5484 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 5485 * changed and should be routed based on its new L3 header. 5486 * (This is an L3 redirect, as opposed to L2 redirect 5487 * represented by BPF_REDIRECT above). 5488 */ 5489 BPF_LWT_REROUTE = 128, 5490}; 5491 5492struct bpf_sock { 5493 __u32 bound_dev_if; 5494 __u32 family; 5495 __u32 type; 5496 __u32 protocol; 5497 __u32 mark; 5498 __u32 priority; 5499 /* IP address also allows 1 and 2 bytes access */ 5500 __u32 src_ip4; 5501 __u32 src_ip6[4]; 5502 __u32 src_port; /* host byte order */ 5503 __u32 dst_port; /* network byte order */ 5504 __u32 dst_ip4; 5505 __u32 dst_ip6[4]; 5506 __u32 state; 5507 __s32 rx_queue_mapping; 5508}; 5509 5510struct bpf_tcp_sock { 5511 __u32 snd_cwnd; /* Sending congestion window */ 5512 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5513 __u32 rtt_min; 5514 __u32 snd_ssthresh; /* Slow start size threshold */ 5515 __u32 rcv_nxt; /* What we want to receive next */ 5516 __u32 snd_nxt; /* Next sequence we send */ 5517 __u32 snd_una; /* First byte we want an ack for */ 5518 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5519 __u32 ecn_flags; /* ECN status bits. */ 5520 __u32 rate_delivered; /* saved rate sample: packets delivered */ 5521 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5522 __u32 packets_out; /* Packets which are "in flight" */ 5523 __u32 retrans_out; /* Retransmitted packets out */ 5524 __u32 total_retrans; /* Total retransmits for entire connection */ 5525 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5526 * total number of segments in. 5527 */ 5528 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5529 * total number of data segments in. 5530 */ 5531 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5532 * The total number of segments sent. 5533 */ 5534 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5535 * total number of data segments sent. 5536 */ 5537 __u32 lost_out; /* Lost packets */ 5538 __u32 sacked_out; /* SACK'd packets */ 5539 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5540 * sum(delta(rcv_nxt)), or how many bytes 5541 * were acked. 5542 */ 5543 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5544 * sum(delta(snd_una)), or how many bytes 5545 * were acked. 5546 */ 5547 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 5548 * total number of DSACK blocks received 5549 */ 5550 __u32 delivered; /* Total data packets delivered incl. rexmits */ 5551 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5552 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5553}; 5554 5555struct bpf_sock_tuple { 5556 union { 5557 struct { 5558 __be32 saddr; 5559 __be32 daddr; 5560 __be16 sport; 5561 __be16 dport; 5562 } ipv4; 5563 struct { 5564 __be32 saddr[4]; 5565 __be32 daddr[4]; 5566 __be16 sport; 5567 __be16 dport; 5568 } ipv6; 5569 }; 5570}; 5571 5572struct bpf_xdp_sock { 5573 __u32 queue_id; 5574}; 5575 5576#define XDP_PACKET_HEADROOM 256 5577 5578/* User return codes for XDP prog type. 5579 * A valid XDP program must return one of these defined values. All other 5580 * return codes are reserved for future use. Unknown return codes will 5581 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 5582 */ 5583enum xdp_action { 5584 XDP_ABORTED = 0, 5585 XDP_DROP, 5586 XDP_PASS, 5587 XDP_TX, 5588 XDP_REDIRECT, 5589}; 5590 5591/* user accessible metadata for XDP packet hook 5592 * new fields must be added to the end of this structure 5593 */ 5594struct xdp_md { 5595 __u32 data; 5596 __u32 data_end; 5597 __u32 data_meta; 5598 /* Below access go through struct xdp_rxq_info */ 5599 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 5600 __u32 rx_queue_index; /* rxq->queue_index */ 5601 5602 __u32 egress_ifindex; /* txq->dev->ifindex */ 5603}; 5604 5605/* DEVMAP map-value layout 5606 * 5607 * The struct data-layout of map-value is a configuration interface. 5608 * New members can only be added to the end of this structure. 5609 */ 5610struct bpf_devmap_val { 5611 __u32 ifindex; /* device index */ 5612 union { 5613 int fd; /* prog fd on map write */ 5614 __u32 id; /* prog id on map read */ 5615 } bpf_prog; 5616}; 5617 5618/* CPUMAP map-value layout 5619 * 5620 * The struct data-layout of map-value is a configuration interface. 5621 * New members can only be added to the end of this structure. 5622 */ 5623struct bpf_cpumap_val { 5624 __u32 qsize; /* queue size to remote target CPU */ 5625 union { 5626 int fd; /* prog fd on map write */ 5627 __u32 id; /* prog id on map read */ 5628 } bpf_prog; 5629}; 5630 5631enum sk_action { 5632 SK_DROP = 0, 5633 SK_PASS, 5634}; 5635 5636/* user accessible metadata for SK_MSG packet hook, new fields must 5637 * be added to the end of this structure 5638 */ 5639struct sk_msg_md { 5640 __bpf_md_ptr(void *, data); 5641 __bpf_md_ptr(void *, data_end); 5642 5643 __u32 family; 5644 __u32 remote_ip4; /* Stored in network byte order */ 5645 __u32 local_ip4; /* Stored in network byte order */ 5646 __u32 remote_ip6[4]; /* Stored in network byte order */ 5647 __u32 local_ip6[4]; /* Stored in network byte order */ 5648 __u32 remote_port; /* Stored in network byte order */ 5649 __u32 local_port; /* stored in host byte order */ 5650 __u32 size; /* Total size of sk_msg */ 5651 5652 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 5653}; 5654 5655struct sk_reuseport_md { 5656 /* 5657 * Start of directly accessible data. It begins from 5658 * the tcp/udp header. 5659 */ 5660 __bpf_md_ptr(void *, data); 5661 /* End of directly accessible data */ 5662 __bpf_md_ptr(void *, data_end); 5663 /* 5664 * Total length of packet (starting from the tcp/udp header). 5665 * Note that the directly accessible bytes (data_end - data) 5666 * could be less than this "len". Those bytes could be 5667 * indirectly read by a helper "bpf_skb_load_bytes()". 5668 */ 5669 __u32 len; 5670 /* 5671 * Eth protocol in the mac header (network byte order). e.g. 5672 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 5673 */ 5674 __u32 eth_protocol; 5675 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 5676 __u32 bind_inany; /* Is sock bound to an INANY address? */ 5677 __u32 hash; /* A hash of the packet 4 tuples */ 5678 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 5679 * new incoming connection request (e.g. selecting a listen sk for 5680 * the received SYN in the TCP case). reuse->sk is one of the sk 5681 * in the reuseport group. The bpf prog can use reuse->sk to learn 5682 * the local listening ip/port without looking into the skb. 5683 * 5684 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 5685 * reuse->migrating_sk is the socket that needs to be migrated 5686 * to another listening socket. migrating_sk could be a fullsock 5687 * sk that is fully established or a reqsk that is in-the-middle 5688 * of 3-way handshake. 5689 */ 5690 __bpf_md_ptr(struct bpf_sock *, sk); 5691 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 5692}; 5693 5694#define BPF_TAG_SIZE 8 5695 5696struct bpf_prog_info { 5697 __u32 type; 5698 __u32 id; 5699 __u8 tag[BPF_TAG_SIZE]; 5700 __u32 jited_prog_len; 5701 __u32 xlated_prog_len; 5702 __aligned_u64 jited_prog_insns; 5703 __aligned_u64 xlated_prog_insns; 5704 __u64 load_time; /* ns since boottime */ 5705 __u32 created_by_uid; 5706 __u32 nr_map_ids; 5707 __aligned_u64 map_ids; 5708 char name[BPF_OBJ_NAME_LEN]; 5709 __u32 ifindex; 5710 __u32 gpl_compatible:1; 5711 __u32 :31; /* alignment pad */ 5712 __u64 netns_dev; 5713 __u64 netns_ino; 5714 __u32 nr_jited_ksyms; 5715 __u32 nr_jited_func_lens; 5716 __aligned_u64 jited_ksyms; 5717 __aligned_u64 jited_func_lens; 5718 __u32 btf_id; 5719 __u32 func_info_rec_size; 5720 __aligned_u64 func_info; 5721 __u32 nr_func_info; 5722 __u32 nr_line_info; 5723 __aligned_u64 line_info; 5724 __aligned_u64 jited_line_info; 5725 __u32 nr_jited_line_info; 5726 __u32 line_info_rec_size; 5727 __u32 jited_line_info_rec_size; 5728 __u32 nr_prog_tags; 5729 __aligned_u64 prog_tags; 5730 __u64 run_time_ns; 5731 __u64 run_cnt; 5732 __u64 recursion_misses; 5733 __u32 verified_insns; 5734} __attribute__((aligned(8))); 5735 5736struct bpf_map_info { 5737 __u32 type; 5738 __u32 id; 5739 __u32 key_size; 5740 __u32 value_size; 5741 __u32 max_entries; 5742 __u32 map_flags; 5743 char name[BPF_OBJ_NAME_LEN]; 5744 __u32 ifindex; 5745 __u32 btf_vmlinux_value_type_id; 5746 __u64 netns_dev; 5747 __u64 netns_ino; 5748 __u32 btf_id; 5749 __u32 btf_key_type_id; 5750 __u32 btf_value_type_id; 5751 __u32 :32; /* alignment pad */ 5752 __u64 map_extra; 5753} __attribute__((aligned(8))); 5754 5755struct bpf_btf_info { 5756 __aligned_u64 btf; 5757 __u32 btf_size; 5758 __u32 id; 5759 __aligned_u64 name; 5760 __u32 name_len; 5761 __u32 kernel_btf; 5762} __attribute__((aligned(8))); 5763 5764struct bpf_link_info { 5765 __u32 type; 5766 __u32 id; 5767 __u32 prog_id; 5768 union { 5769 struct { 5770 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5771 __u32 tp_name_len; /* in/out: tp_name buffer len */ 5772 } raw_tracepoint; 5773 struct { 5774 __u32 attach_type; 5775 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 5776 __u32 target_btf_id; /* BTF type id inside the object */ 5777 } tracing; 5778 struct { 5779 __u64 cgroup_id; 5780 __u32 attach_type; 5781 } cgroup; 5782 struct { 5783 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 5784 __u32 target_name_len; /* in/out: target_name buffer len */ 5785 union { 5786 struct { 5787 __u32 map_id; 5788 } map; 5789 }; 5790 } iter; 5791 struct { 5792 __u32 netns_ino; 5793 __u32 attach_type; 5794 } netns; 5795 struct { 5796 __u32 ifindex; 5797 } xdp; 5798 }; 5799} __attribute__((aligned(8))); 5800 5801/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 5802 * by user and intended to be used by socket (e.g. to bind to, depends on 5803 * attach type). 5804 */ 5805struct bpf_sock_addr { 5806 __u32 user_family; /* Allows 4-byte read, but no write. */ 5807 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5808 * Stored in network byte order. 5809 */ 5810 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5811 * Stored in network byte order. 5812 */ 5813 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 5814 * Stored in network byte order 5815 */ 5816 __u32 family; /* Allows 4-byte read, but no write */ 5817 __u32 type; /* Allows 4-byte read, but no write */ 5818 __u32 protocol; /* Allows 4-byte read, but no write */ 5819 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5820 * Stored in network byte order. 5821 */ 5822 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5823 * Stored in network byte order. 5824 */ 5825 __bpf_md_ptr(struct bpf_sock *, sk); 5826}; 5827 5828/* User bpf_sock_ops struct to access socket values and specify request ops 5829 * and their replies. 5830 * Some of this fields are in network (bigendian) byte order and may need 5831 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 5832 * New fields can only be added at the end of this structure 5833 */ 5834struct bpf_sock_ops { 5835 __u32 op; 5836 union { 5837 __u32 args[4]; /* Optionally passed to bpf program */ 5838 __u32 reply; /* Returned by bpf program */ 5839 __u32 replylong[4]; /* Optionally returned by bpf prog */ 5840 }; 5841 __u32 family; 5842 __u32 remote_ip4; /* Stored in network byte order */ 5843 __u32 local_ip4; /* Stored in network byte order */ 5844 __u32 remote_ip6[4]; /* Stored in network byte order */ 5845 __u32 local_ip6[4]; /* Stored in network byte order */ 5846 __u32 remote_port; /* Stored in network byte order */ 5847 __u32 local_port; /* stored in host byte order */ 5848 __u32 is_fullsock; /* Some TCP fields are only valid if 5849 * there is a full socket. If not, the 5850 * fields read as zero. 5851 */ 5852 __u32 snd_cwnd; 5853 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5854 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 5855 __u32 state; 5856 __u32 rtt_min; 5857 __u32 snd_ssthresh; 5858 __u32 rcv_nxt; 5859 __u32 snd_nxt; 5860 __u32 snd_una; 5861 __u32 mss_cache; 5862 __u32 ecn_flags; 5863 __u32 rate_delivered; 5864 __u32 rate_interval_us; 5865 __u32 packets_out; 5866 __u32 retrans_out; 5867 __u32 total_retrans; 5868 __u32 segs_in; 5869 __u32 data_segs_in; 5870 __u32 segs_out; 5871 __u32 data_segs_out; 5872 __u32 lost_out; 5873 __u32 sacked_out; 5874 __u32 sk_txhash; 5875 __u64 bytes_received; 5876 __u64 bytes_acked; 5877 __bpf_md_ptr(struct bpf_sock *, sk); 5878 /* [skb_data, skb_data_end) covers the whole TCP header. 5879 * 5880 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 5881 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 5882 * header has not been written. 5883 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 5884 * been written so far. 5885 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 5886 * the 3WHS. 5887 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 5888 * the 3WHS. 5889 * 5890 * bpf_load_hdr_opt() can also be used to read a particular option. 5891 */ 5892 __bpf_md_ptr(void *, skb_data); 5893 __bpf_md_ptr(void *, skb_data_end); 5894 __u32 skb_len; /* The total length of a packet. 5895 * It includes the header, options, 5896 * and payload. 5897 */ 5898 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 5899 * an easy way to check for tcp_flags 5900 * without parsing skb_data. 5901 * 5902 * In particular, the skb_tcp_flags 5903 * will still be available in 5904 * BPF_SOCK_OPS_HDR_OPT_LEN even though 5905 * the outgoing header has not 5906 * been written yet. 5907 */ 5908}; 5909 5910/* Definitions for bpf_sock_ops_cb_flags */ 5911enum { 5912 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 5913 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 5914 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 5915 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 5916 /* Call bpf for all received TCP headers. The bpf prog will be 5917 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5918 * 5919 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5920 * for the header option related helpers that will be useful 5921 * to the bpf programs. 5922 * 5923 * It could be used at the client/active side (i.e. connect() side) 5924 * when the server told it that the server was in syncookie 5925 * mode and required the active side to resend the bpf-written 5926 * options. The active side can keep writing the bpf-options until 5927 * it received a valid packet from the server side to confirm 5928 * the earlier packet (and options) has been received. The later 5929 * example patch is using it like this at the active side when the 5930 * server is in syncookie mode. 5931 * 5932 * The bpf prog will usually turn this off in the common cases. 5933 */ 5934 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 5935 /* Call bpf when kernel has received a header option that 5936 * the kernel cannot handle. The bpf prog will be called under 5937 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 5938 * 5939 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5940 * for the header option related helpers that will be useful 5941 * to the bpf programs. 5942 */ 5943 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 5944 /* Call bpf when the kernel is writing header options for the 5945 * outgoing packet. The bpf prog will first be called 5946 * to reserve space in a skb under 5947 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 5948 * the bpf prog will be called to write the header option(s) 5949 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5950 * 5951 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 5952 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 5953 * related helpers that will be useful to the bpf programs. 5954 * 5955 * The kernel gets its chance to reserve space and write 5956 * options first before the BPF program does. 5957 */ 5958 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 5959/* Mask of all currently supported cb flags */ 5960 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 5961}; 5962 5963/* List of known BPF sock_ops operators. 5964 * New entries can only be added at the end 5965 */ 5966enum { 5967 BPF_SOCK_OPS_VOID, 5968 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 5969 * -1 if default value should be used 5970 */ 5971 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 5972 * window (in packets) or -1 if default 5973 * value should be used 5974 */ 5975 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 5976 * active connection is initialized 5977 */ 5978 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 5979 * active connection is 5980 * established 5981 */ 5982 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 5983 * passive connection is 5984 * established 5985 */ 5986 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 5987 * needs ECN 5988 */ 5989 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 5990 * based on the path and may be 5991 * dependent on the congestion control 5992 * algorithm. In general it indicates 5993 * a congestion threshold. RTTs above 5994 * this indicate congestion 5995 */ 5996 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 5997 * Arg1: value of icsk_retransmits 5998 * Arg2: value of icsk_rto 5999 * Arg3: whether RTO has expired 6000 */ 6001 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6002 * Arg1: sequence number of 1st byte 6003 * Arg2: # segments 6004 * Arg3: return value of 6005 * tcp_transmit_skb (0 => success) 6006 */ 6007 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6008 * Arg1: old_state 6009 * Arg2: new_state 6010 */ 6011 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6012 * socket transition to LISTEN state. 6013 */ 6014 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6015 */ 6016 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6017 * It will be called to handle 6018 * the packets received at 6019 * an already established 6020 * connection. 6021 * 6022 * sock_ops->skb_data: 6023 * Referring to the received skb. 6024 * It covers the TCP header only. 6025 * 6026 * bpf_load_hdr_opt() can also 6027 * be used to search for a 6028 * particular option. 6029 */ 6030 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6031 * header option later in 6032 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6033 * Arg1: bool want_cookie. (in 6034 * writing SYNACK only) 6035 * 6036 * sock_ops->skb_data: 6037 * Not available because no header has 6038 * been written yet. 6039 * 6040 * sock_ops->skb_tcp_flags: 6041 * The tcp_flags of the 6042 * outgoing skb. (e.g. SYN, ACK, FIN). 6043 * 6044 * bpf_reserve_hdr_opt() should 6045 * be used to reserve space. 6046 */ 6047 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6048 * Arg1: bool want_cookie. (in 6049 * writing SYNACK only) 6050 * 6051 * sock_ops->skb_data: 6052 * Referring to the outgoing skb. 6053 * It covers the TCP header 6054 * that has already been written 6055 * by the kernel and the 6056 * earlier bpf-progs. 6057 * 6058 * sock_ops->skb_tcp_flags: 6059 * The tcp_flags of the outgoing 6060 * skb. (e.g. SYN, ACK, FIN). 6061 * 6062 * bpf_store_hdr_opt() should 6063 * be used to write the 6064 * option. 6065 * 6066 * bpf_load_hdr_opt() can also 6067 * be used to search for a 6068 * particular option that 6069 * has already been written 6070 * by the kernel or the 6071 * earlier bpf-progs. 6072 */ 6073}; 6074 6075/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6076 * changes between the TCP and BPF versions. Ideally this should never happen. 6077 * If it does, we need to add code to convert them before calling 6078 * the BPF sock_ops function. 6079 */ 6080enum { 6081 BPF_TCP_ESTABLISHED = 1, 6082 BPF_TCP_SYN_SENT, 6083 BPF_TCP_SYN_RECV, 6084 BPF_TCP_FIN_WAIT1, 6085 BPF_TCP_FIN_WAIT2, 6086 BPF_TCP_TIME_WAIT, 6087 BPF_TCP_CLOSE, 6088 BPF_TCP_CLOSE_WAIT, 6089 BPF_TCP_LAST_ACK, 6090 BPF_TCP_LISTEN, 6091 BPF_TCP_CLOSING, /* Now a valid state */ 6092 BPF_TCP_NEW_SYN_RECV, 6093 6094 BPF_TCP_MAX_STATES /* Leave at the end! */ 6095}; 6096 6097enum { 6098 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6099 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6100 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6101 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6102 /* Copy the SYN pkt to optval 6103 * 6104 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6105 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6106 * to only getting from the saved_syn. It can either get the 6107 * syn packet from: 6108 * 6109 * 1. the just-received SYN packet (only available when writing the 6110 * SYNACK). It will be useful when it is not necessary to 6111 * save the SYN packet for latter use. It is also the only way 6112 * to get the SYN during syncookie mode because the syn 6113 * packet cannot be saved during syncookie. 6114 * 6115 * OR 6116 * 6117 * 2. the earlier saved syn which was done by 6118 * bpf_setsockopt(TCP_SAVE_SYN). 6119 * 6120 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 6121 * SYN packet is obtained. 6122 * 6123 * If the bpf-prog does not need the IP[46] header, the 6124 * bpf-prog can avoid parsing the IP header by using 6125 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 6126 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 6127 * 6128 * >0: Total number of bytes copied 6129 * -ENOSPC: Not enough space in optval. Only optlen number of 6130 * bytes is copied. 6131 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 6132 * is not saved by setsockopt(TCP_SAVE_SYN). 6133 */ 6134 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 6135 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6136 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 6137}; 6138 6139enum { 6140 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 6141}; 6142 6143/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 6144 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6145 */ 6146enum { 6147 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 6148 * total option spaces 6149 * required for an established 6150 * sk in order to calculate the 6151 * MSS. No skb is actually 6152 * sent. 6153 */ 6154 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 6155 * when sending a SYN. 6156 */ 6157}; 6158 6159struct bpf_perf_event_value { 6160 __u64 counter; 6161 __u64 enabled; 6162 __u64 running; 6163}; 6164 6165enum { 6166 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 6167 BPF_DEVCG_ACC_READ = (1ULL << 1), 6168 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 6169}; 6170 6171enum { 6172 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 6173 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 6174}; 6175 6176struct bpf_cgroup_dev_ctx { 6177 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 6178 __u32 access_type; 6179 __u32 major; 6180 __u32 minor; 6181}; 6182 6183struct bpf_raw_tracepoint_args { 6184 __u64 args[0]; 6185}; 6186 6187/* DIRECT: Skip the FIB rules and go to FIB table associated with device 6188 * OUTPUT: Do lookup from egress perspective; default is ingress 6189 */ 6190enum { 6191 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 6192 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 6193}; 6194 6195enum { 6196 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 6197 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 6198 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 6199 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 6200 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 6201 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 6202 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 6203 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 6204 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6205}; 6206 6207struct bpf_fib_lookup { 6208 /* input: network family for lookup (AF_INET, AF_INET6) 6209 * output: network family of egress nexthop 6210 */ 6211 __u8 family; 6212 6213 /* set if lookup is to consider L4 data - e.g., FIB rules */ 6214 __u8 l4_protocol; 6215 __be16 sport; 6216 __be16 dport; 6217 6218 union { /* used for MTU check */ 6219 /* input to lookup */ 6220 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 6221 6222 /* output: MTU value */ 6223 __u16 mtu_result; 6224 }; 6225 /* input: L3 device index for lookup 6226 * output: device index from FIB lookup 6227 */ 6228 __u32 ifindex; 6229 6230 union { 6231 /* inputs to lookup */ 6232 __u8 tos; /* AF_INET */ 6233 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 6234 6235 /* output: metric of fib result (IPv4/IPv6 only) */ 6236 __u32 rt_metric; 6237 }; 6238 6239 union { 6240 __be32 ipv4_src; 6241 __u32 ipv6_src[4]; /* in6_addr; network order */ 6242 }; 6243 6244 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6245 * network header. output: bpf_fib_lookup sets to gateway address 6246 * if FIB lookup returns gateway route 6247 */ 6248 union { 6249 __be32 ipv4_dst; 6250 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6251 }; 6252 6253 /* output */ 6254 __be16 h_vlan_proto; 6255 __be16 h_vlan_TCI; 6256 __u8 smac[6]; /* ETH_ALEN */ 6257 __u8 dmac[6]; /* ETH_ALEN */ 6258}; 6259 6260struct bpf_redir_neigh { 6261 /* network family for lookup (AF_INET, AF_INET6) */ 6262 __u32 nh_family; 6263 /* network address of nexthop; skips fib lookup to find gateway */ 6264 union { 6265 __be32 ipv4_nh; 6266 __u32 ipv6_nh[4]; /* in6_addr; network order */ 6267 }; 6268}; 6269 6270/* bpf_check_mtu flags*/ 6271enum bpf_check_mtu_flags { 6272 BPF_MTU_CHK_SEGS = (1U << 0), 6273}; 6274 6275enum bpf_check_mtu_ret { 6276 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 6277 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6278 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 6279}; 6280 6281enum bpf_task_fd_type { 6282 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 6283 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 6284 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 6285 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 6286 BPF_FD_TYPE_UPROBE, /* filename + offset */ 6287 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 6288}; 6289 6290enum { 6291 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 6292 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 6293 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 6294}; 6295 6296struct bpf_flow_keys { 6297 __u16 nhoff; 6298 __u16 thoff; 6299 __u16 addr_proto; /* ETH_P_* of valid addrs */ 6300 __u8 is_frag; 6301 __u8 is_first_frag; 6302 __u8 is_encap; 6303 __u8 ip_proto; 6304 __be16 n_proto; 6305 __be16 sport; 6306 __be16 dport; 6307 union { 6308 struct { 6309 __be32 ipv4_src; 6310 __be32 ipv4_dst; 6311 }; 6312 struct { 6313 __u32 ipv6_src[4]; /* in6_addr; network order */ 6314 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6315 }; 6316 }; 6317 __u32 flags; 6318 __be32 flow_label; 6319}; 6320 6321struct bpf_func_info { 6322 __u32 insn_off; 6323 __u32 type_id; 6324}; 6325 6326#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6327#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6328 6329struct bpf_line_info { 6330 __u32 insn_off; 6331 __u32 file_name_off; 6332 __u32 line_off; 6333 __u32 line_col; 6334}; 6335 6336struct bpf_spin_lock { 6337 __u32 val; 6338}; 6339 6340struct bpf_timer { 6341 __u64 :64; 6342 __u64 :64; 6343} __attribute__((aligned(8))); 6344 6345struct bpf_sysctl { 6346 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 6347 * Allows 1,2,4-byte read, but no write. 6348 */ 6349 __u32 file_pos; /* Sysctl file position to read from, write to. 6350 * Allows 1,2,4-byte read an 4-byte write. 6351 */ 6352}; 6353 6354struct bpf_sockopt { 6355 __bpf_md_ptr(struct bpf_sock *, sk); 6356 __bpf_md_ptr(void *, optval); 6357 __bpf_md_ptr(void *, optval_end); 6358 6359 __s32 level; 6360 __s32 optname; 6361 __s32 optlen; 6362 __s32 retval; 6363}; 6364 6365struct bpf_pidns_info { 6366 __u32 pid; 6367 __u32 tgid; 6368}; 6369 6370/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6371struct bpf_sk_lookup { 6372 union { 6373 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 6374 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 6375 }; 6376 6377 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6378 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6379 __u32 remote_ip4; /* Network byte order */ 6380 __u32 remote_ip6[4]; /* Network byte order */ 6381 __u32 remote_port; /* Network byte order */ 6382 __u32 local_ip4; /* Network byte order */ 6383 __u32 local_ip6[4]; /* Network byte order */ 6384 __u32 local_port; /* Host byte order */ 6385 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 6386}; 6387 6388/* 6389 * struct btf_ptr is used for typed pointer representation; the 6390 * type id is used to render the pointer data as the appropriate type 6391 * via the bpf_snprintf_btf() helper described above. A flags field - 6392 * potentially to specify additional details about the BTF pointer 6393 * (rather than its mode of display) - is included for future use. 6394 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6395 */ 6396struct btf_ptr { 6397 void *ptr; 6398 __u32 type_id; 6399 __u32 flags; /* BTF ptr flags; unused at present. */ 6400}; 6401 6402/* 6403 * Flags to control bpf_snprintf_btf() behaviour. 6404 * - BTF_F_COMPACT: no formatting around type information 6405 * - BTF_F_NONAME: no struct/union member names/types 6406 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6407 * equivalent to %px. 6408 * - BTF_F_ZERO: show zero-valued struct/union members; they 6409 * are not displayed by default 6410 */ 6411enum { 6412 BTF_F_COMPACT = (1ULL << 0), 6413 BTF_F_NONAME = (1ULL << 1), 6414 BTF_F_PTR_RAW = (1ULL << 2), 6415 BTF_F_ZERO = (1ULL << 3), 6416}; 6417 6418/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 6419 * has to be adjusted by relocations. It is emitted by llvm and passed to 6420 * libbpf and later to the kernel. 6421 */ 6422enum bpf_core_relo_kind { 6423 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 6424 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 6425 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 6426 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 6427 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 6428 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 6429 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 6430 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 6431 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 6432 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 6433 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 6434 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 6435}; 6436 6437/* 6438 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 6439 * and from libbpf to the kernel. 6440 * 6441 * CO-RE relocation captures the following data: 6442 * - insn_off - instruction offset (in bytes) within a BPF program that needs 6443 * its insn->imm field to be relocated with actual field info; 6444 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 6445 * type or field; 6446 * - access_str_off - offset into corresponding .BTF string section. String 6447 * interpretation depends on specific relocation kind: 6448 * - for field-based relocations, string encodes an accessed field using 6449 * a sequence of field and array indices, separated by colon (:). It's 6450 * conceptually very close to LLVM's getelementptr ([0]) instruction's 6451 * arguments for identifying offset to a field. 6452 * - for type-based relocations, strings is expected to be just "0"; 6453 * - for enum value-based relocations, string contains an index of enum 6454 * value within its enum type; 6455 * - kind - one of enum bpf_core_relo_kind; 6456 * 6457 * Example: 6458 * struct sample { 6459 * int a; 6460 * struct { 6461 * int b[10]; 6462 * }; 6463 * }; 6464 * 6465 * struct sample *s = ...; 6466 * int *x = &s->a; // encoded as "0:0" (a is field #0) 6467 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 6468 * // b is field #0 inside anon struct, accessing elem #5) 6469 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 6470 * 6471 * type_id for all relocs in this example will capture BTF type id of 6472 * `struct sample`. 6473 * 6474 * Such relocation is emitted when using __builtin_preserve_access_index() 6475 * Clang built-in, passing expression that captures field address, e.g.: 6476 * 6477 * bpf_probe_read(&dst, sizeof(dst), 6478 * __builtin_preserve_access_index(&src->a.b.c)); 6479 * 6480 * In this case Clang will emit field relocation recording necessary data to 6481 * be able to find offset of embedded `a.b.c` field within `src` struct. 6482 * 6483 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 6484 */ 6485struct bpf_core_relo { 6486 __u32 insn_off; 6487 __u32 type_id; 6488 __u32 access_str_off; 6489 enum bpf_core_relo_kind kind; 6490}; 6491 6492#endif /* _UAPI__LINUX_BPF_H__ */