iproute2/include/uapi/linux/bpf.h
<<
>>
Prefs
   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 __LINUX_BPF_H__
   9#define __LINUX_BPF_H__
  10
  11#include <linux/types.h>
  12#include <linux/bpf_common.h>
  13
  14/* Extended instruction set based on top of classic BPF */
  15
  16/* instruction classes */
  17#define BPF_JMP32       0x06    /* jmp mode in word width */
  18#define BPF_ALU64       0x07    /* alu mode in double word width */
  19
  20/* ld/ldx fields */
  21#define BPF_DW          0x18    /* double word (64-bit) */
  22#define BPF_MEMSX       0x80    /* load with sign extension */
  23#define BPF_ATOMIC      0xc0    /* atomic memory ops - op type in immediate */
  24#define BPF_XADD        0xc0    /* exclusive add - legacy name */
  25
  26/* alu/jmp fields */
  27#define BPF_MOV         0xb0    /* mov reg to reg */
  28#define BPF_ARSH        0xc0    /* sign extending arithmetic shift right */
  29
  30/* change endianness of a register */
  31#define BPF_END         0xd0    /* flags for endianness conversion: */
  32#define BPF_TO_LE       0x00    /* convert to little-endian */
  33#define BPF_TO_BE       0x08    /* convert to big-endian */
  34#define BPF_FROM_LE     BPF_TO_LE
  35#define BPF_FROM_BE     BPF_TO_BE
  36
  37/* jmp encodings */
  38#define BPF_JNE         0x50    /* jump != */
  39#define BPF_JLT         0xa0    /* LT is unsigned, '<' */
  40#define BPF_JLE         0xb0    /* LE is unsigned, '<=' */
  41#define BPF_JSGT        0x60    /* SGT is signed '>', GT in x86 */
  42#define BPF_JSGE        0x70    /* SGE is signed '>=', GE in x86 */
  43#define BPF_JSLT        0xc0    /* SLT is signed, '<' */
  44#define BPF_JSLE        0xd0    /* SLE is signed, '<=' */
  45#define BPF_JCOND       0xe0    /* conditional pseudo jumps: may_goto, goto_or_nop */
  46#define BPF_CALL        0x80    /* function call */
  47#define BPF_EXIT        0x90    /* function return */
  48
  49/* atomic op type fields (stored in immediate) */
  50#define BPF_FETCH       0x01    /* not an opcode on its own, used to build others */
  51#define BPF_XCHG        (0xe0 | BPF_FETCH)      /* atomic exchange */
  52#define BPF_CMPXCHG     (0xf0 | BPF_FETCH)      /* atomic compare-and-write */
  53
  54#define BPF_LOAD_ACQ    0x100   /* load-acquire */
  55#define BPF_STORE_REL   0x110   /* store-release */
  56
  57enum bpf_cond_pseudo_jmp {
  58        BPF_MAY_GOTO = 0,
  59};
  60
  61/* Register numbers */
  62enum {
  63        BPF_REG_0 = 0,
  64        BPF_REG_1,
  65        BPF_REG_2,
  66        BPF_REG_3,
  67        BPF_REG_4,
  68        BPF_REG_5,
  69        BPF_REG_6,
  70        BPF_REG_7,
  71        BPF_REG_8,
  72        BPF_REG_9,
  73        BPF_REG_10,
  74        __MAX_BPF_REG,
  75};
  76
  77/* BPF has 10 general purpose 64-bit registers and stack frame. */
  78#define MAX_BPF_REG     __MAX_BPF_REG
  79
  80struct bpf_insn {
  81        __u8    code;           /* opcode */
  82        __u8    dst_reg:4;      /* dest register */
  83        __u8    src_reg:4;      /* source register */
  84        __s16   off;            /* signed offset */
  85        __s32   imm;            /* signed immediate constant */
  86};
  87
  88/* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for
  89 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for
  90 * the trailing flexible array member) instead.
  91 */
  92struct bpf_lpm_trie_key {
  93        __u32   prefixlen;      /* up to 32 for AF_INET, 128 for AF_INET6 */
  94        __u8    data[0];        /* Arbitrary size */
  95};
  96
  97/* Header for bpf_lpm_trie_key structs */
  98struct bpf_lpm_trie_key_hdr {
  99        __u32   prefixlen;
 100};
 101
 102/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */
 103struct bpf_lpm_trie_key_u8 {
 104        union {
 105                struct bpf_lpm_trie_key_hdr     hdr;
 106                __u32                           prefixlen;
 107        };
 108        __u8    data[];         /* Arbitrary size */
 109};
 110
 111struct bpf_cgroup_storage_key {
 112        __u64   cgroup_inode_id;        /* cgroup inode id */
 113        __u32   attach_type;            /* program attach type (enum bpf_attach_type) */
 114};
 115
 116enum bpf_cgroup_iter_order {
 117        BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
 118        BPF_CGROUP_ITER_SELF_ONLY,              /* process only a single object. */
 119        BPF_CGROUP_ITER_DESCENDANTS_PRE,        /* walk descendants in pre-order. */
 120        BPF_CGROUP_ITER_DESCENDANTS_POST,       /* walk descendants in post-order. */
 121        BPF_CGROUP_ITER_ANCESTORS_UP,           /* walk ancestors upward. */
 122};
 123
 124union bpf_iter_link_info {
 125        struct {
 126                __u32   map_fd;
 127        } map;
 128        struct {
 129                enum bpf_cgroup_iter_order order;
 130
 131                /* At most one of cgroup_fd and cgroup_id can be non-zero. If
 132                 * both are zero, the walk starts from the default cgroup v2
 133                 * root. For walking v1 hierarchy, one should always explicitly
 134                 * specify cgroup_fd.
 135                 */
 136                __u32   cgroup_fd;
 137                __u64   cgroup_id;
 138        } cgroup;
 139        /* Parameters of task iterators. */
 140        struct {
 141                __u32   tid;
 142                __u32   pid;
 143                __u32   pid_fd;
 144        } task;
 145};
 146
 147/* BPF syscall commands, see bpf(2) man-page for more details. */
 148/**
 149 * DOC: eBPF Syscall Preamble
 150 *
 151 * The operation to be performed by the **bpf**\ () system call is determined
 152 * by the *cmd* argument. Each operation takes an accompanying argument,
 153 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
 154 * below). The size argument is the size of the union pointed to by *attr*.
 155 */
 156/**
 157 * DOC: eBPF Syscall Commands
 158 *
 159 * BPF_MAP_CREATE
 160 *      Description
 161 *              Create a map and return a file descriptor that refers to the
 162 *              map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
 163 *              is automatically enabled for the new file descriptor.
 164 *
 165 *              Applying **close**\ (2) to the file descriptor returned by
 166 *              **BPF_MAP_CREATE** will delete the map (but see NOTES).
 167 *
 168 *      Return
 169 *              A new file descriptor (a nonnegative integer), or -1 if an
 170 *              error occurred (in which case, *errno* is set appropriately).
 171 *
 172 * BPF_MAP_LOOKUP_ELEM
 173 *      Description
 174 *              Look up an element with a given *key* in the map referred to
 175 *              by the file descriptor *map_fd*.
 176 *
 177 *              The *flags* argument may be specified as one of the
 178 *              following:
 179 *
 180 *              **BPF_F_LOCK**
 181 *                      Look up the value of a spin-locked map without
 182 *                      returning the lock. This must be specified if the
 183 *                      elements contain a spinlock.
 184 *
 185 *      Return
 186 *              Returns zero on success. On error, -1 is returned and *errno*
 187 *              is set appropriately.
 188 *
 189 * BPF_MAP_UPDATE_ELEM
 190 *      Description
 191 *              Create or update an element (key/value pair) in a specified map.
 192 *
 193 *              The *flags* argument should be specified as one of the
 194 *              following:
 195 *
 196 *              **BPF_ANY**
 197 *                      Create a new element or update an existing element.
 198 *              **BPF_NOEXIST**
 199 *                      Create a new element only if it did not exist.
 200 *              **BPF_EXIST**
 201 *                      Update an existing element.
 202 *              **BPF_F_LOCK**
 203 *                      Update a spin_lock-ed map element.
 204 *
 205 *      Return
 206 *              Returns zero on success. On error, -1 is returned and *errno*
 207 *              is set appropriately.
 208 *
 209 *              May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
 210 *              **E2BIG**, **EEXIST**, or **ENOENT**.
 211 *
 212 *              **E2BIG**
 213 *                      The number of elements in the map reached the
 214 *                      *max_entries* limit specified at map creation time.
 215 *              **EEXIST**
 216 *                      If *flags* specifies **BPF_NOEXIST** and the element
 217 *                      with *key* already exists in the map.
 218 *              **ENOENT**
 219 *                      If *flags* specifies **BPF_EXIST** and the element with
 220 *                      *key* does not exist in the map.
 221 *
 222 * BPF_MAP_DELETE_ELEM
 223 *      Description
 224 *              Look up and delete an element by key in a specified map.
 225 *
 226 *      Return
 227 *              Returns zero on success. On error, -1 is returned and *errno*
 228 *              is set appropriately.
 229 *
 230 * BPF_MAP_GET_NEXT_KEY
 231 *      Description
 232 *              Look up an element by key in a specified map and return the key
 233 *              of the next element. Can be used to iterate over all elements
 234 *              in the map.
 235 *
 236 *      Return
 237 *              Returns zero on success. On error, -1 is returned and *errno*
 238 *              is set appropriately.
 239 *
 240 *              The following cases can be used to iterate over all elements of
 241 *              the map:
 242 *
 243 *              * If *key* is not found, the operation returns zero and sets
 244 *                the *next_key* pointer to the key of the first element.
 245 *              * If *key* is found, the operation returns zero and sets the
 246 *                *next_key* pointer to the key of the next element.
 247 *              * If *key* is the last element, returns -1 and *errno* is set
 248 *                to **ENOENT**.
 249 *
 250 *              May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
 251 *              **EINVAL** on error.
 252 *
 253 * BPF_PROG_LOAD
 254 *      Description
 255 *              Verify and load an eBPF program, returning a new file
 256 *              descriptor associated with the program.
 257 *
 258 *              Applying **close**\ (2) to the file descriptor returned by
 259 *              **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
 260 *
 261 *              The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
 262 *              automatically enabled for the new file descriptor.
 263 *
 264 *      Return
 265 *              A new file descriptor (a nonnegative integer), or -1 if an
 266 *              error occurred (in which case, *errno* is set appropriately).
 267 *
 268 * BPF_OBJ_PIN
 269 *      Description
 270 *              Pin an eBPF program or map referred by the specified *bpf_fd*
 271 *              to the provided *pathname* on the filesystem.
 272 *
 273 *              The *pathname* argument must not contain a dot (".").
 274 *
 275 *              On success, *pathname* retains a reference to the eBPF object,
 276 *              preventing deallocation of the object when the original
 277 *              *bpf_fd* is closed. This allow the eBPF object to live beyond
 278 *              **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
 279 *              process.
 280 *
 281 *              Applying **unlink**\ (2) or similar calls to the *pathname*
 282 *              unpins the object from the filesystem, removing the reference.
 283 *              If no other file descriptors or filesystem nodes refer to the
 284 *              same object, it will be deallocated (see NOTES).
 285 *
 286 *              The filesystem type for the parent directory of *pathname* must
 287 *              be **BPF_FS_MAGIC**.
 288 *
 289 *      Return
 290 *              Returns zero on success. On error, -1 is returned and *errno*
 291 *              is set appropriately.
 292 *
 293 * BPF_OBJ_GET
 294 *      Description
 295 *              Open a file descriptor for the eBPF object pinned to the
 296 *              specified *pathname*.
 297 *
 298 *      Return
 299 *              A new file descriptor (a nonnegative integer), or -1 if an
 300 *              error occurred (in which case, *errno* is set appropriately).
 301 *
 302 * BPF_PROG_ATTACH
 303 *      Description
 304 *              Attach an eBPF program to a *target_fd* at the specified
 305 *              *attach_type* hook.
 306 *
 307 *              The *attach_type* specifies the eBPF attachment point to
 308 *              attach the program to, and must be one of *bpf_attach_type*
 309 *              (see below).
 310 *
 311 *              The *attach_bpf_fd* must be a valid file descriptor for a
 312 *              loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
 313 *              or sock_ops type corresponding to the specified *attach_type*.
 314 *
 315 *              The *target_fd* must be a valid file descriptor for a kernel
 316 *              object which depends on the attach type of *attach_bpf_fd*:
 317 *
 318 *              **BPF_PROG_TYPE_CGROUP_DEVICE**,
 319 *              **BPF_PROG_TYPE_CGROUP_SKB**,
 320 *              **BPF_PROG_TYPE_CGROUP_SOCK**,
 321 *              **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
 322 *              **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
 323 *              **BPF_PROG_TYPE_CGROUP_SYSCTL**,
 324 *              **BPF_PROG_TYPE_SOCK_OPS**
 325 *
 326 *                      Control Group v2 hierarchy with the eBPF controller
 327 *                      enabled. Requires the kernel to be compiled with
 328 *                      **CONFIG_CGROUP_BPF**.
 329 *
 330 *              **BPF_PROG_TYPE_FLOW_DISSECTOR**
 331 *
 332 *                      Network namespace (eg /proc/self/ns/net).
 333 *
 334 *              **BPF_PROG_TYPE_LIRC_MODE2**
 335 *
 336 *                      LIRC device path (eg /dev/lircN). Requires the kernel
 337 *                      to be compiled with **CONFIG_BPF_LIRC_MODE2**.
 338 *
 339 *              **BPF_PROG_TYPE_SK_SKB**,
 340 *              **BPF_PROG_TYPE_SK_MSG**
 341 *
 342 *                      eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
 343 *
 344 *      Return
 345 *              Returns zero on success. On error, -1 is returned and *errno*
 346 *              is set appropriately.
 347 *
 348 * BPF_PROG_DETACH
 349 *      Description
 350 *              Detach the eBPF program associated with the *target_fd* at the
 351 *              hook specified by *attach_type*. The program must have been
 352 *              previously attached using **BPF_PROG_ATTACH**.
 353 *
 354 *      Return
 355 *              Returns zero on success. On error, -1 is returned and *errno*
 356 *              is set appropriately.
 357 *
 358 * BPF_PROG_TEST_RUN
 359 *      Description
 360 *              Run the eBPF program associated with the *prog_fd* a *repeat*
 361 *              number of times against a provided program context *ctx_in* and
 362 *              data *data_in*, and return the modified program context
 363 *              *ctx_out*, *data_out* (for example, packet data), result of the
 364 *              execution *retval*, and *duration* of the test run.
 365 *
 366 *              The sizes of the buffers provided as input and output
 367 *              parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
 368 *              be provided in the corresponding variables *ctx_size_in*,
 369 *              *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
 370 *              of these parameters are not provided (ie set to NULL), the
 371 *              corresponding size field must be zero.
 372 *
 373 *              Some program types have particular requirements:
 374 *
 375 *              **BPF_PROG_TYPE_SK_LOOKUP**
 376 *                      *data_in* and *data_out* must be NULL.
 377 *
 378 *              **BPF_PROG_TYPE_RAW_TRACEPOINT**,
 379 *              **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
 380 *
 381 *                      *ctx_out*, *data_in* and *data_out* must be NULL.
 382 *                      *repeat* must be zero.
 383 *
 384 *              BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
 385 *
 386 *      Return
 387 *              Returns zero on success. On error, -1 is returned and *errno*
 388 *              is set appropriately.
 389 *
 390 *              **ENOSPC**
 391 *                      Either *data_size_out* or *ctx_size_out* is too small.
 392 *              **ENOTSUPP**
 393 *                      This command is not supported by the program type of
 394 *                      the program referred to by *prog_fd*.
 395 *
 396 * BPF_PROG_GET_NEXT_ID
 397 *      Description
 398 *              Fetch the next eBPF program currently loaded into the kernel.
 399 *
 400 *              Looks for the eBPF program with an id greater than *start_id*
 401 *              and updates *next_id* on success. If no other eBPF programs
 402 *              remain with ids higher than *start_id*, returns -1 and sets
 403 *              *errno* to **ENOENT**.
 404 *
 405 *      Return
 406 *              Returns zero on success. On error, or when no id remains, -1
 407 *              is returned and *errno* is set appropriately.
 408 *
 409 * BPF_MAP_GET_NEXT_ID
 410 *      Description
 411 *              Fetch the next eBPF map currently loaded into the kernel.
 412 *
 413 *              Looks for the eBPF map with an id greater than *start_id*
 414 *              and updates *next_id* on success. If no other eBPF maps
 415 *              remain with ids higher than *start_id*, returns -1 and sets
 416 *              *errno* to **ENOENT**.
 417 *
 418 *      Return
 419 *              Returns zero on success. On error, or when no id remains, -1
 420 *              is returned and *errno* is set appropriately.
 421 *
 422 * BPF_PROG_GET_FD_BY_ID
 423 *      Description
 424 *              Open a file descriptor for the eBPF program corresponding to
 425 *              *prog_id*.
 426 *
 427 *      Return
 428 *              A new file descriptor (a nonnegative integer), or -1 if an
 429 *              error occurred (in which case, *errno* is set appropriately).
 430 *
 431 * BPF_MAP_GET_FD_BY_ID
 432 *      Description
 433 *              Open a file descriptor for the eBPF map corresponding to
 434 *              *map_id*.
 435 *
 436 *      Return
 437 *              A new file descriptor (a nonnegative integer), or -1 if an
 438 *              error occurred (in which case, *errno* is set appropriately).
 439 *
 440 * BPF_OBJ_GET_INFO_BY_FD
 441 *      Description
 442 *              Obtain information about the eBPF object corresponding to
 443 *              *bpf_fd*.
 444 *
 445 *              Populates up to *info_len* bytes of *info*, which will be in
 446 *              one of the following formats depending on the eBPF object type
 447 *              of *bpf_fd*:
 448 *
 449 *              * **struct bpf_prog_info**
 450 *              * **struct bpf_map_info**
 451 *              * **struct bpf_btf_info**
 452 *              * **struct bpf_link_info**
 453 *              * **struct bpf_token_info**
 454 *
 455 *      Return
 456 *              Returns zero on success. On error, -1 is returned and *errno*
 457 *              is set appropriately.
 458 *
 459 * BPF_PROG_QUERY
 460 *      Description
 461 *              Obtain information about eBPF programs associated with the
 462 *              specified *attach_type* hook.
 463 *
 464 *              The *target_fd* must be a valid file descriptor for a kernel
 465 *              object which depends on the attach type of *attach_bpf_fd*:
 466 *
 467 *              **BPF_PROG_TYPE_CGROUP_DEVICE**,
 468 *              **BPF_PROG_TYPE_CGROUP_SKB**,
 469 *              **BPF_PROG_TYPE_CGROUP_SOCK**,
 470 *              **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
 471 *              **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
 472 *              **BPF_PROG_TYPE_CGROUP_SYSCTL**,
 473 *              **BPF_PROG_TYPE_SOCK_OPS**
 474 *
 475 *                      Control Group v2 hierarchy with the eBPF controller
 476 *                      enabled. Requires the kernel to be compiled with
 477 *                      **CONFIG_CGROUP_BPF**.
 478 *
 479 *              **BPF_PROG_TYPE_FLOW_DISSECTOR**
 480 *
 481 *                      Network namespace (eg /proc/self/ns/net).
 482 *
 483 *              **BPF_PROG_TYPE_LIRC_MODE2**
 484 *
 485 *                      LIRC device path (eg /dev/lircN). Requires the kernel
 486 *                      to be compiled with **CONFIG_BPF_LIRC_MODE2**.
 487 *
 488 *              **BPF_PROG_QUERY** always fetches the number of programs
 489 *              attached and the *attach_flags* which were used to attach those
 490 *              programs. Additionally, if *prog_ids* is nonzero and the number
 491 *              of attached programs is less than *prog_cnt*, populates
 492 *              *prog_ids* with the eBPF program ids of the programs attached
 493 *              at *target_fd*.
 494 *
 495 *              The following flags may alter the result:
 496 *
 497 *              **BPF_F_QUERY_EFFECTIVE**
 498 *                      Only return information regarding programs which are
 499 *                      currently effective at the specified *target_fd*.
 500 *
 501 *      Return
 502 *              Returns zero on success. On error, -1 is returned and *errno*
 503 *              is set appropriately.
 504 *
 505 * BPF_RAW_TRACEPOINT_OPEN
 506 *      Description
 507 *              Attach an eBPF program to a tracepoint *name* to access kernel
 508 *              internal arguments of the tracepoint in their raw form.
 509 *
 510 *              The *prog_fd* must be a valid file descriptor associated with
 511 *              a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
 512 *
 513 *              No ABI guarantees are made about the content of tracepoint
 514 *              arguments exposed to the corresponding eBPF program.
 515 *
 516 *              Applying **close**\ (2) to the file descriptor returned by
 517 *              **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
 518 *
 519 *      Return
 520 *              A new file descriptor (a nonnegative integer), or -1 if an
 521 *              error occurred (in which case, *errno* is set appropriately).
 522 *
 523 * BPF_BTF_LOAD
 524 *      Description
 525 *              Verify and load BPF Type Format (BTF) metadata into the kernel,
 526 *              returning a new file descriptor associated with the metadata.
 527 *              BTF is described in more detail at
 528 *              https://www.kernel.org/doc/html/latest/bpf/btf.html.
 529 *
 530 *              The *btf* parameter must point to valid memory providing
 531 *              *btf_size* bytes of BTF binary metadata.
 532 *
 533 *              The returned file descriptor can be passed to other **bpf**\ ()
 534 *              subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
 535 *              associate the BTF with those objects.
 536 *
 537 *              Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
 538 *              parameters to specify a *btf_log_buf*, *btf_log_size* and
 539 *              *btf_log_level* which allow the kernel to return freeform log
 540 *              output regarding the BTF verification process.
 541 *
 542 *      Return
 543 *              A new file descriptor (a nonnegative integer), or -1 if an
 544 *              error occurred (in which case, *errno* is set appropriately).
 545 *
 546 * BPF_BTF_GET_FD_BY_ID
 547 *      Description
 548 *              Open a file descriptor for the BPF Type Format (BTF)
 549 *              corresponding to *btf_id*.
 550 *
 551 *      Return
 552 *              A new file descriptor (a nonnegative integer), or -1 if an
 553 *              error occurred (in which case, *errno* is set appropriately).
 554 *
 555 * BPF_TASK_FD_QUERY
 556 *      Description
 557 *              Obtain information about eBPF programs associated with the
 558 *              target process identified by *pid* and *fd*.
 559 *
 560 *              If the *pid* and *fd* are associated with a tracepoint, kprobe
 561 *              or uprobe perf event, then the *prog_id* and *fd_type* will
 562 *              be populated with the eBPF program id and file descriptor type
 563 *              of type **bpf_task_fd_type**. If associated with a kprobe or
 564 *              uprobe, the  *probe_offset* and *probe_addr* will also be
 565 *              populated. Optionally, if *buf* is provided, then up to
 566 *              *buf_len* bytes of *buf* will be populated with the name of
 567 *              the tracepoint, kprobe or uprobe.
 568 *
 569 *              The resulting *prog_id* may be introspected in deeper detail
 570 *              using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
 571 *
 572 *      Return
 573 *              Returns zero on success. On error, -1 is returned and *errno*
 574 *              is set appropriately.
 575 *
 576 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
 577 *      Description
 578 *              Look up an element with the given *key* in the map referred to
 579 *              by the file descriptor *fd*, and if found, delete the element.
 580 *
 581 *              For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
 582 *              types, the *flags* argument needs to be set to 0, but for other
 583 *              map types, it may be specified as:
 584 *
 585 *              **BPF_F_LOCK**
 586 *                      Look up and delete the value of a spin-locked map
 587 *                      without returning the lock. This must be specified if
 588 *                      the elements contain a spinlock.
 589 *
 590 *              The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
 591 *              implement this command as a "pop" operation, deleting the top
 592 *              element rather than one corresponding to *key*.
 593 *              The *key* and *key_len* parameters should be zeroed when
 594 *              issuing this operation for these map types.
 595 *
 596 *              This command is only valid for the following map types:
 597 *              * **BPF_MAP_TYPE_QUEUE**
 598 *              * **BPF_MAP_TYPE_STACK**
 599 *              * **BPF_MAP_TYPE_HASH**
 600 *              * **BPF_MAP_TYPE_PERCPU_HASH**
 601 *              * **BPF_MAP_TYPE_LRU_HASH**
 602 *              * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
 603 *
 604 *      Return
 605 *              Returns zero on success. On error, -1 is returned and *errno*
 606 *              is set appropriately.
 607 *
 608 * BPF_MAP_FREEZE
 609 *      Description
 610 *              Freeze the permissions of the specified map.
 611 *
 612 *              Write permissions may be frozen by passing zero *flags*.
 613 *              Upon success, no future syscall invocations may alter the
 614 *              map state of *map_fd*. Write operations from eBPF programs
 615 *              are still possible for a frozen map.
 616 *
 617 *              Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
 618 *
 619 *      Return
 620 *              Returns zero on success. On error, -1 is returned and *errno*
 621 *              is set appropriately.
 622 *
 623 * BPF_BTF_GET_NEXT_ID
 624 *      Description
 625 *              Fetch the next BPF Type Format (BTF) object currently loaded
 626 *              into the kernel.
 627 *
 628 *              Looks for the BTF object with an id greater than *start_id*
 629 *              and updates *next_id* on success. If no other BTF objects
 630 *              remain with ids higher than *start_id*, returns -1 and sets
 631 *              *errno* to **ENOENT**.
 632 *
 633 *      Return
 634 *              Returns zero on success. On error, or when no id remains, -1
 635 *              is returned and *errno* is set appropriately.
 636 *
 637 * BPF_MAP_LOOKUP_BATCH
 638 *      Description
 639 *              Iterate and fetch multiple elements in a map.
 640 *
 641 *              Two opaque values are used to manage batch operations,
 642 *              *in_batch* and *out_batch*. Initially, *in_batch* must be set
 643 *              to NULL to begin the batched operation. After each subsequent
 644 *              **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
 645 *              *out_batch* as the *in_batch* for the next operation to
 646 *              continue iteration from the current point. Both *in_batch* and
 647 *              *out_batch* must point to memory large enough to hold a key,
 648 *              except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH,
 649 *              LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters
 650 *              must be at least 4 bytes wide regardless of key size.
 651 *
 652 *              The *keys* and *values* are output parameters which must point
 653 *              to memory large enough to hold *count* items based on the key
 654 *              and value size of the map *map_fd*. The *keys* buffer must be
 655 *              of *key_size* * *count*. The *values* buffer must be of
 656 *              *value_size* * *count*.
 657 *
 658 *              The *elem_flags* argument may be specified as one of the
 659 *              following:
 660 *
 661 *              **BPF_F_LOCK**
 662 *                      Look up the value of a spin-locked map without
 663 *                      returning the lock. This must be specified if the
 664 *                      elements contain a spinlock.
 665 *
 666 *              On success, *count* elements from the map are copied into the
 667 *              user buffer, with the keys copied into *keys* and the values
 668 *              copied into the corresponding indices in *values*.
 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 **ENOSPC** to indicate that *keys* or
 678 *              *values* is too small to dump an entire bucket during
 679 *              iteration of a hash-based map type.
 680 *
 681 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
 682 *      Description
 683 *              Iterate and delete all elements in a map.
 684 *
 685 *              This operation has the same behavior as
 686 *              **BPF_MAP_LOOKUP_BATCH** with two exceptions:
 687 *
 688 *              * Every element that is successfully returned is also deleted
 689 *                from the map. This is at least *count* elements. Note that
 690 *                *count* is both an input and an output parameter.
 691 *              * Upon returning with *errno* set to **EFAULT**, up to
 692 *                *count* elements may be deleted without returning the keys
 693 *                and values of the deleted elements.
 694 *
 695 *      Return
 696 *              Returns zero on success. On error, -1 is returned and *errno*
 697 *              is set appropriately.
 698 *
 699 * BPF_MAP_UPDATE_BATCH
 700 *      Description
 701 *              Update multiple elements in a map by *key*.
 702 *
 703 *              The *keys* and *values* are input parameters which must point
 704 *              to memory large enough to hold *count* items based on the key
 705 *              and value size of the map *map_fd*. The *keys* buffer must be
 706 *              of *key_size* * *count*. The *values* buffer must be of
 707 *              *value_size* * *count*.
 708 *
 709 *              Each element specified in *keys* is sequentially updated to the
 710 *              value in the corresponding index in *values*. The *in_batch*
 711 *              and *out_batch* parameters are ignored and should be zeroed.
 712 *
 713 *              The *elem_flags* argument should be specified as one of the
 714 *              following:
 715 *
 716 *              **BPF_ANY**
 717 *                      Create new elements or update a existing elements.
 718 *              **BPF_NOEXIST**
 719 *                      Create new elements only if they do not exist.
 720 *              **BPF_EXIST**
 721 *                      Update existing elements.
 722 *              **BPF_F_LOCK**
 723 *                      Update spin_lock-ed map elements. This must be
 724 *                      specified if the map value contains a spinlock.
 725 *
 726 *              On success, *count* elements from the map are updated.
 727 *
 728 *              If an error is returned and *errno* is not **EFAULT**, *count*
 729 *              is set to the number of successfully processed elements.
 730 *
 731 *      Return
 732 *              Returns zero on success. On error, -1 is returned and *errno*
 733 *              is set appropriately.
 734 *
 735 *              May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
 736 *              **E2BIG**. **E2BIG** indicates that the number of elements in
 737 *              the map reached the *max_entries* limit specified at map
 738 *              creation time.
 739 *
 740 *              May set *errno* to one of the following error codes under
 741 *              specific circumstances:
 742 *
 743 *              **EEXIST**
 744 *                      If *flags* specifies **BPF_NOEXIST** and the element
 745 *                      with *key* already exists in the map.
 746 *              **ENOENT**
 747 *                      If *flags* specifies **BPF_EXIST** and the element with
 748 *                      *key* does not exist in the map.
 749 *
 750 * BPF_MAP_DELETE_BATCH
 751 *      Description
 752 *              Delete multiple elements in a map by *key*.
 753 *
 754 *              The *keys* parameter is an input parameter which must point
 755 *              to memory large enough to hold *count* items based on the key
 756 *              size of the map *map_fd*, that is, *key_size* * *count*.
 757 *
 758 *              Each element specified in *keys* is sequentially deleted. The
 759 *              *in_batch*, *out_batch*, and *values* parameters are ignored
 760 *              and should be zeroed.
 761 *
 762 *              The *elem_flags* argument may be specified as one of the
 763 *              following:
 764 *
 765 *              **BPF_F_LOCK**
 766 *                      Look up the value of a spin-locked map without
 767 *                      returning the lock. This must be specified if the
 768 *                      elements contain a spinlock.
 769 *
 770 *              On success, *count* elements from the map are updated.
 771 *
 772 *              If an error is returned and *errno* is not **EFAULT**, *count*
 773 *              is set to the number of successfully processed elements. If
 774 *              *errno* is **EFAULT**, up to *count* elements may be been
 775 *              deleted.
 776 *
 777 *      Return
 778 *              Returns zero on success. On error, -1 is returned and *errno*
 779 *              is set appropriately.
 780 *
 781 * BPF_LINK_CREATE
 782 *      Description
 783 *              Attach an eBPF program to a *target_fd* at the specified
 784 *              *attach_type* hook and return a file descriptor handle for
 785 *              managing the link.
 786 *
 787 *      Return
 788 *              A new file descriptor (a nonnegative integer), or -1 if an
 789 *              error occurred (in which case, *errno* is set appropriately).
 790 *
 791 * BPF_LINK_UPDATE
 792 *      Description
 793 *              Update the eBPF program in the specified *link_fd* to
 794 *              *new_prog_fd*.
 795 *
 796 *      Return
 797 *              Returns zero on success. On error, -1 is returned and *errno*
 798 *              is set appropriately.
 799 *
 800 * BPF_LINK_GET_FD_BY_ID
 801 *      Description
 802 *              Open a file descriptor for the eBPF Link corresponding to
 803 *              *link_id*.
 804 *
 805 *      Return
 806 *              A new file descriptor (a nonnegative integer), or -1 if an
 807 *              error occurred (in which case, *errno* is set appropriately).
 808 *
 809 * BPF_LINK_GET_NEXT_ID
 810 *      Description
 811 *              Fetch the next eBPF link currently loaded into the kernel.
 812 *
 813 *              Looks for the eBPF link with an id greater than *start_id*
 814 *              and updates *next_id* on success. If no other eBPF links
 815 *              remain with ids higher than *start_id*, returns -1 and sets
 816 *              *errno* to **ENOENT**.
 817 *
 818 *      Return
 819 *              Returns zero on success. On error, or when no id remains, -1
 820 *              is returned and *errno* is set appropriately.
 821 *
 822 * BPF_ENABLE_STATS
 823 *      Description
 824 *              Enable eBPF runtime statistics gathering.
 825 *
 826 *              Runtime statistics gathering for the eBPF runtime is disabled
 827 *              by default to minimize the corresponding performance overhead.
 828 *              This command enables statistics globally.
 829 *
 830 *              Multiple programs may independently enable statistics.
 831 *              After gathering the desired statistics, eBPF runtime statistics
 832 *              may be disabled again by calling **close**\ (2) for the file
 833 *              descriptor returned by this function. Statistics will only be
 834 *              disabled system-wide when all outstanding file descriptors
 835 *              returned by prior calls for this subcommand are closed.
 836 *
 837 *      Return
 838 *              A new file descriptor (a nonnegative integer), or -1 if an
 839 *              error occurred (in which case, *errno* is set appropriately).
 840 *
 841 * BPF_ITER_CREATE
 842 *      Description
 843 *              Create an iterator on top of the specified *link_fd* (as
 844 *              previously created using **BPF_LINK_CREATE**) and return a
 845 *              file descriptor that can be used to trigger the iteration.
 846 *
 847 *              If the resulting file descriptor is pinned to the filesystem
 848 *              using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
 849 *              for that path will trigger the iterator to read kernel state
 850 *              using the eBPF program attached to *link_fd*.
 851 *
 852 *      Return
 853 *              A new file descriptor (a nonnegative integer), or -1 if an
 854 *              error occurred (in which case, *errno* is set appropriately).
 855 *
 856 * BPF_LINK_DETACH
 857 *      Description
 858 *              Forcefully detach the specified *link_fd* from its
 859 *              corresponding attachment point.
 860 *
 861 *      Return
 862 *              Returns zero on success. On error, -1 is returned and *errno*
 863 *              is set appropriately.
 864 *
 865 * BPF_PROG_BIND_MAP
 866 *      Description
 867 *              Bind a map to the lifetime of an eBPF program.
 868 *
 869 *              The map identified by *map_fd* is bound to the program
 870 *              identified by *prog_fd* and only released when *prog_fd* is
 871 *              released. This may be used in cases where metadata should be
 872 *              associated with a program which otherwise does not contain any
 873 *              references to the map (for example, embedded in the eBPF
 874 *              program instructions).
 875 *
 876 *      Return
 877 *              Returns zero on success. On error, -1 is returned and *errno*
 878 *              is set appropriately.
 879 *
 880 * BPF_TOKEN_CREATE
 881 *      Description
 882 *              Create BPF token with embedded information about what
 883 *              BPF-related functionality it allows:
 884 *              - a set of allowed bpf() syscall commands;
 885 *              - a set of allowed BPF map types to be created with
 886 *              BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed;
 887 *              - a set of allowed BPF program types and BPF program attach
 888 *              types to be loaded with BPF_PROG_LOAD command, if
 889 *              BPF_PROG_LOAD itself is allowed.
 890 *
 891 *              BPF token is created (derived) from an instance of BPF FS,
 892 *              assuming it has necessary delegation mount options specified.
 893 *              This BPF token can be passed as an extra parameter to various
 894 *              bpf() syscall commands to grant BPF subsystem functionality to
 895 *              unprivileged processes.
 896 *
 897 *              When created, BPF token is "associated" with the owning
 898 *              user namespace of BPF FS instance (super block) that it was
 899 *              derived from, and subsequent BPF operations performed with
 900 *              BPF token would be performing capabilities checks (i.e.,
 901 *              CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within
 902 *              that user namespace. Without BPF token, such capabilities
 903 *              have to be granted in init user namespace, making bpf()
 904 *              syscall incompatible with user namespace, for the most part.
 905 *
 906 *      Return
 907 *              A new file descriptor (a nonnegative integer), or -1 if an
 908 *              error occurred (in which case, *errno* is set appropriately).
 909 *
 910 * BPF_PROG_STREAM_READ_BY_FD
 911 *      Description
 912 *              Read data of a program's BPF stream. The program is identified
 913 *              by *prog_fd*, and the stream is identified by the *stream_id*.
 914 *              The data is copied to a buffer pointed to by *stream_buf*, and
 915 *              filled less than or equal to *stream_buf_len* bytes.
 916 *
 917 *      Return
 918 *              Number of bytes read from the stream on success, or -1 if an
 919 *              error occurred (in which case, *errno* is set appropriately).
 920 *
 921 * NOTES
 922 *      eBPF objects (maps and programs) can be shared between processes.
 923 *
 924 *      * After **fork**\ (2), the child inherits file descriptors
 925 *        referring to the same eBPF objects.
 926 *      * File descriptors referring to eBPF objects can be transferred over
 927 *        **unix**\ (7) domain sockets.
 928 *      * File descriptors referring to eBPF objects can be duplicated in the
 929 *        usual way, using **dup**\ (2) and similar calls.
 930 *      * File descriptors referring to eBPF objects can be pinned to the
 931 *        filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
 932 *
 933 *      An eBPF object is deallocated only after all file descriptors referring
 934 *      to the object have been closed and no references remain pinned to the
 935 *      filesystem or attached (for example, bound to a program or device).
 936 */
 937enum bpf_cmd {
 938        BPF_MAP_CREATE,
 939        BPF_MAP_LOOKUP_ELEM,
 940        BPF_MAP_UPDATE_ELEM,
 941        BPF_MAP_DELETE_ELEM,
 942        BPF_MAP_GET_NEXT_KEY,
 943        BPF_PROG_LOAD,
 944        BPF_OBJ_PIN,
 945        BPF_OBJ_GET,
 946        BPF_PROG_ATTACH,
 947        BPF_PROG_DETACH,
 948        BPF_PROG_TEST_RUN,
 949        BPF_PROG_RUN = BPF_PROG_TEST_RUN,
 950        BPF_PROG_GET_NEXT_ID,
 951        BPF_MAP_GET_NEXT_ID,
 952        BPF_PROG_GET_FD_BY_ID,
 953        BPF_MAP_GET_FD_BY_ID,
 954        BPF_OBJ_GET_INFO_BY_FD,
 955        BPF_PROG_QUERY,
 956        BPF_RAW_TRACEPOINT_OPEN,
 957        BPF_BTF_LOAD,
 958        BPF_BTF_GET_FD_BY_ID,
 959        BPF_TASK_FD_QUERY,
 960        BPF_MAP_LOOKUP_AND_DELETE_ELEM,
 961        BPF_MAP_FREEZE,
 962        BPF_BTF_GET_NEXT_ID,
 963        BPF_MAP_LOOKUP_BATCH,
 964        BPF_MAP_LOOKUP_AND_DELETE_BATCH,
 965        BPF_MAP_UPDATE_BATCH,
 966        BPF_MAP_DELETE_BATCH,
 967        BPF_LINK_CREATE,
 968        BPF_LINK_UPDATE,
 969        BPF_LINK_GET_FD_BY_ID,
 970        BPF_LINK_GET_NEXT_ID,
 971        BPF_ENABLE_STATS,
 972        BPF_ITER_CREATE,
 973        BPF_LINK_DETACH,
 974        BPF_PROG_BIND_MAP,
 975        BPF_TOKEN_CREATE,
 976        BPF_PROG_STREAM_READ_BY_FD,
 977        __MAX_BPF_CMD,
 978};
 979
 980enum bpf_map_type {
 981        BPF_MAP_TYPE_UNSPEC,
 982        BPF_MAP_TYPE_HASH,
 983        BPF_MAP_TYPE_ARRAY,
 984        BPF_MAP_TYPE_PROG_ARRAY,
 985        BPF_MAP_TYPE_PERF_EVENT_ARRAY,
 986        BPF_MAP_TYPE_PERCPU_HASH,
 987        BPF_MAP_TYPE_PERCPU_ARRAY,
 988        BPF_MAP_TYPE_STACK_TRACE,
 989        BPF_MAP_TYPE_CGROUP_ARRAY,
 990        BPF_MAP_TYPE_LRU_HASH,
 991        BPF_MAP_TYPE_LRU_PERCPU_HASH,
 992        BPF_MAP_TYPE_LPM_TRIE,
 993        BPF_MAP_TYPE_ARRAY_OF_MAPS,
 994        BPF_MAP_TYPE_HASH_OF_MAPS,
 995        BPF_MAP_TYPE_DEVMAP,
 996        BPF_MAP_TYPE_SOCKMAP,
 997        BPF_MAP_TYPE_CPUMAP,
 998        BPF_MAP_TYPE_XSKMAP,
 999        BPF_MAP_TYPE_SOCKHASH,
1000        BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
1001        /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
1002         * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
1003         * both cgroup-attached and other progs and supports all functionality
1004         * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
1005         * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
1006         */
1007        BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
1008        BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
1009        BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
1010        /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs
1011         * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +
1012         * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1013         * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1014         * deprecated.
1015         */
1016        BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
1017        BPF_MAP_TYPE_QUEUE,
1018        BPF_MAP_TYPE_STACK,
1019        BPF_MAP_TYPE_SK_STORAGE,
1020        BPF_MAP_TYPE_DEVMAP_HASH,
1021        BPF_MAP_TYPE_STRUCT_OPS,
1022        BPF_MAP_TYPE_RINGBUF,
1023        BPF_MAP_TYPE_INODE_STORAGE,
1024        BPF_MAP_TYPE_TASK_STORAGE,
1025        BPF_MAP_TYPE_BLOOM_FILTER,
1026        BPF_MAP_TYPE_USER_RINGBUF,
1027        BPF_MAP_TYPE_CGRP_STORAGE,
1028        BPF_MAP_TYPE_ARENA,
1029        __MAX_BPF_MAP_TYPE
1030};
1031
1032/* Note that tracing related programs such as
1033 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
1034 * are not subject to a stable API since kernel internal data
1035 * structures can change from release to release and may
1036 * therefore break existing tracing BPF programs. Tracing BPF
1037 * programs correspond to /a/ specific kernel which is to be
1038 * analyzed, and not /a/ specific kernel /and/ all future ones.
1039 */
1040enum bpf_prog_type {
1041        BPF_PROG_TYPE_UNSPEC,
1042        BPF_PROG_TYPE_SOCKET_FILTER,
1043        BPF_PROG_TYPE_KPROBE,
1044        BPF_PROG_TYPE_SCHED_CLS,
1045        BPF_PROG_TYPE_SCHED_ACT,
1046        BPF_PROG_TYPE_TRACEPOINT,
1047        BPF_PROG_TYPE_XDP,
1048        BPF_PROG_TYPE_PERF_EVENT,
1049        BPF_PROG_TYPE_CGROUP_SKB,
1050        BPF_PROG_TYPE_CGROUP_SOCK,
1051        BPF_PROG_TYPE_LWT_IN,
1052        BPF_PROG_TYPE_LWT_OUT,
1053        BPF_PROG_TYPE_LWT_XMIT,
1054        BPF_PROG_TYPE_SOCK_OPS,
1055        BPF_PROG_TYPE_SK_SKB,
1056        BPF_PROG_TYPE_CGROUP_DEVICE,
1057        BPF_PROG_TYPE_SK_MSG,
1058        BPF_PROG_TYPE_RAW_TRACEPOINT,
1059        BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
1060        BPF_PROG_TYPE_LWT_SEG6LOCAL,
1061        BPF_PROG_TYPE_LIRC_MODE2,
1062        BPF_PROG_TYPE_SK_REUSEPORT,
1063        BPF_PROG_TYPE_FLOW_DISSECTOR,
1064        BPF_PROG_TYPE_CGROUP_SYSCTL,
1065        BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
1066        BPF_PROG_TYPE_CGROUP_SOCKOPT,
1067        BPF_PROG_TYPE_TRACING,
1068        BPF_PROG_TYPE_STRUCT_OPS,
1069        BPF_PROG_TYPE_EXT,
1070        BPF_PROG_TYPE_LSM,
1071        BPF_PROG_TYPE_SK_LOOKUP,
1072        BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
1073        BPF_PROG_TYPE_NETFILTER,
1074        __MAX_BPF_PROG_TYPE
1075};
1076
1077enum bpf_attach_type {
1078        BPF_CGROUP_INET_INGRESS,
1079        BPF_CGROUP_INET_EGRESS,
1080        BPF_CGROUP_INET_SOCK_CREATE,
1081        BPF_CGROUP_SOCK_OPS,
1082        BPF_SK_SKB_STREAM_PARSER,
1083        BPF_SK_SKB_STREAM_VERDICT,
1084        BPF_CGROUP_DEVICE,
1085        BPF_SK_MSG_VERDICT,
1086        BPF_CGROUP_INET4_BIND,
1087        BPF_CGROUP_INET6_BIND,
1088        BPF_CGROUP_INET4_CONNECT,
1089        BPF_CGROUP_INET6_CONNECT,
1090        BPF_CGROUP_INET4_POST_BIND,
1091        BPF_CGROUP_INET6_POST_BIND,
1092        BPF_CGROUP_UDP4_SENDMSG,
1093        BPF_CGROUP_UDP6_SENDMSG,
1094        BPF_LIRC_MODE2,
1095        BPF_FLOW_DISSECTOR,
1096        BPF_CGROUP_SYSCTL,
1097        BPF_CGROUP_UDP4_RECVMSG,
1098        BPF_CGROUP_UDP6_RECVMSG,
1099        BPF_CGROUP_GETSOCKOPT,
1100        BPF_CGROUP_SETSOCKOPT,
1101        BPF_TRACE_RAW_TP,
1102        BPF_TRACE_FENTRY,
1103        BPF_TRACE_FEXIT,
1104        BPF_MODIFY_RETURN,
1105        BPF_LSM_MAC,
1106        BPF_TRACE_ITER,
1107        BPF_CGROUP_INET4_GETPEERNAME,
1108        BPF_CGROUP_INET6_GETPEERNAME,
1109        BPF_CGROUP_INET4_GETSOCKNAME,
1110        BPF_CGROUP_INET6_GETSOCKNAME,
1111        BPF_XDP_DEVMAP,
1112        BPF_CGROUP_INET_SOCK_RELEASE,
1113        BPF_XDP_CPUMAP,
1114        BPF_SK_LOOKUP,
1115        BPF_XDP,
1116        BPF_SK_SKB_VERDICT,
1117        BPF_SK_REUSEPORT_SELECT,
1118        BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1119        BPF_PERF_EVENT,
1120        BPF_TRACE_KPROBE_MULTI,
1121        BPF_LSM_CGROUP,
1122        BPF_STRUCT_OPS,
1123        BPF_NETFILTER,
1124        BPF_TCX_INGRESS,
1125        BPF_TCX_EGRESS,
1126        BPF_TRACE_UPROBE_MULTI,
1127        BPF_CGROUP_UNIX_CONNECT,
1128        BPF_CGROUP_UNIX_SENDMSG,
1129        BPF_CGROUP_UNIX_RECVMSG,
1130        BPF_CGROUP_UNIX_GETPEERNAME,
1131        BPF_CGROUP_UNIX_GETSOCKNAME,
1132        BPF_NETKIT_PRIMARY,
1133        BPF_NETKIT_PEER,
1134        BPF_TRACE_KPROBE_SESSION,
1135        BPF_TRACE_UPROBE_SESSION,
1136        __MAX_BPF_ATTACH_TYPE
1137};
1138
1139#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1140
1141/* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[]
1142 * in sync with the definitions below.
1143 */
1144enum bpf_link_type {
1145        BPF_LINK_TYPE_UNSPEC = 0,
1146        BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1147        BPF_LINK_TYPE_TRACING = 2,
1148        BPF_LINK_TYPE_CGROUP = 3,
1149        BPF_LINK_TYPE_ITER = 4,
1150        BPF_LINK_TYPE_NETNS = 5,
1151        BPF_LINK_TYPE_XDP = 6,
1152        BPF_LINK_TYPE_PERF_EVENT = 7,
1153        BPF_LINK_TYPE_KPROBE_MULTI = 8,
1154        BPF_LINK_TYPE_STRUCT_OPS = 9,
1155        BPF_LINK_TYPE_NETFILTER = 10,
1156        BPF_LINK_TYPE_TCX = 11,
1157        BPF_LINK_TYPE_UPROBE_MULTI = 12,
1158        BPF_LINK_TYPE_NETKIT = 13,
1159        BPF_LINK_TYPE_SOCKMAP = 14,
1160        __MAX_BPF_LINK_TYPE,
1161};
1162
1163#define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE
1164
1165enum bpf_perf_event_type {
1166        BPF_PERF_EVENT_UNSPEC = 0,
1167        BPF_PERF_EVENT_UPROBE = 1,
1168        BPF_PERF_EVENT_URETPROBE = 2,
1169        BPF_PERF_EVENT_KPROBE = 3,
1170        BPF_PERF_EVENT_KRETPROBE = 4,
1171        BPF_PERF_EVENT_TRACEPOINT = 5,
1172        BPF_PERF_EVENT_EVENT = 6,
1173};
1174
1175/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1176 *
1177 * NONE(default): No further bpf programs allowed in the subtree.
1178 *
1179 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1180 * the program in this cgroup yields to sub-cgroup program.
1181 *
1182 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1183 * that cgroup program gets run in addition to the program in this cgroup.
1184 *
1185 * Only one program is allowed to be attached to a cgroup with
1186 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1187 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1188 * release old program and attach the new one. Attach flags has to match.
1189 *
1190 * Multiple programs are allowed to be attached to a cgroup with
1191 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1192 * (those that were attached first, run first)
1193 * The programs of sub-cgroup are executed first, then programs of
1194 * this cgroup and then programs of parent cgroup.
1195 * When children program makes decision (like picking TCP CA or sock bind)
1196 * parent program has a chance to override it.
1197 *
1198 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1199 * programs for a cgroup. Though it's possible to replace an old program at
1200 * any position by also specifying BPF_F_REPLACE flag and position itself in
1201 * replace_bpf_fd attribute. Old program at this position will be released.
1202 *
1203 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1204 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1205 * Ex1:
1206 * cgrp1 (MULTI progs A, B) ->
1207 *    cgrp2 (OVERRIDE prog C) ->
1208 *      cgrp3 (MULTI prog D) ->
1209 *        cgrp4 (OVERRIDE prog E) ->
1210 *          cgrp5 (NONE prog F)
1211 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1212 * if prog F is detached, the execution is E,D,A,B
1213 * if prog F and D are detached, the execution is E,A,B
1214 * if prog F, E and D are detached, the execution is C,A,B
1215 *
1216 * All eligible programs are executed regardless of return code from
1217 * earlier programs.
1218 */
1219#define BPF_F_ALLOW_OVERRIDE    (1U << 0)
1220#define BPF_F_ALLOW_MULTI       (1U << 1)
1221/* Generic attachment flags. */
1222#define BPF_F_REPLACE           (1U << 2)
1223#define BPF_F_BEFORE            (1U << 3)
1224#define BPF_F_AFTER             (1U << 4)
1225#define BPF_F_ID                (1U << 5)
1226#define BPF_F_PREORDER          (1U << 6)
1227#define BPF_F_LINK              BPF_F_LINK /* 1 << 13 */
1228
1229/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1230 * verifier will perform strict alignment checking as if the kernel
1231 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1232 * and NET_IP_ALIGN defined to 2.
1233 */
1234#define BPF_F_STRICT_ALIGNMENT  (1U << 0)
1235
1236/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
1237 * verifier will allow any alignment whatsoever.  On platforms
1238 * with strict alignment requirements for loads ands stores (such
1239 * as sparc and mips) the verifier validates that all loads and
1240 * stores provably follow this requirement.  This flag turns that
1241 * checking and enforcement off.
1242 *
1243 * It is mostly used for testing when we want to validate the
1244 * context and memory access aspects of the verifier, but because
1245 * of an unaligned access the alignment check would trigger before
1246 * the one we are interested in.
1247 */
1248#define BPF_F_ANY_ALIGNMENT     (1U << 1)
1249
1250/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1251 * Verifier does sub-register def/use analysis and identifies instructions whose
1252 * def only matters for low 32-bit, high 32-bit is never referenced later
1253 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1254 * that it is safe to ignore clearing high 32-bit for these instructions. This
1255 * saves some back-ends a lot of code-gen. However such optimization is not
1256 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1257 * hence hasn't used verifier's analysis result. But, we really want to have a
1258 * way to be able to verify the correctness of the described optimization on
1259 * x86_64 on which testsuites are frequently exercised.
1260 *
1261 * So, this flag is introduced. Once it is set, verifier will randomize high
1262 * 32-bit for those instructions who has been identified as safe to ignore them.
1263 * Then, if verifier is not doing correct analysis, such randomization will
1264 * regress tests to expose bugs.
1265 */
1266#define BPF_F_TEST_RND_HI32     (1U << 2)
1267
1268/* The verifier internal test flag. Behavior is undefined */
1269#define BPF_F_TEST_STATE_FREQ   (1U << 3)
1270
1271/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1272 * restrict map and helper usage for such programs. Sleepable BPF programs can
1273 * only be attached to hooks where kernel execution context allows sleeping.
1274 * Such programs are allowed to use helpers that may sleep like
1275 * bpf_copy_from_user().
1276 */
1277#define BPF_F_SLEEPABLE         (1U << 4)
1278
1279/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1280 * fully support xdp frags.
1281 */
1282#define BPF_F_XDP_HAS_FRAGS     (1U << 5)
1283
1284/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
1285 * program becomes device-bound but can access XDP metadata.
1286 */
1287#define BPF_F_XDP_DEV_BOUND_ONLY        (1U << 6)
1288
1289/* The verifier internal test flag. Behavior is undefined */
1290#define BPF_F_TEST_REG_INVARIANTS       (1U << 7)
1291
1292/* link_create.kprobe_multi.flags used in LINK_CREATE command for
1293 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1294 */
1295enum {
1296        BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
1297};
1298
1299/* link_create.uprobe_multi.flags used in LINK_CREATE command for
1300 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
1301 */
1302enum {
1303        BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
1304};
1305
1306/* link_create.netfilter.flags used in LINK_CREATE command for
1307 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
1308 */
1309#define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
1310
1311/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1312 * the following extensions:
1313 *
1314 * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]
1315 * insn[0].imm:      map fd or fd_idx
1316 * insn[1].imm:      0
1317 * insn[0].off:      0
1318 * insn[1].off:      0
1319 * ldimm64 rewrite:  address of map
1320 * verifier type:    CONST_PTR_TO_MAP
1321 */
1322#define BPF_PSEUDO_MAP_FD       1
1323#define BPF_PSEUDO_MAP_IDX      5
1324
1325/* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE
1326 * insn[0].imm:      map fd or fd_idx
1327 * insn[1].imm:      offset into value
1328 * insn[0].off:      0
1329 * insn[1].off:      0
1330 * ldimm64 rewrite:  address of map[0]+offset
1331 * verifier type:    PTR_TO_MAP_VALUE
1332 */
1333#define BPF_PSEUDO_MAP_VALUE            2
1334#define BPF_PSEUDO_MAP_IDX_VALUE        6
1335
1336/* insn[0].src_reg:  BPF_PSEUDO_BTF_ID
1337 * insn[0].imm:      kernel btd id of VAR
1338 * insn[1].imm:      0
1339 * insn[0].off:      0
1340 * insn[1].off:      0
1341 * ldimm64 rewrite:  address of the kernel variable
1342 * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1343 *                   is struct/union.
1344 */
1345#define BPF_PSEUDO_BTF_ID       3
1346/* insn[0].src_reg:  BPF_PSEUDO_FUNC
1347 * insn[0].imm:      insn offset to the func
1348 * insn[1].imm:      0
1349 * insn[0].off:      0
1350 * insn[1].off:      0
1351 * ldimm64 rewrite:  address of the function
1352 * verifier type:    PTR_TO_FUNC.
1353 */
1354#define BPF_PSEUDO_FUNC         4
1355
1356/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1357 * offset to another bpf function
1358 */
1359#define BPF_PSEUDO_CALL         1
1360/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1361 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1362 */
1363#define BPF_PSEUDO_KFUNC_CALL   2
1364
1365enum bpf_addr_space_cast {
1366        BPF_ADDR_SPACE_CAST = 1,
1367};
1368
1369/* flags for BPF_MAP_UPDATE_ELEM command */
1370enum {
1371        BPF_ANY         = 0, /* create new element or update existing */
1372        BPF_NOEXIST     = 1, /* create new element if it didn't exist */
1373        BPF_EXIST       = 2, /* update existing element */
1374        BPF_F_LOCK      = 4, /* spin_lock-ed map_lookup/map_update */
1375};
1376
1377/* flags for BPF_MAP_CREATE command */
1378enum {
1379        BPF_F_NO_PREALLOC       = (1U << 0),
1380/* Instead of having one common LRU list in the
1381 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1382 * which can scale and perform better.
1383 * Note, the LRU nodes (including free nodes) cannot be moved
1384 * across different LRU lists.
1385 */
1386        BPF_F_NO_COMMON_LRU     = (1U << 1),
1387/* Specify numa node during map creation */
1388        BPF_F_NUMA_NODE         = (1U << 2),
1389
1390/* Flags for accessing BPF object from syscall side. */
1391        BPF_F_RDONLY            = (1U << 3),
1392        BPF_F_WRONLY            = (1U << 4),
1393
1394/* Flag for stack_map, store build_id+offset instead of pointer */
1395        BPF_F_STACK_BUILD_ID    = (1U << 5),
1396
1397/* Zero-initialize hash function seed. This should only be used for testing. */
1398        BPF_F_ZERO_SEED         = (1U << 6),
1399
1400/* Flags for accessing BPF object from program side. */
1401        BPF_F_RDONLY_PROG       = (1U << 7),
1402        BPF_F_WRONLY_PROG       = (1U << 8),
1403
1404/* Clone map from listener for newly accepted socket */
1405        BPF_F_CLONE             = (1U << 9),
1406
1407/* Enable memory-mapping BPF map */
1408        BPF_F_MMAPABLE          = (1U << 10),
1409
1410/* Share perf_event among processes */
1411        BPF_F_PRESERVE_ELEMS    = (1U << 11),
1412
1413/* Create a map that is suitable to be an inner map with dynamic max entries */
1414        BPF_F_INNER_MAP         = (1U << 12),
1415
1416/* Create a map that will be registered/unregesitered by the backed bpf_link */
1417        BPF_F_LINK              = (1U << 13),
1418
1419/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1420        BPF_F_PATH_FD           = (1U << 14),
1421
1422/* Flag for value_type_btf_obj_fd, the fd is available */
1423        BPF_F_VTYPE_BTF_OBJ_FD  = (1U << 15),
1424
1425/* BPF token FD is passed in a corresponding command's token_fd field */
1426        BPF_F_TOKEN_FD          = (1U << 16),
1427
1428/* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */
1429        BPF_F_SEGV_ON_FAULT     = (1U << 17),
1430
1431/* Do not translate kernel bpf_arena pointers to user pointers */
1432        BPF_F_NO_USER_CONV      = (1U << 18),
1433};
1434
1435/* Flags for BPF_PROG_QUERY. */
1436
1437/* Query effective (directly attached + inherited from ancestor cgroups)
1438 * programs that will be executed for events within a cgroup.
1439 * attach_flags with this flag are always returned 0.
1440 */
1441#define BPF_F_QUERY_EFFECTIVE   (1U << 0)
1442
1443/* Flags for BPF_PROG_TEST_RUN */
1444
1445/* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1446#define BPF_F_TEST_RUN_ON_CPU   (1U << 0)
1447/* If set, XDP frames will be transmitted after processing */
1448#define BPF_F_TEST_XDP_LIVE_FRAMES      (1U << 1)
1449/* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */
1450#define BPF_F_TEST_SKB_CHECKSUM_COMPLETE        (1U << 2)
1451
1452/* type for BPF_ENABLE_STATS */
1453enum bpf_stats_type {
1454        /* enabled run_time_ns and run_cnt */
1455        BPF_STATS_RUN_TIME = 0,
1456};
1457
1458enum bpf_stack_build_id_status {
1459        /* user space need an empty entry to identify end of a trace */
1460        BPF_STACK_BUILD_ID_EMPTY = 0,
1461        /* with valid build_id and offset */
1462        BPF_STACK_BUILD_ID_VALID = 1,
1463        /* couldn't get build_id, fallback to ip */
1464        BPF_STACK_BUILD_ID_IP = 2,
1465};
1466
1467#define BPF_BUILD_ID_SIZE 20
1468struct bpf_stack_build_id {
1469        __s32           status;
1470        unsigned char   build_id[BPF_BUILD_ID_SIZE];
1471        union {
1472                __u64   offset;
1473                __u64   ip;
1474        };
1475};
1476
1477#define BPF_OBJ_NAME_LEN 16U
1478
1479enum {
1480        BPF_STREAM_STDOUT = 1,
1481        BPF_STREAM_STDERR = 2,
1482};
1483
1484union bpf_attr {
1485        struct { /* anonymous struct used by BPF_MAP_CREATE command */
1486                __u32   map_type;       /* one of enum bpf_map_type */
1487                __u32   key_size;       /* size of key in bytes */
1488                __u32   value_size;     /* size of value in bytes */
1489                __u32   max_entries;    /* max number of entries in a map */
1490                __u32   map_flags;      /* BPF_MAP_CREATE related
1491                                         * flags defined above.
1492                                         */
1493                __u32   inner_map_fd;   /* fd pointing to the inner map */
1494                __u32   numa_node;      /* numa node (effective only if
1495                                         * BPF_F_NUMA_NODE is set).
1496                                         */
1497                char    map_name[BPF_OBJ_NAME_LEN];
1498                __u32   map_ifindex;    /* ifindex of netdev to create on */
1499                __u32   btf_fd;         /* fd pointing to a BTF type data */
1500                __u32   btf_key_type_id;        /* BTF type_id of the key */
1501                __u32   btf_value_type_id;      /* BTF type_id of the value */
1502                __u32   btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1503                                                   * struct stored as the
1504                                                   * map value
1505                                                   */
1506                /* Any per-map-type extra fields
1507                 *
1508                 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1509                 * number of hash functions (if 0, the bloom filter will default
1510                 * to using 5 hash functions).
1511                 *
1512                 * BPF_MAP_TYPE_ARENA - contains the address where user space
1513                 * is going to mmap() the arena. It has to be page aligned.
1514                 */
1515                __u64   map_extra;
1516
1517                __s32   value_type_btf_obj_fd;  /* fd pointing to a BTF
1518                                                 * type data for
1519                                                 * btf_vmlinux_value_type_id.
1520                                                 */
1521                /* BPF token FD to use with BPF_MAP_CREATE operation.
1522                 * If provided, map_flags should have BPF_F_TOKEN_FD flag set.
1523                 */
1524                __s32   map_token_fd;
1525        };
1526
1527        struct { /* anonymous struct used by BPF_MAP_*_ELEM and BPF_MAP_FREEZE commands */
1528                __u32           map_fd;
1529                __aligned_u64   key;
1530                union {
1531                        __aligned_u64 value;
1532                        __aligned_u64 next_key;
1533                };
1534                __u64           flags;
1535        };
1536
1537        struct { /* struct used by BPF_MAP_*_BATCH commands */
1538                __aligned_u64   in_batch;       /* start batch,
1539                                                 * NULL to start from beginning
1540                                                 */
1541                __aligned_u64   out_batch;      /* output: next start batch */
1542                __aligned_u64   keys;
1543                __aligned_u64   values;
1544                __u32           count;          /* input/output:
1545                                                 * input: # of key/value
1546                                                 * elements
1547                                                 * output: # of filled elements
1548                                                 */
1549                __u32           map_fd;
1550                __u64           elem_flags;
1551                __u64           flags;
1552        } batch;
1553
1554        struct { /* anonymous struct used by BPF_PROG_LOAD command */
1555                __u32           prog_type;      /* one of enum bpf_prog_type */
1556                __u32           insn_cnt;
1557                __aligned_u64   insns;
1558                __aligned_u64   license;
1559                __u32           log_level;      /* verbosity level of verifier */
1560                __u32           log_size;       /* size of user buffer */
1561                __aligned_u64   log_buf;        /* user supplied buffer */
1562                __u32           kern_version;   /* not used */
1563                __u32           prog_flags;
1564                char            prog_name[BPF_OBJ_NAME_LEN];
1565                __u32           prog_ifindex;   /* ifindex of netdev to prep for */
1566                /* For some prog types expected attach type must be known at
1567                 * load time to verify attach type specific parts of prog
1568                 * (context accesses, allowed helpers, etc).
1569                 */
1570                __u32           expected_attach_type;
1571                __u32           prog_btf_fd;    /* fd pointing to BTF type data */
1572                __u32           func_info_rec_size;     /* userspace bpf_func_info size */
1573                __aligned_u64   func_info;      /* func info */
1574                __u32           func_info_cnt;  /* number of bpf_func_info records */
1575                __u32           line_info_rec_size;     /* userspace bpf_line_info size */
1576                __aligned_u64   line_info;      /* line info */
1577                __u32           line_info_cnt;  /* number of bpf_line_info records */
1578                __u32           attach_btf_id;  /* in-kernel BTF type id to attach to */
1579                union {
1580                        /* valid prog_fd to attach to bpf prog */
1581                        __u32           attach_prog_fd;
1582                        /* or valid module BTF object fd or 0 to attach to vmlinux */
1583                        __u32           attach_btf_obj_fd;
1584                };
1585                __u32           core_relo_cnt;  /* number of bpf_core_relo */
1586                __aligned_u64   fd_array;       /* array of FDs */
1587                __aligned_u64   core_relos;
1588                __u32           core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1589                /* output: actual total log contents size (including termintaing zero).
1590                 * It could be both larger than original log_size (if log was
1591                 * truncated), or smaller (if log buffer wasn't filled completely).
1592                 */
1593                __u32           log_true_size;
1594                /* BPF token FD to use with BPF_PROG_LOAD operation.
1595                 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set.
1596                 */
1597                __s32           prog_token_fd;
1598                /* The fd_array_cnt can be used to pass the length of the
1599                 * fd_array array. In this case all the [map] file descriptors
1600                 * passed in this array will be bound to the program, even if
1601                 * the maps are not referenced directly. The functionality is
1602                 * similar to the BPF_PROG_BIND_MAP syscall, but maps can be
1603                 * used by the verifier during the program load. If provided,
1604                 * then the fd_array[0,...,fd_array_cnt-1] is expected to be
1605                 * continuous.
1606                 */
1607                __u32           fd_array_cnt;
1608        };
1609
1610        struct { /* anonymous struct used by BPF_OBJ_* commands */
1611                __aligned_u64   pathname;
1612                __u32           bpf_fd;
1613                __u32           file_flags;
1614                /* Same as dirfd in openat() syscall; see openat(2)
1615                 * manpage for details of path FD and pathname semantics;
1616                 * path_fd should accompanied by BPF_F_PATH_FD flag set in
1617                 * file_flags field, otherwise it should be set to zero;
1618                 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
1619                 */
1620                __s32           path_fd;
1621        };
1622
1623        struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1624                union {
1625                        __u32   target_fd;      /* target object to attach to or ... */
1626                        __u32   target_ifindex; /* target ifindex */
1627                };
1628                __u32           attach_bpf_fd;
1629                __u32           attach_type;
1630                __u32           attach_flags;
1631                __u32           replace_bpf_fd;
1632                union {
1633                        __u32   relative_fd;
1634                        __u32   relative_id;
1635                };
1636                __u64           expected_revision;
1637        };
1638
1639        struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1640                __u32           prog_fd;
1641                __u32           retval;
1642                __u32           data_size_in;   /* input: len of data_in */
1643                __u32           data_size_out;  /* input/output: len of data_out
1644                                                 *   returns ENOSPC if data_out
1645                                                 *   is too small.
1646                                                 */
1647                __aligned_u64   data_in;
1648                __aligned_u64   data_out;
1649                __u32           repeat;
1650                __u32           duration;
1651                __u32           ctx_size_in;    /* input: len of ctx_in */
1652                __u32           ctx_size_out;   /* input/output: len of ctx_out
1653                                                 *   returns ENOSPC if ctx_out
1654                                                 *   is too small.
1655                                                 */
1656                __aligned_u64   ctx_in;
1657                __aligned_u64   ctx_out;
1658                __u32           flags;
1659                __u32           cpu;
1660                __u32           batch_size;
1661        } test;
1662
1663        struct { /* anonymous struct used by BPF_*_GET_*_ID */
1664                union {
1665                        __u32           start_id;
1666                        __u32           prog_id;
1667                        __u32           map_id;
1668                        __u32           btf_id;
1669                        __u32           link_id;
1670                };
1671                __u32           next_id;
1672                __u32           open_flags;
1673                __s32           fd_by_id_token_fd;
1674        };
1675
1676        struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1677                __u32           bpf_fd;
1678                __u32           info_len;
1679                __aligned_u64   info;
1680        } info;
1681
1682        struct { /* anonymous struct used by BPF_PROG_QUERY command */
1683                union {
1684                        __u32   target_fd;      /* target object to query or ... */
1685                        __u32   target_ifindex; /* target ifindex */
1686                };
1687                __u32           attach_type;
1688                __u32           query_flags;
1689                __u32           attach_flags;
1690                __aligned_u64   prog_ids;
1691                union {
1692                        __u32   prog_cnt;
1693                        __u32   count;
1694                };
1695                __u32           :32;
1696                /* output: per-program attach_flags.
1697                 * not allowed to be set during effective query.
1698                 */
1699                __aligned_u64   prog_attach_flags;
1700                __aligned_u64   link_ids;
1701                __aligned_u64   link_attach_flags;
1702                __u64           revision;
1703        } query;
1704
1705        struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1706                __u64           name;
1707                __u32           prog_fd;
1708                __u32           :32;
1709                __aligned_u64   cookie;
1710        } raw_tracepoint;
1711
1712        struct { /* anonymous struct for BPF_BTF_LOAD */
1713                __aligned_u64   btf;
1714                __aligned_u64   btf_log_buf;
1715                __u32           btf_size;
1716                __u32           btf_log_size;
1717                __u32           btf_log_level;
1718                /* output: actual total log contents size (including termintaing zero).
1719                 * It could be both larger than original log_size (if log was
1720                 * truncated), or smaller (if log buffer wasn't filled completely).
1721                 */
1722                __u32           btf_log_true_size;
1723                __u32           btf_flags;
1724                /* BPF token FD to use with BPF_BTF_LOAD operation.
1725                 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set.
1726                 */
1727                __s32           btf_token_fd;
1728        };
1729
1730        struct {
1731                __u32           pid;            /* input: pid */
1732                __u32           fd;             /* input: fd */
1733                __u32           flags;          /* input: flags */
1734                __u32           buf_len;        /* input/output: buf len */
1735                __aligned_u64   buf;            /* input/output:
1736                                                 *   tp_name for tracepoint
1737                                                 *   symbol for kprobe
1738                                                 *   filename for uprobe
1739                                                 */
1740                __u32           prog_id;        /* output: prod_id */
1741                __u32           fd_type;        /* output: BPF_FD_TYPE_* */
1742                __u64           probe_offset;   /* output: probe_offset */
1743                __u64           probe_addr;     /* output: probe_addr */
1744        } task_fd_query;
1745
1746        struct { /* struct used by BPF_LINK_CREATE command */
1747                union {
1748                        __u32           prog_fd;        /* eBPF program to attach */
1749                        __u32           map_fd;         /* struct_ops to attach */
1750                };
1751                union {
1752                        __u32   target_fd;      /* target object to attach to or ... */
1753                        __u32   target_ifindex; /* target ifindex */
1754                };
1755                __u32           attach_type;    /* attach type */
1756                __u32           flags;          /* extra flags */
1757                union {
1758                        __u32   target_btf_id;  /* btf_id of target to attach to */
1759                        struct {
1760                                __aligned_u64   iter_info;      /* extra bpf_iter_link_info */
1761                                __u32           iter_info_len;  /* iter_info length */
1762                        };
1763                        struct {
1764                                /* black box user-provided value passed through
1765                                 * to BPF program at the execution time and
1766                                 * accessible through bpf_get_attach_cookie() BPF helper
1767                                 */
1768                                __u64           bpf_cookie;
1769                        } perf_event;
1770                        struct {
1771                                __u32           flags;
1772                                __u32           cnt;
1773                                __aligned_u64   syms;
1774                                __aligned_u64   addrs;
1775                                __aligned_u64   cookies;
1776                        } kprobe_multi;
1777                        struct {
1778                                /* this is overlaid with the target_btf_id above. */
1779                                __u32           target_btf_id;
1780                                /* black box user-provided value passed through
1781                                 * to BPF program at the execution time and
1782                                 * accessible through bpf_get_attach_cookie() BPF helper
1783                                 */
1784                                __u64           cookie;
1785                        } tracing;
1786                        struct {
1787                                __u32           pf;
1788                                __u32           hooknum;
1789                                __s32           priority;
1790                                __u32           flags;
1791                        } netfilter;
1792                        struct {
1793                                union {
1794                                        __u32   relative_fd;
1795                                        __u32   relative_id;
1796                                };
1797                                __u64           expected_revision;
1798                        } tcx;
1799                        struct {
1800                                __aligned_u64   path;
1801                                __aligned_u64   offsets;
1802                                __aligned_u64   ref_ctr_offsets;
1803                                __aligned_u64   cookies;
1804                                __u32           cnt;
1805                                __u32           flags;
1806                                __u32           pid;
1807                        } uprobe_multi;
1808                        struct {
1809                                union {
1810                                        __u32   relative_fd;
1811                                        __u32   relative_id;
1812                                };
1813                                __u64           expected_revision;
1814                        } netkit;
1815                        struct {
1816                                union {
1817                                        __u32   relative_fd;
1818                                        __u32   relative_id;
1819                                };
1820                                __u64           expected_revision;
1821                        } cgroup;
1822                };
1823        } link_create;
1824
1825        struct { /* struct used by BPF_LINK_UPDATE command */
1826                __u32           link_fd;        /* link fd */
1827                union {
1828                        /* new program fd to update link with */
1829                        __u32           new_prog_fd;
1830                        /* new struct_ops map fd to update link with */
1831                        __u32           new_map_fd;
1832                };
1833                __u32           flags;          /* extra flags */
1834                union {
1835                        /* expected link's program fd; is specified only if
1836                         * BPF_F_REPLACE flag is set in flags.
1837                         */
1838                        __u32           old_prog_fd;
1839                        /* expected link's map fd; is specified only
1840                         * if BPF_F_REPLACE flag is set.
1841                         */
1842                        __u32           old_map_fd;
1843                };
1844        } link_update;
1845
1846        struct {
1847                __u32           link_fd;
1848        } link_detach;
1849
1850        struct { /* struct used by BPF_ENABLE_STATS command */
1851                __u32           type;
1852        } enable_stats;
1853
1854        struct { /* struct used by BPF_ITER_CREATE command */
1855                __u32           link_fd;
1856                __u32           flags;
1857        } iter_create;
1858
1859        struct { /* struct used by BPF_PROG_BIND_MAP command */
1860                __u32           prog_fd;
1861                __u32           map_fd;
1862                __u32           flags;          /* extra flags */
1863        } prog_bind_map;
1864
1865        struct { /* struct used by BPF_TOKEN_CREATE command */
1866                __u32           flags;
1867                __u32           bpffs_fd;
1868        } token_create;
1869
1870        struct {
1871                __aligned_u64   stream_buf;
1872                __u32           stream_buf_len;
1873                __u32           stream_id;
1874                __u32           prog_fd;
1875        } prog_stream_read;
1876
1877} __attribute__((aligned(8)));
1878
1879/* The description below is an attempt at providing documentation to eBPF
1880 * developers about the multiple available eBPF helper functions. It can be
1881 * parsed and used to produce a manual page. The workflow is the following,
1882 * and requires the rst2man utility:
1883 *
1884 *     $ ./scripts/bpf_doc.py \
1885 *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1886 *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1887 *     $ man /tmp/bpf-helpers.7
1888 *
1889 * Note that in order to produce this external documentation, some RST
1890 * formatting is used in the descriptions to get "bold" and "italics" in
1891 * manual pages. Also note that the few trailing white spaces are
1892 * intentional, removing them would break paragraphs for rst2man.
1893 *
1894 * Start of BPF helper function descriptions:
1895 *
1896 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1897 *      Description
1898 *              Perform a lookup in *map* for an entry associated to *key*.
1899 *      Return
1900 *              Map value associated to *key*, or **NULL** if no entry was
1901 *              found.
1902 *
1903 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1904 *      Description
1905 *              Add or update the value of the entry associated to *key* in
1906 *              *map* with *value*. *flags* is one of:
1907 *
1908 *              **BPF_NOEXIST**
1909 *                      The entry for *key* must not exist in the map.
1910 *              **BPF_EXIST**
1911 *                      The entry for *key* must already exist in the map.
1912 *              **BPF_ANY**
1913 *                      No condition on the existence of the entry for *key*.
1914 *
1915 *              Flag value **BPF_NOEXIST** cannot be used for maps of types
1916 *              **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all
1917 *              elements always exist), the helper would return an error.
1918 *      Return
1919 *              0 on success, or a negative error in case of failure.
1920 *
1921 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1922 *      Description
1923 *              Delete entry with *key* from *map*.
1924 *      Return
1925 *              0 on success, or a negative error in case of failure.
1926 *
1927 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1928 *      Description
1929 *              For tracing programs, safely attempt to read *size* bytes from
1930 *              kernel space address *unsafe_ptr* and store the data in *dst*.
1931 *
1932 *              Generally, use **bpf_probe_read_user**\ () or
1933 *              **bpf_probe_read_kernel**\ () instead.
1934 *      Return
1935 *              0 on success, or a negative error in case of failure.
1936 *
1937 * u64 bpf_ktime_get_ns(void)
1938 *      Description
1939 *              Return the time elapsed since system boot, in nanoseconds.
1940 *              Does not include time the system was suspended.
1941 *              See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1942 *      Return
1943 *              Current *ktime*.
1944 *
1945 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1946 *      Description
1947 *              This helper is a "printk()-like" facility for debugging. It
1948 *              prints a message defined by format *fmt* (of size *fmt_size*)
1949 *              to file *\/sys/kernel/tracing/trace* from TraceFS, if
1950 *              available. It can take up to three additional **u64**
1951 *              arguments (as an eBPF helpers, the total number of arguments is
1952 *              limited to five).
1953 *
1954 *              Each time the helper is called, it appends a line to the trace.
1955 *              Lines are discarded while *\/sys/kernel/tracing/trace* is
1956 *              open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
1957 *              The format of the trace is customizable, and the exact output
1958 *              one will get depends on the options set in
1959 *              *\/sys/kernel/tracing/trace_options* (see also the
1960 *              *README* file under the same directory). However, it usually
1961 *              defaults to something like:
1962 *
1963 *              ::
1964 *
1965 *                      telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1966 *
1967 *              In the above:
1968 *
1969 *                      * ``telnet`` is the name of the current task.
1970 *                      * ``470`` is the PID of the current task.
1971 *                      * ``001`` is the CPU number on which the task is
1972 *                        running.
1973 *                      * In ``.N..``, each character refers to a set of
1974 *                        options (whether irqs are enabled, scheduling
1975 *                        options, whether hard/softirqs are running, level of
1976 *                        preempt_disabled respectively). **N** means that
1977 *                        **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1978 *                        are set.
1979 *                      * ``419421.045894`` is a timestamp.
1980 *                      * ``0x00000001`` is a fake value used by BPF for the
1981 *                        instruction pointer register.
1982 *                      * ``<formatted msg>`` is the message formatted with
1983 *                        *fmt*.
1984 *
1985 *              The conversion specifiers supported by *fmt* are similar, but
1986 *              more limited than for printk(). They are **%d**, **%i**,
1987 *              **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1988 *              **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1989 *              of field, padding with zeroes, etc.) is available, and the
1990 *              helper will return **-EINVAL** (but print nothing) if it
1991 *              encounters an unknown specifier.
1992 *
1993 *              Also, note that **bpf_trace_printk**\ () is slow, and should
1994 *              only be used for debugging purposes. For this reason, a notice
1995 *              block (spanning several lines) is printed to kernel logs and
1996 *              states that the helper should not be used "for production use"
1997 *              the first time this helper is used (or more precisely, when
1998 *              **trace_printk**\ () buffers are allocated). For passing values
1999 *              to user space, perf events should be preferred.
2000 *      Return
2001 *              The number of bytes written to the buffer, or a negative error
2002 *              in case of failure.
2003 *
2004 * u32 bpf_get_prandom_u32(void)
2005 *      Description
2006 *              Get a pseudo-random number.
2007 *
2008 *              From a security point of view, this helper uses its own
2009 *              pseudo-random internal state, and cannot be used to infer the
2010 *              seed of other random functions in the kernel. However, it is
2011 *              essential to note that the generator used by the helper is not
2012 *              cryptographically secure.
2013 *      Return
2014 *              A random 32-bit unsigned value.
2015 *
2016 * u32 bpf_get_smp_processor_id(void)
2017 *      Description
2018 *              Get the SMP (symmetric multiprocessing) processor id. Note that
2019 *              all programs run with migration disabled, which means that the
2020 *              SMP processor id is stable during all the execution of the
2021 *              program.
2022 *      Return
2023 *              The SMP id of the processor running the program.
2024 *      Attributes
2025 *              __bpf_fastcall
2026 *
2027 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
2028 *      Description
2029 *              Store *len* bytes from address *from* into the packet
2030 *              associated to *skb*, at *offset*. The *flags* are a combination
2031 *              of the following values:
2032 *
2033 *              **BPF_F_RECOMPUTE_CSUM**
2034 *                      Automatically update *skb*\ **->csum** after storing the
2035 *                      bytes.
2036 *              **BPF_F_INVALIDATE_HASH**
2037 *                      Set *skb*\ **->hash**, *skb*\ **->swhash** and *skb*\
2038 *                      **->l4hash** to 0.
2039 *
2040 *              A call to this helper is susceptible to change the underlying
2041 *              packet buffer. Therefore, at load time, all checks on pointers
2042 *              previously done by the verifier are invalidated and must be
2043 *              performed again, if the helper is used in combination with
2044 *              direct packet access.
2045 *      Return
2046 *              0 on success, or a negative error in case of failure.
2047 *
2048 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
2049 *      Description
2050 *              Recompute the layer 3 (e.g. IP) checksum for the packet
2051 *              associated to *skb*. Computation is incremental, so the helper
2052 *              must know the former value of the header field that was
2053 *              modified (*from*), the new value of this field (*to*), and the
2054 *              number of bytes (2 or 4) for this field, stored in *size*.
2055 *              Alternatively, it is possible to store the difference between
2056 *              the previous and the new values of the header field in *to*, by
2057 *              setting *from* and *size* to 0. For both methods, *offset*
2058 *              indicates the location of the IP checksum within the packet.
2059 *
2060 *              This helper works in combination with **bpf_csum_diff**\ (),
2061 *              which does not update the checksum in-place, but offers more
2062 *              flexibility and can handle sizes larger than 2 or 4 for the
2063 *              checksum to update.
2064 *
2065 *              A call to this helper is susceptible to change the underlying
2066 *              packet buffer. Therefore, at load time, all checks on pointers
2067 *              previously done by the verifier are invalidated and must be
2068 *              performed again, if the helper is used in combination with
2069 *              direct packet access.
2070 *      Return
2071 *              0 on success, or a negative error in case of failure.
2072 *
2073 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
2074 *      Description
2075 *              Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
2076 *              packet associated to *skb*. Computation is incremental, so the
2077 *              helper must know the former value of the header field that was
2078 *              modified (*from*), the new value of this field (*to*), and the
2079 *              number of bytes (2 or 4) for this field, stored on the lowest
2080 *              four bits of *flags*. Alternatively, it is possible to store
2081 *              the difference between the previous and the new values of the
2082 *              header field in *to*, by setting *from* and the four lowest
2083 *              bits of *flags* to 0. For both methods, *offset* indicates the
2084 *              location of the IP checksum within the packet. In addition to
2085 *              the size of the field, *flags* can be added (bitwise OR) actual
2086 *              flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
2087 *              untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
2088 *              for updates resulting in a null checksum the value is set to
2089 *              **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
2090 *              that the modified header field is part of the pseudo-header.
2091 *              Flag **BPF_F_IPV6** should be set for IPv6 packets.
2092 *
2093 *              This helper works in combination with **bpf_csum_diff**\ (),
2094 *              which does not update the checksum in-place, but offers more
2095 *              flexibility and can handle sizes larger than 2 or 4 for the
2096 *              checksum to update.
2097 *
2098 *              A call to this helper is susceptible to change the underlying
2099 *              packet buffer. Therefore, at load time, all checks on pointers
2100 *              previously done by the verifier are invalidated and must be
2101 *              performed again, if the helper is used in combination with
2102 *              direct packet access.
2103 *      Return
2104 *              0 on success, or a negative error in case of failure.
2105 *
2106 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
2107 *      Description
2108 *              This special helper is used to trigger a "tail call", or in
2109 *              other words, to jump into another eBPF program. The same stack
2110 *              frame is used (but values on stack and in registers for the
2111 *              caller are not accessible to the callee). This mechanism allows
2112 *              for program chaining, either for raising the maximum number of
2113 *              available eBPF instructions, or to execute given programs in
2114 *              conditional blocks. For security reasons, there is an upper
2115 *              limit to the number of successive tail calls that can be
2116 *              performed.
2117 *
2118 *              Upon call of this helper, the program attempts to jump into a
2119 *              program referenced at index *index* in *prog_array_map*, a
2120 *              special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
2121 *              *ctx*, a pointer to the context.
2122 *
2123 *              If the call succeeds, the kernel immediately runs the first
2124 *              instruction of the new program. This is not a function call,
2125 *              and it never returns to the previous program. If the call
2126 *              fails, then the helper has no effect, and the caller continues
2127 *              to run its subsequent instructions. A call can fail if the
2128 *              destination program for the jump does not exist (i.e. *index*
2129 *              is superior to the number of entries in *prog_array_map*), or
2130 *              if the maximum number of tail calls has been reached for this
2131 *              chain of programs. This limit is defined in the kernel by the
2132 *              macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
2133 *              which is currently set to 33.
2134 *      Return
2135 *              0 on success, or a negative error in case of failure.
2136 *
2137 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
2138 *      Description
2139 *              Clone and redirect the packet associated to *skb* to another
2140 *              net device of index *ifindex*. Both ingress and egress
2141 *              interfaces can be used for redirection. The **BPF_F_INGRESS**
2142 *              value in *flags* is used to make the distinction (ingress path
2143 *              is selected if the flag is present, egress path otherwise).
2144 *              This is the only flag supported for now.
2145 *
2146 *              In comparison with **bpf_redirect**\ () helper,
2147 *              **bpf_clone_redirect**\ () has the associated cost of
2148 *              duplicating the packet buffer, but this can be executed out of
2149 *              the eBPF program. Conversely, **bpf_redirect**\ () is more
2150 *              efficient, but it is handled through an action code where the
2151 *              redirection happens only after the eBPF program has returned.
2152 *
2153 *              A call to this helper is susceptible to change the underlying
2154 *              packet buffer. Therefore, at load time, all checks on pointers
2155 *              previously done by the verifier are invalidated and must be
2156 *              performed again, if the helper is used in combination with
2157 *              direct packet access.
2158 *      Return
2159 *              0 on success, or a negative error in case of failure. Positive
2160 *              error indicates a potential drop or congestion in the target
2161 *              device. The particular positive error codes are not defined.
2162 *
2163 * u64 bpf_get_current_pid_tgid(void)
2164 *      Description
2165 *              Get the current pid and tgid.
2166 *      Return
2167 *              A 64-bit integer containing the current tgid and pid, and
2168 *              created as such:
2169 *              *current_task*\ **->tgid << 32 \|**
2170 *              *current_task*\ **->pid**.
2171 *
2172 * u64 bpf_get_current_uid_gid(void)
2173 *      Description
2174 *              Get the current uid and gid.
2175 *      Return
2176 *              A 64-bit integer containing the current GID and UID, and
2177 *              created as such: *current_gid* **<< 32 \|** *current_uid*.
2178 *
2179 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
2180 *      Description
2181 *              Copy the **comm** attribute of the current task into *buf* of
2182 *              *size_of_buf*. The **comm** attribute contains the name of
2183 *              the executable (excluding the path) for the current task. The
2184 *              *size_of_buf* must be strictly positive. On success, the
2185 *              helper makes sure that the *buf* is NUL-terminated. On failure,
2186 *              it is filled with zeroes.
2187 *      Return
2188 *              0 on success, or a negative error in case of failure.
2189 *
2190 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
2191 *      Description
2192 *              Retrieve the classid for the current task, i.e. for the net_cls
2193 *              cgroup to which *skb* belongs.
2194 *
2195 *              This helper can be used on TC egress path, but not on ingress.
2196 *
2197 *              The net_cls cgroup provides an interface to tag network packets
2198 *              based on a user-provided identifier for all traffic coming from
2199 *              the tasks belonging to the related cgroup. See also the related
2200 *              kernel documentation, available from the Linux sources in file
2201 *              *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
2202 *
2203 *              The Linux kernel has two versions for cgroups: there are
2204 *              cgroups v1 and cgroups v2. Both are available to users, who can
2205 *              use a mixture of them, but note that the net_cls cgroup is for
2206 *              cgroup v1 only. This makes it incompatible with BPF programs
2207 *              run on cgroups, which is a cgroup-v2-only feature (a socket can
2208 *              only hold data for one version of cgroups at a time).
2209 *
2210 *              This helper is only available is the kernel was compiled with
2211 *              the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
2212 *              "**y**" or to "**m**".
2213 *      Return
2214 *              The classid, or 0 for the default unconfigured classid.
2215 *
2216 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
2217 *      Description
2218 *              Push a *vlan_tci* (VLAN tag control information) of protocol
2219 *              *vlan_proto* to the packet associated to *skb*, then update
2220 *              the checksum. Note that if *vlan_proto* is different from
2221 *              **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
2222 *              be **ETH_P_8021Q**.
2223 *
2224 *              A call to this helper is susceptible to change the underlying
2225 *              packet buffer. Therefore, at load time, all checks on pointers
2226 *              previously done by the verifier are invalidated and must be
2227 *              performed again, if the helper is used in combination with
2228 *              direct packet access.
2229 *      Return
2230 *              0 on success, or a negative error in case of failure.
2231 *
2232 * long bpf_skb_vlan_pop(struct sk_buff *skb)
2233 *      Description
2234 *              Pop a VLAN header from the packet associated to *skb*.
2235 *
2236 *              A call to this helper is susceptible to change the underlying
2237 *              packet buffer. Therefore, at load time, all checks on pointers
2238 *              previously done by the verifier are invalidated and must be
2239 *              performed again, if the helper is used in combination with
2240 *              direct packet access.
2241 *      Return
2242 *              0 on success, or a negative error in case of failure.
2243 *
2244 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2245 *      Description
2246 *              Get tunnel metadata. This helper takes a pointer *key* to an
2247 *              empty **struct bpf_tunnel_key** of **size**, that will be
2248 *              filled with tunnel metadata for the packet associated to *skb*.
2249 *              The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
2250 *              indicates that the tunnel is based on IPv6 protocol instead of
2251 *              IPv4.
2252 *
2253 *              The **struct bpf_tunnel_key** is an object that generalizes the
2254 *              principal parameters used by various tunneling protocols into a
2255 *              single struct. This way, it can be used to easily make a
2256 *              decision based on the contents of the encapsulation header,
2257 *              "summarized" in this struct. In particular, it holds the IP
2258 *              address of the remote end (IPv4 or IPv6, depending on the case)
2259 *              in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
2260 *              this struct exposes the *key*\ **->tunnel_id**, which is
2261 *              generally mapped to a VNI (Virtual Network Identifier), making
2262 *              it programmable together with the **bpf_skb_set_tunnel_key**\
2263 *              () helper.
2264 *
2265 *              Let's imagine that the following code is part of a program
2266 *              attached to the TC ingress interface, on one end of a GRE
2267 *              tunnel, and is supposed to filter out all messages coming from
2268 *              remote ends with IPv4 address other than 10.0.0.1:
2269 *
2270 *              ::
2271 *
2272 *                      int ret;
2273 *                      struct bpf_tunnel_key key = {};
2274 *
2275 *                      ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
2276 *                      if (ret < 0)
2277 *                              return TC_ACT_SHOT;     // drop packet
2278 *
2279 *                      if (key.remote_ipv4 != 0x0a000001)
2280 *                              return TC_ACT_SHOT;     // drop packet
2281 *
2282 *                      return TC_ACT_OK;               // accept packet
2283 *
2284 *              This interface can also be used with all encapsulation devices
2285 *              that can operate in "collect metadata" mode: instead of having
2286 *              one network device per specific configuration, the "collect
2287 *              metadata" mode only requires a single device where the
2288 *              configuration can be extracted from this helper.
2289 *
2290 *              This can be used together with various tunnels such as VXLan,
2291 *              Geneve, GRE or IP in IP (IPIP).
2292 *      Return
2293 *              0 on success, or a negative error in case of failure.
2294 *
2295 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2296 *      Description
2297 *              Populate tunnel metadata for packet associated to *skb.* The
2298 *              tunnel metadata is set to the contents of *key*, of *size*. The
2299 *              *flags* can be set to a combination of the following values:
2300 *
2301 *              **BPF_F_TUNINFO_IPV6**
2302 *                      Indicate that the tunnel is based on IPv6 protocol
2303 *                      instead of IPv4.
2304 *              **BPF_F_ZERO_CSUM_TX**
2305 *                      For IPv4 packets, add a flag to tunnel metadata
2306 *                      indicating that checksum computation should be skipped
2307 *                      and checksum set to zeroes.
2308 *              **BPF_F_DONT_FRAGMENT**
2309 *                      Add a flag to tunnel metadata indicating that the
2310 *                      packet should not be fragmented.
2311 *              **BPF_F_SEQ_NUMBER**
2312 *                      Add a flag to tunnel metadata indicating that a
2313 *                      sequence number should be added to tunnel header before
2314 *                      sending the packet. This flag was added for GRE
2315 *                      encapsulation, but might be used with other protocols
2316 *                      as well in the future.
2317 *              **BPF_F_NO_TUNNEL_KEY**
2318 *                      Add a flag to tunnel metadata indicating that no tunnel
2319 *                      key should be set in the resulting tunnel header.
2320 *
2321 *              Here is a typical usage on the transmit path:
2322 *
2323 *              ::
2324 *
2325 *                      struct bpf_tunnel_key key;
2326 *                           populate key ...
2327 *                      bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2328 *                      bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2329 *
2330 *              See also the description of the **bpf_skb_get_tunnel_key**\ ()
2331 *              helper for additional information.
2332 *      Return
2333 *              0 on success, or a negative error in case of failure.
2334 *
2335 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2336 *      Description
2337 *              Read the value of a perf event counter. This helper relies on a
2338 *              *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2339 *              the perf event counter is selected when *map* is updated with
2340 *              perf event file descriptors. The *map* is an array whose size
2341 *              is the number of available CPUs, and each cell contains a value
2342 *              relative to one CPU. The value to retrieve is indicated by
2343 *              *flags*, that contains the index of the CPU to look up, masked
2344 *              with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2345 *              **BPF_F_CURRENT_CPU** to indicate that the value for the
2346 *              current CPU should be retrieved.
2347 *
2348 *              Note that before Linux 4.13, only hardware perf event can be
2349 *              retrieved.
2350 *
2351 *              Also, be aware that the newer helper
2352 *              **bpf_perf_event_read_value**\ () is recommended over
2353 *              **bpf_perf_event_read**\ () in general. The latter has some ABI
2354 *              quirks where error and counter value are used as a return code
2355 *              (which is wrong to do since ranges may overlap). This issue is
2356 *              fixed with **bpf_perf_event_read_value**\ (), which at the same
2357 *              time provides more features over the **bpf_perf_event_read**\
2358 *              () interface. Please refer to the description of
2359 *              **bpf_perf_event_read_value**\ () for details.
2360 *      Return
2361 *              The value of the perf event counter read from the map, or a
2362 *              negative error code in case of failure.
2363 *
2364 * long bpf_redirect(u32 ifindex, u64 flags)
2365 *      Description
2366 *              Redirect the packet to another net device of index *ifindex*.
2367 *              This helper is somewhat similar to **bpf_clone_redirect**\
2368 *              (), except that the packet is not cloned, which provides
2369 *              increased performance.
2370 *
2371 *              Except for XDP, both ingress and egress interfaces can be used
2372 *              for redirection. The **BPF_F_INGRESS** value in *flags* is used
2373 *              to make the distinction (ingress path is selected if the flag
2374 *              is present, egress path otherwise). Currently, XDP only
2375 *              supports redirection to the egress interface, and accepts no
2376 *              flag at all.
2377 *
2378 *              The same effect can also be attained with the more generic
2379 *              **bpf_redirect_map**\ (), which uses a BPF map to store the
2380 *              redirect target instead of providing it directly to the helper.
2381 *      Return
2382 *              For XDP, the helper returns **XDP_REDIRECT** on success or
2383 *              **XDP_ABORTED** on error. For other program types, the values
2384 *              are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2385 *              error.
2386 *
2387 * u32 bpf_get_route_realm(struct sk_buff *skb)
2388 *      Description
2389 *              Retrieve the realm or the route, that is to say the
2390 *              **tclassid** field of the destination for the *skb*. The
2391 *              identifier retrieved is a user-provided tag, similar to the
2392 *              one used with the net_cls cgroup (see description for
2393 *              **bpf_get_cgroup_classid**\ () helper), but here this tag is
2394 *              held by a route (a destination entry), not by a task.
2395 *
2396 *              Retrieving this identifier works with the clsact TC egress hook
2397 *              (see also **tc-bpf(8)**), or alternatively on conventional
2398 *              classful egress qdiscs, but not on TC ingress path. In case of
2399 *              clsact TC egress hook, this has the advantage that, internally,
2400 *              the destination entry has not been dropped yet in the transmit
2401 *              path. Therefore, the destination entry does not need to be
2402 *              artificially held via **netif_keep_dst**\ () for a classful
2403 *              qdisc until the *skb* is freed.
2404 *
2405 *              This helper is available only if the kernel was compiled with
2406 *              **CONFIG_IP_ROUTE_CLASSID** configuration option.
2407 *      Return
2408 *              The realm of the route for the packet associated to *skb*, or 0
2409 *              if none was found.
2410 *
2411 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2412 *      Description
2413 *              Write raw *data* blob into a special BPF perf event held by
2414 *              *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2415 *              event must have the following attributes: **PERF_SAMPLE_RAW**
2416 *              as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2417 *              **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2418 *
2419 *              The *flags* are used to indicate the index in *map* for which
2420 *              the value must be put, masked with **BPF_F_INDEX_MASK**.
2421 *              Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2422 *              to indicate that the index of the current CPU core should be
2423 *              used.
2424 *
2425 *              The value to write, of *size*, is passed through eBPF stack and
2426 *              pointed by *data*.
2427 *
2428 *              The context of the program *ctx* needs also be passed to the
2429 *              helper.
2430 *
2431 *              On user space, a program willing to read the values needs to
2432 *              call **perf_event_open**\ () on the perf event (either for
2433 *              one or for all CPUs) and to store the file descriptor into the
2434 *              *map*. This must be done before the eBPF program can send data
2435 *              into it. An example is available in file
2436 *              *samples/bpf/trace_output_user.c* in the Linux kernel source
2437 *              tree (the eBPF program counterpart is in
2438 *              *samples/bpf/trace_output.bpf.c*).
2439 *
2440 *              **bpf_perf_event_output**\ () achieves better performance
2441 *              than **bpf_trace_printk**\ () for sharing data with user
2442 *              space, and is much better suitable for streaming data from eBPF
2443 *              programs.
2444 *
2445 *              Note that this helper is not restricted to tracing use cases
2446 *              and can be used with programs attached to TC or XDP as well,
2447 *              where it allows for passing data to user space listeners. Data
2448 *              can be:
2449 *
2450 *              * Only custom structs,
2451 *              * Only the packet payload, or
2452 *              * A combination of both.
2453 *      Return
2454 *              0 on success, or a negative error in case of failure.
2455 *
2456 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2457 *      Description
2458 *              This helper was provided as an easy way to load data from a
2459 *              packet. It can be used to load *len* bytes from *offset* from
2460 *              the packet associated to *skb*, into the buffer pointed by
2461 *              *to*.
2462 *
2463 *              Since Linux 4.7, usage of this helper has mostly been replaced
2464 *              by "direct packet access", enabling packet data to be
2465 *              manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2466 *              pointing respectively to the first byte of packet data and to
2467 *              the byte after the last byte of packet data. However, it
2468 *              remains useful if one wishes to read large quantities of data
2469 *              at once from a packet into the eBPF stack.
2470 *      Return
2471 *              0 on success, or a negative error in case of failure.
2472 *
2473 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2474 *      Description
2475 *              Walk a user or a kernel stack and return its id. To achieve
2476 *              this, the helper needs *ctx*, which is a pointer to the context
2477 *              on which the tracing program is executed, and a pointer to a
2478 *              *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2479 *
2480 *              The last argument, *flags*, holds the number of stack frames to
2481 *              skip (from 0 to 255), masked with
2482 *              **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2483 *              a combination of the following flags:
2484 *
2485 *              **BPF_F_USER_STACK**
2486 *                      Collect a user space stack instead of a kernel stack.
2487 *              **BPF_F_FAST_STACK_CMP**
2488 *                      Compare stacks by hash only.
2489 *              **BPF_F_REUSE_STACKID**
2490 *                      If two different stacks hash into the same *stackid*,
2491 *                      discard the old one.
2492 *
2493 *              The stack id retrieved is a 32 bit long integer handle which
2494 *              can be further combined with other data (including other stack
2495 *              ids) and used as a key into maps. This can be useful for
2496 *              generating a variety of graphs (such as flame graphs or off-cpu
2497 *              graphs).
2498 *
2499 *              For walking a stack, this helper is an improvement over
2500 *              **bpf_probe_read**\ (), which can be used with unrolled loops
2501 *              but is not efficient and consumes a lot of eBPF instructions.
2502 *              Instead, **bpf_get_stackid**\ () can collect up to
2503 *              **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2504 *              this limit can be controlled with the **sysctl** program, and
2505 *              that it should be manually increased in order to profile long
2506 *              user stacks (such as stacks for Java programs). To do so, use:
2507 *
2508 *              ::
2509 *
2510 *                      # sysctl kernel.perf_event_max_stack=<new value>
2511 *      Return
2512 *              The positive or null stack id on success, or a negative error
2513 *              in case of failure.
2514 *
2515 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2516 *      Description
2517 *              Compute a checksum difference, from the raw buffer pointed by
2518 *              *from*, of length *from_size* (that must be a multiple of 4),
2519 *              towards the raw buffer pointed by *to*, of size *to_size*
2520 *              (same remark). An optional *seed* can be added to the value
2521 *              (this can be cascaded, the seed may come from a previous call
2522 *              to the helper).
2523 *
2524 *              This is flexible enough to be used in several ways:
2525 *
2526 *              * With *from_size* == 0, *to_size* > 0 and *seed* set to
2527 *                checksum, it can be used when pushing new data.
2528 *              * With *from_size* > 0, *to_size* == 0 and *seed* set to
2529 *                checksum, it can be used when removing data from a packet.
2530 *              * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2531 *                can be used to compute a diff. Note that *from_size* and
2532 *                *to_size* do not need to be equal.
2533 *
2534 *              This helper can be used in combination with
2535 *              **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2536 *              which one can feed in the difference computed with
2537 *              **bpf_csum_diff**\ ().
2538 *      Return
2539 *              The checksum result, or a negative error code in case of
2540 *              failure.
2541 *
2542 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2543 *      Description
2544 *              Retrieve tunnel options metadata for the packet associated to
2545 *              *skb*, and store the raw tunnel option data to the buffer *opt*
2546 *              of *size*.
2547 *
2548 *              This helper can be used with encapsulation devices that can
2549 *              operate in "collect metadata" mode (please refer to the related
2550 *              note in the description of **bpf_skb_get_tunnel_key**\ () for
2551 *              more details). A particular example where this can be used is
2552 *              in combination with the Geneve encapsulation protocol, where it
2553 *              allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2554 *              and retrieving arbitrary TLVs (Type-Length-Value headers) from
2555 *              the eBPF program. This allows for full customization of these
2556 *              headers.
2557 *      Return
2558 *              The size of the option data retrieved.
2559 *
2560 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2561 *      Description
2562 *              Set tunnel options metadata for the packet associated to *skb*
2563 *              to the option data contained in the raw buffer *opt* of *size*.
2564 *
2565 *              See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2566 *              helper for additional information.
2567 *      Return
2568 *              0 on success, or a negative error in case of failure.
2569 *
2570 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2571 *      Description
2572 *              Change the protocol of the *skb* to *proto*. Currently
2573 *              supported are transition from IPv4 to IPv6, and from IPv6 to
2574 *              IPv4. The helper takes care of the groundwork for the
2575 *              transition, including resizing the socket buffer. The eBPF
2576 *              program is expected to fill the new headers, if any, via
2577 *              **skb_store_bytes**\ () and to recompute the checksums with
2578 *              **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2579 *              (). The main case for this helper is to perform NAT64
2580 *              operations out of an eBPF program.
2581 *
2582 *              Internally, the GSO type is marked as dodgy so that headers are
2583 *              checked and segments are recalculated by the GSO/GRO engine.
2584 *              The size for GSO target is adapted as well.
2585 *
2586 *              All values for *flags* are reserved for future usage, and must
2587 *              be left at zero.
2588 *
2589 *              A call to this helper is susceptible to change the underlying
2590 *              packet buffer. Therefore, at load time, all checks on pointers
2591 *              previously done by the verifier are invalidated and must be
2592 *              performed again, if the helper is used in combination with
2593 *              direct packet access.
2594 *      Return
2595 *              0 on success, or a negative error in case of failure.
2596 *
2597 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2598 *      Description
2599 *              Change the packet type for the packet associated to *skb*. This
2600 *              comes down to setting *skb*\ **->pkt_type** to *type*, except
2601 *              the eBPF program does not have a write access to *skb*\
2602 *              **->pkt_type** beside this helper. Using a helper here allows
2603 *              for graceful handling of errors.
2604 *
2605 *              The major use case is to change incoming *skb*s to
2606 *              **PACKET_HOST** in a programmatic way instead of having to
2607 *              recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2608 *              example.
2609 *
2610 *              Note that *type* only allows certain values. At this time, they
2611 *              are:
2612 *
2613 *              **PACKET_HOST**
2614 *                      Packet is for us.
2615 *              **PACKET_BROADCAST**
2616 *                      Send packet to all.
2617 *              **PACKET_MULTICAST**
2618 *                      Send packet to group.
2619 *              **PACKET_OTHERHOST**
2620 *                      Send packet to someone else.
2621 *      Return
2622 *              0 on success, or a negative error in case of failure.
2623 *
2624 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2625 *      Description
2626 *              Check whether *skb* is a descendant of the cgroup2 held by
2627 *              *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2628 *      Return
2629 *              The return value depends on the result of the test, and can be:
2630 *
2631 *              * 0, if the *skb* failed the cgroup2 descendant test.
2632 *              * 1, if the *skb* succeeded the cgroup2 descendant test.
2633 *              * A negative error code, if an error occurred.
2634 *
2635 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2636 *      Description
2637 *              Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2638 *              not set, in particular if the hash was cleared due to mangling,
2639 *              recompute this hash. Later accesses to the hash can be done
2640 *              directly with *skb*\ **->hash**.
2641 *
2642 *              Calling **bpf_set_hash_invalid**\ (), changing a packet
2643 *              prototype with **bpf_skb_change_proto**\ (), or calling
2644 *              **bpf_skb_store_bytes**\ () with the
2645 *              **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2646 *              the hash and to trigger a new computation for the next call to
2647 *              **bpf_get_hash_recalc**\ ().
2648 *      Return
2649 *              The 32-bit hash.
2650 *
2651 * u64 bpf_get_current_task(void)
2652 *      Description
2653 *              Get the current task.
2654 *      Return
2655 *              A pointer to the current task struct.
2656 *
2657 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2658 *      Description
2659 *              Attempt in a safe way to write *len* bytes from the buffer
2660 *              *src* to *dst* in memory. It only works for threads that are in
2661 *              user context, and *dst* must be a valid user space address.
2662 *
2663 *              This helper should not be used to implement any kind of
2664 *              security mechanism because of TOC-TOU attacks, but rather to
2665 *              debug, divert, and manipulate execution of semi-cooperative
2666 *              processes.
2667 *
2668 *              Keep in mind that this feature is meant for experiments, and it
2669 *              has a risk of crashing the system and running programs.
2670 *              Therefore, when an eBPF program using this helper is attached,
2671 *              a warning including PID and process name is printed to kernel
2672 *              logs.
2673 *      Return
2674 *              0 on success, or a negative error in case of failure.
2675 *
2676 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2677 *      Description
2678 *              Check whether the probe is being run is the context of a given
2679 *              subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2680 *              *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2681 *      Return
2682 *              The return value depends on the result of the test, and can be:
2683 *
2684 *              * 1, if current task belongs to the cgroup2.
2685 *              * 0, if current task does not belong to the cgroup2.
2686 *              * A negative error code, if an error occurred.
2687 *
2688 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2689 *      Description
2690 *              Resize (trim or grow) the packet associated to *skb* to the
2691 *              new *len*. The *flags* are reserved for future usage, and must
2692 *              be left at zero.
2693 *
2694 *              The basic idea is that the helper performs the needed work to
2695 *              change the size of the packet, then the eBPF program rewrites
2696 *              the rest via helpers like **bpf_skb_store_bytes**\ (),
2697 *              **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2698 *              and others. This helper is a slow path utility intended for
2699 *              replies with control messages. And because it is targeted for
2700 *              slow path, the helper itself can afford to be slow: it
2701 *              implicitly linearizes, unclones and drops offloads from the
2702 *              *skb*.
2703 *
2704 *              A call to this helper is susceptible to change the underlying
2705 *              packet buffer. Therefore, at load time, all checks on pointers
2706 *              previously done by the verifier are invalidated and must be
2707 *              performed again, if the helper is used in combination with
2708 *              direct packet access.
2709 *      Return
2710 *              0 on success, or a negative error in case of failure.
2711 *
2712 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2713 *      Description
2714 *              Pull in non-linear data in case the *skb* is non-linear and not
2715 *              all of *len* are part of the linear section. Make *len* bytes
2716 *              from *skb* readable and writable. If a zero value is passed for
2717 *              *len*, then all bytes in the linear part of *skb* will be made
2718 *              readable and writable.
2719 *
2720 *              This helper is only needed for reading and writing with direct
2721 *              packet access.
2722 *
2723 *              For direct packet access, testing that offsets to access
2724 *              are within packet boundaries (test on *skb*\ **->data_end**) is
2725 *              susceptible to fail if offsets are invalid, or if the requested
2726 *              data is in non-linear parts of the *skb*. On failure the
2727 *              program can just bail out, or in the case of a non-linear
2728 *              buffer, use a helper to make the data available. The
2729 *              **bpf_skb_load_bytes**\ () helper is a first solution to access
2730 *              the data. Another one consists in using **bpf_skb_pull_data**
2731 *              to pull in once the non-linear parts, then retesting and
2732 *              eventually access the data.
2733 *
2734 *              At the same time, this also makes sure the *skb* is uncloned,
2735 *              which is a necessary condition for direct write. As this needs
2736 *              to be an invariant for the write part only, the verifier
2737 *              detects writes and adds a prologue that is calling
2738 *              **bpf_skb_pull_data()** to effectively unclone the *skb* from
2739 *              the very beginning in case it is indeed cloned.
2740 *
2741 *              A call to this helper is susceptible to change the underlying
2742 *              packet buffer. Therefore, at load time, all checks on pointers
2743 *              previously done by the verifier are invalidated and must be
2744 *              performed again, if the helper is used in combination with
2745 *              direct packet access.
2746 *      Return
2747 *              0 on success, or a negative error in case of failure.
2748 *
2749 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2750 *      Description
2751 *              Add the checksum *csum* into *skb*\ **->csum** in case the
2752 *              driver has supplied a checksum for the entire packet into that
2753 *              field. Return an error otherwise. This helper is intended to be
2754 *              used in combination with **bpf_csum_diff**\ (), in particular
2755 *              when the checksum needs to be updated after data has been
2756 *              written into the packet through direct packet access.
2757 *      Return
2758 *              The checksum on success, or a negative error code in case of
2759 *              failure.
2760 *
2761 * void bpf_set_hash_invalid(struct sk_buff *skb)
2762 *      Description
2763 *              Invalidate the current *skb*\ **->hash**. It can be used after
2764 *              mangling on headers through direct packet access, in order to
2765 *              indicate that the hash is outdated and to trigger a
2766 *              recalculation the next time the kernel tries to access this
2767 *              hash or when the **bpf_get_hash_recalc**\ () helper is called.
2768 *      Return
2769 *              void.
2770 *
2771 * long bpf_get_numa_node_id(void)
2772 *      Description
2773 *              Return the id of the current NUMA node. The primary use case
2774 *              for this helper is the selection of sockets for the local NUMA
2775 *              node, when the program is attached to sockets using the
2776 *              **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2777 *              but the helper is also available to other eBPF program types,
2778 *              similarly to **bpf_get_smp_processor_id**\ ().
2779 *      Return
2780 *              The id of current NUMA node.
2781 *
2782 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2783 *      Description
2784 *              Grows headroom of packet associated to *skb* and adjusts the
2785 *              offset of the MAC header accordingly, adding *len* bytes of
2786 *              space. It automatically extends and reallocates memory as
2787 *              required.
2788 *
2789 *              This helper can be used on a layer 3 *skb* to push a MAC header
2790 *              for redirection into a layer 2 device.
2791 *
2792 *              All values for *flags* are reserved for future usage, and must
2793 *              be left at zero.
2794 *
2795 *              A call to this helper is susceptible to change the underlying
2796 *              packet buffer. Therefore, at load time, all checks on pointers
2797 *              previously done by the verifier are invalidated and must be
2798 *              performed again, if the helper is used in combination with
2799 *              direct packet access.
2800 *      Return
2801 *              0 on success, or a negative error in case of failure.
2802 *
2803 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2804 *      Description
2805 *              Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2806 *              it is possible to use a negative value for *delta*. This helper
2807 *              can be used to prepare the packet for pushing or popping
2808 *              headers.
2809 *
2810 *              A call to this helper is susceptible to change the underlying
2811 *              packet buffer. Therefore, at load time, all checks on pointers
2812 *              previously done by the verifier are invalidated and must be
2813 *              performed again, if the helper is used in combination with
2814 *              direct packet access.
2815 *      Return
2816 *              0 on success, or a negative error in case of failure.
2817 *
2818 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2819 *      Description
2820 *              Copy a NUL terminated string from an unsafe kernel address
2821 *              *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2822 *              more details.
2823 *
2824 *              Generally, use **bpf_probe_read_user_str**\ () or
2825 *              **bpf_probe_read_kernel_str**\ () instead.
2826 *      Return
2827 *              On success, the strictly positive length of the string,
2828 *              including the trailing NUL character. On error, a negative
2829 *              value.
2830 *
2831 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2832 *      Description
2833 *              If the **struct sk_buff** pointed by *skb* has a known socket,
2834 *              retrieve the cookie (generated by the kernel) of this socket.
2835 *              If no cookie has been set yet, generate a new cookie. Once
2836 *              generated, the socket cookie remains stable for the life of the
2837 *              socket. This helper can be useful for monitoring per socket
2838 *              networking traffic statistics as it provides a global socket
2839 *              identifier that can be assumed unique.
2840 *      Return
2841 *              A 8-byte long unique number on success, or 0 if the socket
2842 *              field is missing inside *skb*.
2843 *
2844 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2845 *      Description
2846 *              Equivalent to bpf_get_socket_cookie() helper that accepts
2847 *              *skb*, but gets socket from **struct bpf_sock_addr** context.
2848 *      Return
2849 *              A 8-byte long unique number.
2850 *
2851 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2852 *      Description
2853 *              Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2854 *              *skb*, but gets socket from **struct bpf_sock_ops** context.
2855 *      Return
2856 *              A 8-byte long unique number.
2857 *
2858 * u64 bpf_get_socket_cookie(struct sock *sk)
2859 *      Description
2860 *              Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2861 *              *sk*, but gets socket from a BTF **struct sock**. This helper
2862 *              also works for sleepable programs.
2863 *      Return
2864 *              A 8-byte long unique number or 0 if *sk* is NULL.
2865 *
2866 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2867 *      Description
2868 *              Get the owner UID of the socked associated to *skb*.
2869 *      Return
2870 *              The owner UID of the socket associated to *skb*. If the socket
2871 *              is **NULL**, or if it is not a full socket (i.e. if it is a
2872 *              time-wait or a request socket instead), **overflowuid** value
2873 *              is returned (note that **overflowuid** might also be the actual
2874 *              UID value for the socket).
2875 *
2876 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2877 *      Description
2878 *              Set the full hash for *skb* (set the field *skb*\ **->hash**)
2879 *              to value *hash*.
2880 *      Return
2881 *              0
2882 *
2883 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2884 *      Description
2885 *              Emulate a call to **setsockopt()** on the socket associated to
2886 *              *bpf_socket*, which must be a full socket. The *level* at
2887 *              which the option resides and the name *optname* of the option
2888 *              must be specified, see **setsockopt(2)** for more information.
2889 *              The option value of length *optlen* is pointed by *optval*.
2890 *
2891 *              *bpf_socket* should be one of the following:
2892 *
2893 *              * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2894 *              * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
2895 *                **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
2896 *
2897 *              This helper actually implements a subset of **setsockopt()**.
2898 *              It supports the following *level*\ s:
2899 *
2900 *              * **SOL_SOCKET**, which supports the following *optname*\ s:
2901 *                **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2902 *                **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2903 *                **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2904 *                **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2905 *              * **IPPROTO_TCP**, which supports the following *optname*\ s:
2906 *                **TCP_CONGESTION**, **TCP_BPF_IW**,
2907 *                **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2908 *                **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2909 *                **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2910 *                **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2911 *                **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2912 *                **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**.
2913 *              * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2914 *              * **IPPROTO_IPV6**, which supports the following *optname*\ s:
2915 *                **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2916 *      Return
2917 *              0 on success, or a negative error in case of failure.
2918 *
2919 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2920 *      Description
2921 *              Grow or shrink the room for data in the packet associated to
2922 *              *skb* by *len_diff*, and according to the selected *mode*.
2923 *
2924 *              By default, the helper will reset any offloaded checksum
2925 *              indicator of the skb to CHECKSUM_NONE. This can be avoided
2926 *              by the following flag:
2927 *
2928 *              * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2929 *                checksum data of the skb to CHECKSUM_NONE.
2930 *
2931 *              There are two supported modes at this time:
2932 *
2933 *              * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2934 *                (room space is added or removed between the layer 2 and
2935 *                layer 3 headers).
2936 *
2937 *              * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2938 *                (room space is added or removed between the layer 3 and
2939 *                layer 4 headers).
2940 *
2941 *              The following flags are supported at this time:
2942 *
2943 *              * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2944 *                Adjusting mss in this way is not allowed for datagrams.
2945 *
2946 *              * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2947 *                **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2948 *                Any new space is reserved to hold a tunnel header.
2949 *                Configure skb offsets and other fields accordingly.
2950 *
2951 *              * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2952 *                **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2953 *                Use with ENCAP_L3 flags to further specify the tunnel type.
2954 *
2955 *              * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2956 *                Use with ENCAP_L3/L4 flags to further specify the tunnel
2957 *                type; *len* is the length of the inner MAC header.
2958 *
2959 *              * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2960 *                Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2961 *                L2 type as Ethernet.
2962 *
2963 *              * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
2964 *                **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
2965 *                Indicate the new IP header version after decapsulating the outer
2966 *                IP header. Used when the inner and outer IP versions are different.
2967 *
2968 *              A call to this helper is susceptible to change the underlying
2969 *              packet buffer. Therefore, at load time, all checks on pointers
2970 *              previously done by the verifier are invalidated and must be
2971 *              performed again, if the helper is used in combination with
2972 *              direct packet access.
2973 *      Return
2974 *              0 on success, or a negative error in case of failure.
2975 *
2976 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
2977 *      Description
2978 *              Redirect the packet to the endpoint referenced by *map* at
2979 *              index *key*. Depending on its type, this *map* can contain
2980 *              references to net devices (for forwarding packets through other
2981 *              ports), or to CPUs (for redirecting XDP frames to another CPU;
2982 *              but this is only implemented for native XDP (with driver
2983 *              support) as of this writing).
2984 *
2985 *              The lower two bits of *flags* are used as the return code if
2986 *              the map lookup fails. This is so that the return value can be
2987 *              one of the XDP program return codes up to **XDP_TX**, as chosen
2988 *              by the caller. The higher bits of *flags* can be set to
2989 *              BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2990 *
2991 *              With BPF_F_BROADCAST the packet will be broadcasted to all the
2992 *              interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2993 *              interface will be excluded when do broadcasting.
2994 *
2995 *              See also **bpf_redirect**\ (), which only supports redirecting
2996 *              to an ifindex, but doesn't require a map to do so.
2997 *      Return
2998 *              **XDP_REDIRECT** on success, or the value of the two lower bits
2999 *              of the *flags* argument on error.
3000 *
3001 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
3002 *      Description
3003 *              Redirect the packet to the socket referenced by *map* (of type
3004 *              **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3005 *              egress interfaces can be used for redirection. The
3006 *              **BPF_F_INGRESS** value in *flags* is used to make the
3007 *              distinction (ingress path is selected if the flag is present,
3008 *              egress path otherwise). This is the only flag supported for now.
3009 *      Return
3010 *              **SK_PASS** on success, or **SK_DROP** on error.
3011 *
3012 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3013 *      Description
3014 *              Add an entry to, or update a *map* referencing sockets. The
3015 *              *skops* is used as a new value for the entry associated to
3016 *              *key*. *flags* is one of:
3017 *
3018 *              **BPF_NOEXIST**
3019 *                      The entry for *key* must not exist in the map.
3020 *              **BPF_EXIST**
3021 *                      The entry for *key* must already exist in the map.
3022 *              **BPF_ANY**
3023 *                      No condition on the existence of the entry for *key*.
3024 *
3025 *              If the *map* has eBPF programs (parser and verdict), those will
3026 *              be inherited by the socket being added. If the socket is
3027 *              already attached to eBPF programs, this results in an error.
3028 *      Return
3029 *              0 on success, or a negative error in case of failure.
3030 *
3031 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
3032 *      Description
3033 *              Adjust the address pointed by *xdp_md*\ **->data_meta** by
3034 *              *delta* (which can be positive or negative). Note that this
3035 *              operation modifies the address stored in *xdp_md*\ **->data**,
3036 *              so the latter must be loaded only after the helper has been
3037 *              called.
3038 *
3039 *              The use of *xdp_md*\ **->data_meta** is optional and programs
3040 *              are not required to use it. The rationale is that when the
3041 *              packet is processed with XDP (e.g. as DoS filter), it is
3042 *              possible to push further meta data along with it before passing
3043 *              to the stack, and to give the guarantee that an ingress eBPF
3044 *              program attached as a TC classifier on the same device can pick
3045 *              this up for further post-processing. Since TC works with socket
3046 *              buffers, it remains possible to set from XDP the **mark** or
3047 *              **priority** pointers, or other pointers for the socket buffer.
3048 *              Having this scratch space generic and programmable allows for
3049 *              more flexibility as the user is free to store whatever meta
3050 *              data they need.
3051 *
3052 *              A call to this helper is susceptible to change the underlying
3053 *              packet buffer. Therefore, at load time, all checks on pointers
3054 *              previously done by the verifier are invalidated and must be
3055 *              performed again, if the helper is used in combination with
3056 *              direct packet access.
3057 *      Return
3058 *              0 on success, or a negative error in case of failure.
3059 *
3060 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
3061 *      Description
3062 *              Read the value of a perf event counter, and store it into *buf*
3063 *              of size *buf_size*. This helper relies on a *map* of type
3064 *              **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
3065 *              counter is selected when *map* is updated with perf event file
3066 *              descriptors. The *map* is an array whose size is the number of
3067 *              available CPUs, and each cell contains a value relative to one
3068 *              CPU. The value to retrieve is indicated by *flags*, that
3069 *              contains the index of the CPU to look up, masked with
3070 *              **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
3071 *              **BPF_F_CURRENT_CPU** to indicate that the value for the
3072 *              current CPU should be retrieved.
3073 *
3074 *              This helper behaves in a way close to
3075 *              **bpf_perf_event_read**\ () helper, save that instead of
3076 *              just returning the value observed, it fills the *buf*
3077 *              structure. This allows for additional data to be retrieved: in
3078 *              particular, the enabled and running times (in *buf*\
3079 *              **->enabled** and *buf*\ **->running**, respectively) are
3080 *              copied. In general, **bpf_perf_event_read_value**\ () is
3081 *              recommended over **bpf_perf_event_read**\ (), which has some
3082 *              ABI issues and provides fewer functionalities.
3083 *
3084 *              These values are interesting, because hardware PMU (Performance
3085 *              Monitoring Unit) counters are limited resources. When there are
3086 *              more PMU based perf events opened than available counters,
3087 *              kernel will multiplex these events so each event gets certain
3088 *              percentage (but not all) of the PMU time. In case that
3089 *              multiplexing happens, the number of samples or counter value
3090 *              will not reflect the case compared to when no multiplexing
3091 *              occurs. This makes comparison between different runs difficult.
3092 *              Typically, the counter value should be normalized before
3093 *              comparing to other experiments. The usual normalization is done
3094 *              as follows.
3095 *
3096 *              ::
3097 *
3098 *                      normalized_counter = counter * t_enabled / t_running
3099 *
3100 *              Where t_enabled is the time enabled for event and t_running is
3101 *              the time running for event since last normalization. The
3102 *              enabled and running times are accumulated since the perf event
3103 *              open. To achieve scaling factor between two invocations of an
3104 *              eBPF program, users can use CPU id as the key (which is
3105 *              typical for perf array usage model) to remember the previous
3106 *              value and do the calculation inside the eBPF program.
3107 *      Return
3108 *              0 on success, or a negative error in case of failure.
3109 *
3110 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
3111 *      Description
3112 *              For an eBPF program attached to a perf event, retrieve the
3113 *              value of the event counter associated to *ctx* and store it in
3114 *              the structure pointed by *buf* and of size *buf_size*. Enabled
3115 *              and running times are also stored in the structure (see
3116 *              description of helper **bpf_perf_event_read_value**\ () for
3117 *              more details).
3118 *      Return
3119 *              0 on success, or a negative error in case of failure.
3120 *
3121 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
3122 *      Description
3123 *              Emulate a call to **getsockopt()** on the socket associated to
3124 *              *bpf_socket*, which must be a full socket. The *level* at
3125 *              which the option resides and the name *optname* of the option
3126 *              must be specified, see **getsockopt(2)** for more information.
3127 *              The retrieved value is stored in the structure pointed by
3128 *              *opval* and of length *optlen*.
3129 *
3130 *              *bpf_socket* should be one of the following:
3131 *
3132 *              * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
3133 *              * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
3134 *                **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
3135 *
3136 *              This helper actually implements a subset of **getsockopt()**.
3137 *              It supports the same set of *optname*\ s that is supported by
3138 *              the **bpf_setsockopt**\ () helper.  The exceptions are
3139 *              **TCP_BPF_*** is **bpf_setsockopt**\ () only and
3140 *              **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
3141 *      Return
3142 *              0 on success, or a negative error in case of failure.
3143 *
3144 * long bpf_override_return(struct pt_regs *regs, u64 rc)
3145 *      Description
3146 *              Used for error injection, this helper uses kprobes to override
3147 *              the return value of the probed function, and to set it to *rc*.
3148 *              The first argument is the context *regs* on which the kprobe
3149 *              works.
3150 *
3151 *              This helper works by setting the PC (program counter)
3152 *              to an override function which is run in place of the original
3153 *              probed function. This means the probed function is not run at
3154 *              all. The replacement function just returns with the required
3155 *              value.
3156 *
3157 *              This helper has security implications, and thus is subject to
3158 *              restrictions. It is only available if the kernel was compiled
3159 *              with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
3160 *              option, and in this case it only works on functions tagged with
3161 *              **ALLOW_ERROR_INJECTION** in the kernel code.
3162 *      Return
3163 *              0
3164 *
3165 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
3166 *      Description
3167 *              Attempt to set the value of the **bpf_sock_ops_cb_flags** field
3168 *              for the full TCP socket associated to *bpf_sock_ops* to
3169 *              *argval*.
3170 *
3171 *              The primary use of this field is to determine if there should
3172 *              be calls to eBPF programs of type
3173 *              **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
3174 *              code. A program of the same type can change its value, per
3175 *              connection and as necessary, when the connection is
3176 *              established. This field is directly accessible for reading, but
3177 *              this helper must be used for updates in order to return an
3178 *              error if an eBPF program tries to set a callback that is not
3179 *              supported in the current kernel.
3180 *
3181 *              *argval* is a flag array which can combine these flags:
3182 *
3183 *              * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
3184 *              * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
3185 *              * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
3186 *              * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
3187 *
3188 *              Therefore, this function can be used to clear a callback flag by
3189 *              setting the appropriate bit to zero. e.g. to disable the RTO
3190 *              callback:
3191 *
3192 *              **bpf_sock_ops_cb_flags_set(bpf_sock,**
3193 *                      **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
3194 *
3195 *              Here are some examples of where one could call such eBPF
3196 *              program:
3197 *
3198 *              * When RTO fires.
3199 *              * When a packet is retransmitted.
3200 *              * When the connection terminates.
3201 *              * When a packet is sent.
3202 *              * When a packet is received.
3203 *      Return
3204 *              Code **-EINVAL** if the socket is not a full TCP socket;
3205 *              otherwise, a positive number containing the bits that could not
3206 *              be set is returned (which comes down to 0 if all bits were set
3207 *              as required).
3208 *
3209 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
3210 *      Description
3211 *              This helper is used in programs implementing policies at the
3212 *              socket level. If the message *msg* is allowed to pass (i.e. if
3213 *              the verdict eBPF program returns **SK_PASS**), redirect it to
3214 *              the socket referenced by *map* (of type
3215 *              **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3216 *              egress interfaces can be used for redirection. The
3217 *              **BPF_F_INGRESS** value in *flags* is used to make the
3218 *              distinction (ingress path is selected if the flag is present,
3219 *              egress path otherwise). This is the only flag supported for now.
3220 *      Return
3221 *              **SK_PASS** on success, or **SK_DROP** on error.
3222 *
3223 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
3224 *      Description
3225 *              For socket policies, apply the verdict of the eBPF program to
3226 *              the next *bytes* (number of bytes) of message *msg*.
3227 *
3228 *              For example, this helper can be used in the following cases:
3229 *
3230 *              * A single **sendmsg**\ () or **sendfile**\ () system call
3231 *                contains multiple logical messages that the eBPF program is
3232 *                supposed to read and for which it should apply a verdict.
3233 *              * An eBPF program only cares to read the first *bytes* of a
3234 *                *msg*. If the message has a large payload, then setting up
3235 *                and calling the eBPF program repeatedly for all bytes, even
3236 *                though the verdict is already known, would create unnecessary
3237 *                overhead.
3238 *
3239 *              When called from within an eBPF program, the helper sets a
3240 *              counter internal to the BPF infrastructure, that is used to
3241 *              apply the last verdict to the next *bytes*. If *bytes* is
3242 *              smaller than the current data being processed from a
3243 *              **sendmsg**\ () or **sendfile**\ () system call, the first
3244 *              *bytes* will be sent and the eBPF program will be re-run with
3245 *              the pointer for start of data pointing to byte number *bytes*
3246 *              **+ 1**. If *bytes* is larger than the current data being
3247 *              processed, then the eBPF verdict will be applied to multiple
3248 *              **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
3249 *              consumed.
3250 *
3251 *              Note that if a socket closes with the internal counter holding
3252 *              a non-zero value, this is not a problem because data is not
3253 *              being buffered for *bytes* and is sent as it is received.
3254 *      Return
3255 *              0
3256 *
3257 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
3258 *      Description
3259 *              For socket policies, prevent the execution of the verdict eBPF
3260 *              program for message *msg* until *bytes* (byte number) have been
3261 *              accumulated.
3262 *
3263 *              This can be used when one needs a specific number of bytes
3264 *              before a verdict can be assigned, even if the data spans
3265 *              multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
3266 *              case would be a user calling **sendmsg**\ () repeatedly with
3267 *              1-byte long message segments. Obviously, this is bad for
3268 *              performance, but it is still valid. If the eBPF program needs
3269 *              *bytes* bytes to validate a header, this helper can be used to
3270 *              prevent the eBPF program to be called again until *bytes* have
3271 *              been accumulated.
3272 *      Return
3273 *              0
3274 *
3275 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
3276 *      Description
3277 *              For socket policies, pull in non-linear data from user space
3278 *              for *msg* and set pointers *msg*\ **->data** and *msg*\
3279 *              **->data_end** to *start* and *end* bytes offsets into *msg*,
3280 *              respectively.
3281 *
3282 *              If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3283 *              *msg* it can only parse data that the (**data**, **data_end**)
3284 *              pointers have already consumed. For **sendmsg**\ () hooks this
3285 *              is likely the first scatterlist element. But for calls relying
3286 *              on the **sendpage** handler (e.g. **sendfile**\ ()) this will
3287 *              be the range (**0**, **0**) because the data is shared with
3288 *              user space and by default the objective is to avoid allowing
3289 *              user space to modify data while (or after) eBPF verdict is
3290 *              being decided. This helper can be used to pull in data and to
3291 *              set the start and end pointer to given values. Data will be
3292 *              copied if necessary (i.e. if data was not linear and if start
3293 *              and end pointers do not point to the same chunk).
3294 *
3295 *              A call to this helper is susceptible to change the underlying
3296 *              packet buffer. Therefore, at load time, all checks on pointers
3297 *              previously done by the verifier are invalidated and must be
3298 *              performed again, if the helper is used in combination with
3299 *              direct packet access.
3300 *
3301 *              All values for *flags* are reserved for future usage, and must
3302 *              be left at zero.
3303 *      Return
3304 *              0 on success, or a negative error in case of failure.
3305 *
3306 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
3307 *      Description
3308 *              Bind the socket associated to *ctx* to the address pointed by
3309 *              *addr*, of length *addr_len*. This allows for making outgoing
3310 *              connection from the desired IP address, which can be useful for
3311 *              example when all processes inside a cgroup should use one
3312 *              single IP address on a host that has multiple IP configured.
3313 *
3314 *              This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3315 *              domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3316 *              **AF_INET6**). It's advised to pass zero port (**sin_port**
3317 *              or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3318 *              behavior and lets the kernel efficiently pick up an unused
3319 *              port as long as 4-tuple is unique. Passing non-zero port might
3320 *              lead to degraded performance.
3321 *      Return
3322 *              0 on success, or a negative error in case of failure.
3323 *
3324 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3325 *      Description
3326 *              Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3327 *              possible to both shrink and grow the packet tail.
3328 *              Shrink done via *delta* being a negative integer.
3329 *
3330 *              A call to this helper is susceptible to change the underlying
3331 *              packet buffer. Therefore, at load time, all checks on pointers
3332 *              previously done by the verifier are invalidated and must be
3333 *              performed again, if the helper is used in combination with
3334 *              direct packet access.
3335 *      Return
3336 *              0 on success, or a negative error in case of failure.
3337 *
3338 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3339 *      Description
3340 *              Retrieve the XFRM state (IP transform framework, see also
3341 *              **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3342 *
3343 *              The retrieved value is stored in the **struct bpf_xfrm_state**
3344 *              pointed by *xfrm_state* and of length *size*.
3345 *
3346 *              All values for *flags* are reserved for future usage, and must
3347 *              be left at zero.
3348 *
3349 *              This helper is available only if the kernel was compiled with
3350 *              **CONFIG_XFRM** configuration option.
3351 *      Return
3352 *              0 on success, or a negative error in case of failure.
3353 *
3354 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3355 *      Description
3356 *              Return a user or a kernel stack in bpf program provided buffer.
3357 *              To achieve this, the helper needs *ctx*, which is a pointer
3358 *              to the context on which the tracing program is executed.
3359 *              To store the stacktrace, the bpf program provides *buf* with
3360 *              a nonnegative *size*.
3361 *
3362 *              The last argument, *flags*, holds the number of stack frames to
3363 *              skip (from 0 to 255), masked with
3364 *              **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3365 *              the following flags:
3366 *
3367 *              **BPF_F_USER_STACK**
3368 *                      Collect a user space stack instead of a kernel stack.
3369 *              **BPF_F_USER_BUILD_ID**
3370 *                      Collect (build_id, file_offset) instead of ips for user
3371 *                      stack, only valid if **BPF_F_USER_STACK** is also
3372 *                      specified.
3373 *
3374 *                      *file_offset* is an offset relative to the beginning
3375 *                      of the executable or shared object file backing the vma
3376 *                      which the *ip* falls in. It is *not* an offset relative
3377 *                      to that object's base address. Accordingly, it must be
3378 *                      adjusted by adding (sh_addr - sh_offset), where
3379 *                      sh_{addr,offset} correspond to the executable section
3380 *                      containing *file_offset* in the object, for comparisons
3381 *                      to symbols' st_value to be valid.
3382 *
3383 *              **bpf_get_stack**\ () can collect up to
3384 *              **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3385 *              to sufficient large buffer size. Note that
3386 *              this limit can be controlled with the **sysctl** program, and
3387 *              that it should be manually increased in order to profile long
3388 *              user stacks (such as stacks for Java programs). To do so, use:
3389 *
3390 *              ::
3391 *
3392 *                      # sysctl kernel.perf_event_max_stack=<new value>
3393 *      Return
3394 *              The non-negative copied *buf* length equal to or less than
3395 *              *size* on success, or a negative error in case of failure.
3396 *
3397 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3398 *      Description
3399 *              This helper is similar to **bpf_skb_load_bytes**\ () in that
3400 *              it provides an easy way to load *len* bytes from *offset*
3401 *              from the packet associated to *skb*, into the buffer pointed
3402 *              by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3403 *              a fifth argument *start_header* exists in order to select a
3404 *              base offset to start from. *start_header* can be one of:
3405 *
3406 *              **BPF_HDR_START_MAC**
3407 *                      Base offset to load data from is *skb*'s mac header.
3408 *              **BPF_HDR_START_NET**
3409 *                      Base offset to load data from is *skb*'s network header.
3410 *
3411 *              In general, "direct packet access" is the preferred method to
3412 *              access packet data, however, this helper is in particular useful
3413 *              in socket filters where *skb*\ **->data** does not always point
3414 *              to the start of the mac header and where "direct packet access"
3415 *              is not available.
3416 *      Return
3417 *              0 on success, or a negative error in case of failure.
3418 *
3419 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3420 *      Description
3421 *              Do FIB lookup in kernel tables using parameters in *params*.
3422 *              If lookup is successful and result shows packet is to be
3423 *              forwarded, the neighbor tables are searched for the nexthop.
3424 *              If successful (ie., FIB lookup shows forwarding and nexthop
3425 *              is resolved), the nexthop address is returned in ipv4_dst
3426 *              or ipv6_dst based on family, smac is set to mac address of
3427 *              egress device, dmac is set to nexthop mac address, rt_metric
3428 *              is set to metric from route (IPv4/IPv6 only), and ifindex
3429 *              is set to the device index of the nexthop from the FIB lookup.
3430 *
3431 *              *plen* argument is the size of the passed in struct.
3432 *              *flags* argument can be a combination of one or more of the
3433 *              following values:
3434 *
3435 *              **BPF_FIB_LOOKUP_DIRECT**
3436 *                      Do a direct table lookup vs full lookup using FIB
3437 *                      rules.
3438 *              **BPF_FIB_LOOKUP_TBID**
3439 *                      Used with BPF_FIB_LOOKUP_DIRECT.
3440 *                      Use the routing table ID present in *params*->tbid
3441 *                      for the fib lookup.
3442 *              **BPF_FIB_LOOKUP_OUTPUT**
3443 *                      Perform lookup from an egress perspective (default is
3444 *                      ingress).
3445 *              **BPF_FIB_LOOKUP_SKIP_NEIGH**
3446 *                      Skip the neighbour table lookup. *params*->dmac
3447 *                      and *params*->smac will not be set as output. A common
3448 *                      use case is to call **bpf_redirect_neigh**\ () after
3449 *                      doing **bpf_fib_lookup**\ ().
3450 *              **BPF_FIB_LOOKUP_SRC**
3451 *                      Derive and set source IP addr in *params*->ipv{4,6}_src
3452 *                      for the nexthop. If the src addr cannot be derived,
3453 *                      **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
3454 *                      case, *params*->dmac and *params*->smac are not set either.
3455 *              **BPF_FIB_LOOKUP_MARK**
3456 *                      Use the mark present in *params*->mark for the fib lookup.
3457 *                      This option should not be used with BPF_FIB_LOOKUP_DIRECT,
3458 *                      as it only has meaning for full lookups.
3459 *
3460 *              *ctx* is either **struct xdp_md** for XDP programs or
3461 *              **struct sk_buff** tc cls_act programs.
3462 *      Return
3463 *              * < 0 if any input argument is invalid
3464 *              *   0 on success (packet is forwarded, nexthop neighbor exists)
3465 *              * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3466 *                packet is not forwarded or needs assist from full stack
3467 *
3468 *              If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3469 *              was exceeded and output params->mtu_result contains the MTU.
3470 *
3471 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3472 *      Description
3473 *              Add an entry to, or update a sockhash *map* referencing sockets.
3474 *              The *skops* is used as a new value for the entry associated to
3475 *              *key*. *flags* is one of:
3476 *
3477 *              **BPF_NOEXIST**
3478 *                      The entry for *key* must not exist in the map.
3479 *              **BPF_EXIST**
3480 *                      The entry for *key* must already exist in the map.
3481 *              **BPF_ANY**
3482 *                      No condition on the existence of the entry for *key*.
3483 *
3484 *              If the *map* has eBPF programs (parser and verdict), those will
3485 *              be inherited by the socket being added. If the socket is
3486 *              already attached to eBPF programs, this results in an error.
3487 *      Return
3488 *              0 on success, or a negative error in case of failure.
3489 *
3490 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3491 *      Description
3492 *              This helper is used in programs implementing policies at the
3493 *              socket level. If the message *msg* is allowed to pass (i.e. if
3494 *              the verdict eBPF program returns **SK_PASS**), redirect it to
3495 *              the socket referenced by *map* (of type
3496 *              **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3497 *              egress interfaces can be used for redirection. The
3498 *              **BPF_F_INGRESS** value in *flags* is used to make the
3499 *              distinction (ingress path is selected if the flag is present,
3500 *              egress path otherwise). This is the only flag supported for now.
3501 *      Return
3502 *              **SK_PASS** on success, or **SK_DROP** on error.
3503 *
3504 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3505 *      Description
3506 *              This helper is used in programs implementing policies at the
3507 *              skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3508 *              if the verdict eBPF program returns **SK_PASS**), redirect it
3509 *              to the socket referenced by *map* (of type
3510 *              **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3511 *              egress interfaces can be used for redirection. The
3512 *              **BPF_F_INGRESS** value in *flags* is used to make the
3513 *              distinction (ingress path is selected if the flag is present,
3514 *              egress otherwise). This is the only flag supported for now.
3515 *      Return
3516 *              **SK_PASS** on success, or **SK_DROP** on error.
3517 *
3518 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3519 *      Description
3520 *              Encapsulate the packet associated to *skb* within a Layer 3
3521 *              protocol header. This header is provided in the buffer at
3522 *              address *hdr*, with *len* its size in bytes. *type* indicates
3523 *              the protocol of the header and can be one of:
3524 *
3525 *              **BPF_LWT_ENCAP_SEG6**
3526 *                      IPv6 encapsulation with Segment Routing Header
3527 *                      (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3528 *                      the IPv6 header is computed by the kernel.
3529 *              **BPF_LWT_ENCAP_SEG6_INLINE**
3530 *                      Only works if *skb* contains an IPv6 packet. Insert a
3531 *                      Segment Routing Header (**struct ipv6_sr_hdr**) inside
3532 *                      the IPv6 header.
3533 *              **BPF_LWT_ENCAP_IP**
3534 *                      IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3535 *                      must be IPv4 or IPv6, followed by zero or more
3536 *                      additional headers, up to **LWT_BPF_MAX_HEADROOM**
3537 *                      total bytes in all prepended headers. Please note that
3538 *                      if **skb_is_gso**\ (*skb*) is true, no more than two
3539 *                      headers can be prepended, and the inner header, if
3540 *                      present, should be either GRE or UDP/GUE.
3541 *
3542 *              **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3543 *              of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3544 *              be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3545 *              **BPF_PROG_TYPE_LWT_XMIT**.
3546 *
3547 *              A call to this helper is susceptible to change the underlying
3548 *              packet buffer. Therefore, at load time, all checks on pointers
3549 *              previously done by the verifier are invalidated and must be
3550 *              performed again, if the helper is used in combination with
3551 *              direct packet access.
3552 *      Return
3553 *              0 on success, or a negative error in case of failure.
3554 *
3555 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3556 *      Description
3557 *              Store *len* bytes from address *from* into the packet
3558 *              associated to *skb*, at *offset*. Only the flags, tag and TLVs
3559 *              inside the outermost IPv6 Segment Routing Header can be
3560 *              modified through this helper.
3561 *
3562 *              A call to this helper is susceptible to change the underlying
3563 *              packet buffer. Therefore, at load time, all checks on pointers
3564 *              previously done by the verifier are invalidated and must be
3565 *              performed again, if the helper is used in combination with
3566 *              direct packet access.
3567 *      Return
3568 *              0 on success, or a negative error in case of failure.
3569 *
3570 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3571 *      Description
3572 *              Adjust the size allocated to TLVs in the outermost IPv6
3573 *              Segment Routing Header contained in the packet associated to
3574 *              *skb*, at position *offset* by *delta* bytes. Only offsets
3575 *              after the segments are accepted. *delta* can be as well
3576 *              positive (growing) as negative (shrinking).
3577 *
3578 *              A call to this helper is susceptible to change the underlying
3579 *              packet buffer. Therefore, at load time, all checks on pointers
3580 *              previously done by the verifier are invalidated and must be
3581 *              performed again, if the helper is used in combination with
3582 *              direct packet access.
3583 *      Return
3584 *              0 on success, or a negative error in case of failure.
3585 *
3586 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3587 *      Description
3588 *              Apply an IPv6 Segment Routing action of type *action* to the
3589 *              packet associated to *skb*. Each action takes a parameter
3590 *              contained at address *param*, and of length *param_len* bytes.
3591 *              *action* can be one of:
3592 *
3593 *              **SEG6_LOCAL_ACTION_END_X**
3594 *                      End.X action: Endpoint with Layer-3 cross-connect.
3595 *                      Type of *param*: **struct in6_addr**.
3596 *              **SEG6_LOCAL_ACTION_END_T**
3597 *                      End.T action: Endpoint with specific IPv6 table lookup.
3598 *                      Type of *param*: **int**.
3599 *              **SEG6_LOCAL_ACTION_END_B6**
3600 *                      End.B6 action: Endpoint bound to an SRv6 policy.
3601 *                      Type of *param*: **struct ipv6_sr_hdr**.
3602 *              **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3603 *                      End.B6.Encap action: Endpoint bound to an SRv6
3604 *                      encapsulation policy.
3605 *                      Type of *param*: **struct ipv6_sr_hdr**.
3606 *
3607 *              A call to this helper is susceptible to change the underlying
3608 *              packet buffer. Therefore, at load time, all checks on pointers
3609 *              previously done by the verifier are invalidated and must be
3610 *              performed again, if the helper is used in combination with
3611 *              direct packet access.
3612 *      Return
3613 *              0 on success, or a negative error in case of failure.
3614 *
3615 * long bpf_rc_repeat(void *ctx)
3616 *      Description
3617 *              This helper is used in programs implementing IR decoding, to
3618 *              report a successfully decoded repeat key message. This delays
3619 *              the generation of a key up event for previously generated
3620 *              key down event.
3621 *
3622 *              Some IR protocols like NEC have a special IR message for
3623 *              repeating last button, for when a button is held down.
3624 *
3625 *              The *ctx* should point to the lirc sample as passed into
3626 *              the program.
3627 *
3628 *              This helper is only available is the kernel was compiled with
3629 *              the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3630 *              "**y**".
3631 *      Return
3632 *              0
3633 *
3634 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3635 *      Description
3636 *              This helper is used in programs implementing IR decoding, to
3637 *              report a successfully decoded key press with *scancode*,
3638 *              *toggle* value in the given *protocol*. The scancode will be
3639 *              translated to a keycode using the rc keymap, and reported as
3640 *              an input key down event. After a period a key up event is
3641 *              generated. This period can be extended by calling either
3642 *              **bpf_rc_keydown**\ () again with the same values, or calling
3643 *              **bpf_rc_repeat**\ ().
3644 *
3645 *              Some protocols include a toggle bit, in case the button was
3646 *              released and pressed again between consecutive scancodes.
3647 *
3648 *              The *ctx* should point to the lirc sample as passed into
3649 *              the program.
3650 *
3651 *              The *protocol* is the decoded protocol number (see
3652 *              **enum rc_proto** for some predefined values).
3653 *
3654 *              This helper is only available is the kernel was compiled with
3655 *              the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3656 *              "**y**".
3657 *      Return
3658 *              0
3659 *
3660 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3661 *      Description
3662 *              Return the cgroup v2 id of the socket associated with the *skb*.
3663 *              This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3664 *              helper for cgroup v1 by providing a tag resp. identifier that
3665 *              can be matched on or used for map lookups e.g. to implement
3666 *              policy. The cgroup v2 id of a given path in the hierarchy is
3667 *              exposed in user space through the f_handle API in order to get
3668 *              to the same 64-bit id.
3669 *
3670 *              This helper can be used on TC egress path, but not on ingress,
3671 *              and is available only if the kernel was compiled with the
3672 *              **CONFIG_SOCK_CGROUP_DATA** configuration option.
3673 *      Return
3674 *              The id is returned or 0 in case the id could not be retrieved.
3675 *
3676 * u64 bpf_get_current_cgroup_id(void)
3677 *      Description
3678 *              Get the current cgroup id based on the cgroup within which
3679 *              the current task is running.
3680 *      Return
3681 *              A 64-bit integer containing the current cgroup id based
3682 *              on the cgroup within which the current task is running.
3683 *
3684 * void *bpf_get_local_storage(void *map, u64 flags)
3685 *      Description
3686 *              Get the pointer to the local storage area.
3687 *              The type and the size of the local storage is defined
3688 *              by the *map* argument.
3689 *              The *flags* meaning is specific for each map type,
3690 *              and has to be 0 for cgroup local storage.
3691 *
3692 *              Depending on the BPF program type, a local storage area
3693 *              can be shared between multiple instances of the BPF program,
3694 *              running simultaneously.
3695 *
3696 *              A user should care about the synchronization by himself.
3697 *              For example, by using the **BPF_ATOMIC** instructions to alter
3698 *              the shared data.
3699 *      Return
3700 *              A pointer to the local storage area.
3701 *
3702 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3703 *      Description
3704 *              Select a **SO_REUSEPORT** socket from a
3705 *              **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3706 *              It checks the selected socket is matching the incoming
3707 *              request in the socket buffer.
3708 *      Return
3709 *              0 on success, or a negative error in case of failure.
3710 *
3711 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3712 *      Description
3713 *              Return id of cgroup v2 that is ancestor of cgroup associated
3714 *              with the *skb* at the *ancestor_level*.  The root cgroup is at
3715 *              *ancestor_level* zero and each step down the hierarchy
3716 *              increments the level. If *ancestor_level* == level of cgroup
3717 *              associated with *skb*, then return value will be same as that
3718 *              of **bpf_skb_cgroup_id**\ ().
3719 *
3720 *              The helper is useful to implement policies based on cgroups
3721 *              that are upper in hierarchy than immediate cgroup associated
3722 *              with *skb*.
3723 *
3724 *              The format of returned id and helper limitations are same as in
3725 *              **bpf_skb_cgroup_id**\ ().
3726 *      Return
3727 *              The id is returned or 0 in case the id could not be retrieved.
3728 *
3729 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3730 *      Description
3731 *              Look for TCP socket matching *tuple*, optionally in a child
3732 *              network namespace *netns*. The return value must be checked,
3733 *              and if non-**NULL**, released via **bpf_sk_release**\ ().
3734 *
3735 *              The *ctx* should point to the context of the program, such as
3736 *              the skb or socket (depending on the hook in use). This is used
3737 *              to determine the base network namespace for the lookup.
3738 *
3739 *              *tuple_size* must be one of:
3740 *
3741 *              **sizeof**\ (*tuple*\ **->ipv4**)
3742 *                      Look for an IPv4 socket.
3743 *              **sizeof**\ (*tuple*\ **->ipv6**)
3744 *                      Look for an IPv6 socket.
3745 *
3746 *              If the *netns* is a negative signed 32-bit integer, then the
3747 *              socket lookup table in the netns associated with the *ctx*
3748 *              will be used. For the TC hooks, this is the netns of the device
3749 *              in the skb. For socket hooks, this is the netns of the socket.
3750 *              If *netns* is any other signed 32-bit value greater than or
3751 *              equal to zero then it specifies the ID of the netns relative to
3752 *              the netns associated with the *ctx*. *netns* values beyond the
3753 *              range of 32-bit integers are reserved for future use.
3754 *
3755 *              All values for *flags* are reserved for future usage, and must
3756 *              be left at zero.
3757 *
3758 *              This helper is available only if the kernel was compiled with
3759 *              **CONFIG_NET** configuration option.
3760 *      Return
3761 *              Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3762 *              For sockets with reuseport option, the **struct bpf_sock**
3763 *              result is from *reuse*\ **->socks**\ [] using the hash of the
3764 *              tuple.
3765 *
3766 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3767 *      Description
3768 *              Look for UDP socket matching *tuple*, optionally in a child
3769 *              network namespace *netns*. The return value must be checked,
3770 *              and if non-**NULL**, released via **bpf_sk_release**\ ().
3771 *
3772 *              The *ctx* should point to the context of the program, such as
3773 *              the skb or socket (depending on the hook in use). This is used
3774 *              to determine the base network namespace for the lookup.
3775 *
3776 *              *tuple_size* must be one of:
3777 *
3778 *              **sizeof**\ (*tuple*\ **->ipv4**)
3779 *                      Look for an IPv4 socket.
3780 *              **sizeof**\ (*tuple*\ **->ipv6**)
3781 *                      Look for an IPv6 socket.
3782 *
3783 *              If the *netns* is a negative signed 32-bit integer, then the
3784 *              socket lookup table in the netns associated with the *ctx*
3785 *              will be used. For the TC hooks, this is the netns of the device
3786 *              in the skb. For socket hooks, this is the netns of the socket.
3787 *              If *netns* is any other signed 32-bit value greater than or
3788 *              equal to zero then it specifies the ID of the netns relative to
3789 *              the netns associated with the *ctx*. *netns* values beyond the
3790 *              range of 32-bit integers are reserved for future use.
3791 *
3792 *              All values for *flags* are reserved for future usage, and must
3793 *              be left at zero.
3794 *
3795 *              This helper is available only if the kernel was compiled with
3796 *              **CONFIG_NET** configuration option.
3797 *      Return
3798 *              Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3799 *              For sockets with reuseport option, the **struct bpf_sock**
3800 *              result is from *reuse*\ **->socks**\ [] using the hash of the
3801 *              tuple.
3802 *
3803 * long bpf_sk_release(void *sock)
3804 *      Description
3805 *              Release the reference held by *sock*. *sock* must be a
3806 *              non-**NULL** pointer that was returned from
3807 *              **bpf_sk_lookup_xxx**\ ().
3808 *      Return
3809 *              0 on success, or a negative error in case of failure.
3810 *
3811 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3812 *      Description
3813 *              Push an element *value* in *map*. *flags* is one of:
3814 *
3815 *              **BPF_EXIST**
3816 *                      If the queue/stack is full, the oldest element is
3817 *                      removed to make room for this.
3818 *      Return
3819 *              0 on success, or a negative error in case of failure.
3820 *
3821 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3822 *      Description
3823 *              Pop an element from *map*.
3824 *      Return
3825 *              0 on success, or a negative error in case of failure.
3826 *
3827 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3828 *      Description
3829 *              Get an element from *map* without removing it.
3830 *      Return
3831 *              0 on success, or a negative error in case of failure.
3832 *
3833 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3834 *      Description
3835 *              For socket policies, insert *len* bytes into *msg* at offset
3836 *              *start*.
3837 *
3838 *              If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3839 *              *msg* it may want to insert metadata or options into the *msg*.
3840 *              This can later be read and used by any of the lower layer BPF
3841 *              hooks.
3842 *
3843 *              This helper may fail if under memory pressure (a malloc
3844 *              fails) in these cases BPF programs will get an appropriate
3845 *              error and BPF programs will need to handle them.
3846 *      Return
3847 *              0 on success, or a negative error in case of failure.
3848 *
3849 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3850 *      Description
3851 *              Will remove *len* bytes from a *msg* starting at byte *start*.
3852 *              This may result in **ENOMEM** errors under certain situations if
3853 *              an allocation and copy are required due to a full ring buffer.
3854 *              However, the helper will try to avoid doing the allocation
3855 *              if possible. Other errors can occur if input parameters are
3856 *              invalid either due to *start* byte not being valid part of *msg*
3857 *              payload and/or *pop* value being to large.
3858 *      Return
3859 *              0 on success, or a negative error in case of failure.
3860 *
3861 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3862 *      Description
3863 *              This helper is used in programs implementing IR decoding, to
3864 *              report a successfully decoded pointer movement.
3865 *
3866 *              The *ctx* should point to the lirc sample as passed into
3867 *              the program.
3868 *
3869 *              This helper is only available is the kernel was compiled with
3870 *              the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3871 *              "**y**".
3872 *      Return
3873 *              0
3874 *
3875 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3876 *      Description
3877 *              Acquire a spinlock represented by the pointer *lock*, which is
3878 *              stored as part of a value of a map. Taking the lock allows to
3879 *              safely update the rest of the fields in that value. The
3880 *              spinlock can (and must) later be released with a call to
3881 *              **bpf_spin_unlock**\ (\ *lock*\ ).
3882 *
3883 *              Spinlocks in BPF programs come with a number of restrictions
3884 *              and constraints:
3885 *
3886 *              * **bpf_spin_lock** objects are only allowed inside maps of
3887 *                types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3888 *                list could be extended in the future).
3889 *              * BTF description of the map is mandatory.
3890 *              * The BPF program can take ONE lock at a time, since taking two
3891 *                or more could cause dead locks.
3892 *              * Only one **struct bpf_spin_lock** is allowed per map element.
3893 *              * When the lock is taken, calls (either BPF to BPF or helpers)
3894 *                are not allowed.
3895 *              * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3896 *                allowed inside a spinlock-ed region.
3897 *              * The BPF program MUST call **bpf_spin_unlock**\ () to release
3898 *                the lock, on all execution paths, before it returns.
3899 *              * The BPF program can access **struct bpf_spin_lock** only via
3900 *                the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3901 *                helpers. Loading or storing data into the **struct
3902 *                bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3903 *              * To use the **bpf_spin_lock**\ () helper, the BTF description
3904 *                of the map value must be a struct and have **struct
3905 *                bpf_spin_lock** *anyname*\ **;** field at the top level.
3906 *                Nested lock inside another struct is not allowed.
3907 *              * The **struct bpf_spin_lock** *lock* field in a map value must
3908 *                be aligned on a multiple of 4 bytes in that value.
3909 *              * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3910 *                the **bpf_spin_lock** field to user space.
3911 *              * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3912 *                a BPF program, do not update the **bpf_spin_lock** field.
3913 *              * **bpf_spin_lock** cannot be on the stack or inside a
3914 *                networking packet (it can only be inside of a map values).
3915 *              * **bpf_spin_lock** is available to root only.
3916 *              * Tracing programs and socket filter programs cannot use
3917 *                **bpf_spin_lock**\ () due to insufficient preemption checks
3918 *                (but this may change in the future).
3919 *              * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3920 *      Return
3921 *              0
3922 *
3923 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3924 *      Description
3925 *              Release the *lock* previously locked by a call to
3926 *              **bpf_spin_lock**\ (\ *lock*\ ).
3927 *      Return
3928 *              0
3929 *
3930 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3931 *      Description
3932 *              This helper gets a **struct bpf_sock** pointer such
3933 *              that all the fields in this **bpf_sock** can be accessed.
3934 *      Return
3935 *              A **struct bpf_sock** pointer on success, or **NULL** in
3936 *              case of failure.
3937 *
3938 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3939 *      Description
3940 *              This helper gets a **struct bpf_tcp_sock** pointer from a
3941 *              **struct bpf_sock** pointer.
3942 *      Return
3943 *              A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3944 *              case of failure.
3945 *
3946 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3947 *      Description
3948 *              Set ECN (Explicit Congestion Notification) field of IP header
3949 *              to **CE** (Congestion Encountered) if current value is **ECT**
3950 *              (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3951 *              and IPv4.
3952 *      Return
3953 *              1 if the **CE** flag is set (either by the current helper call
3954 *              or because it was already present), 0 if it is not set.
3955 *
3956 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3957 *      Description
3958 *              Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3959 *              **bpf_sk_release**\ () is unnecessary and not allowed.
3960 *      Return
3961 *              A **struct bpf_sock** pointer on success, or **NULL** in
3962 *              case of failure.
3963 *
3964 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3965 *      Description
3966 *              Look for TCP socket matching *tuple*, optionally in a child
3967 *              network namespace *netns*. The return value must be checked,
3968 *              and if non-**NULL**, released via **bpf_sk_release**\ ().
3969 *
3970 *              This function is identical to **bpf_sk_lookup_tcp**\ (), except
3971 *              that it also returns timewait or request sockets. Use
3972 *              **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3973 *              full structure.
3974 *
3975 *              This helper is available only if the kernel was compiled with
3976 *              **CONFIG_NET** configuration option.
3977 *      Return
3978 *              Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3979 *              For sockets with reuseport option, the **struct bpf_sock**
3980 *              result is from *reuse*\ **->socks**\ [] using the hash of the
3981 *              tuple.
3982 *
3983 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3984 *      Description
3985 *              Check whether *iph* and *th* contain a valid SYN cookie ACK for
3986 *              the listening socket in *sk*.
3987 *
3988 *              *iph* points to the start of the IPv4 or IPv6 header, while
3989 *              *iph_len* contains **sizeof**\ (**struct iphdr**) or
3990 *              **sizeof**\ (**struct ipv6hdr**).
3991 *
3992 *              *th* points to the start of the TCP header, while *th_len*
3993 *              contains the length of the TCP header (at least
3994 *              **sizeof**\ (**struct tcphdr**)).
3995 *      Return
3996 *              0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3997 *              error otherwise.
3998 *
3999 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
4000 *      Description
4001 *              Get name of sysctl in /proc/sys/ and copy it into provided by
4002 *              program buffer *buf* of size *buf_len*.
4003 *
4004 *              The buffer is always NUL terminated, unless it's zero-sized.
4005 *
4006 *              If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
4007 *              copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
4008 *              only (e.g. "tcp_mem").
4009 *      Return
4010 *              Number of character copied (not including the trailing NUL).
4011 *
4012 *              **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4013 *              truncated name in this case).
4014 *
4015 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4016 *      Description
4017 *              Get current value of sysctl as it is presented in /proc/sys
4018 *              (incl. newline, etc), and copy it as a string into provided
4019 *              by program buffer *buf* of size *buf_len*.
4020 *
4021 *              The whole value is copied, no matter what file position user
4022 *              space issued e.g. sys_read at.
4023 *
4024 *              The buffer is always NUL terminated, unless it's zero-sized.
4025 *      Return
4026 *              Number of character copied (not including the trailing NUL).
4027 *
4028 *              **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4029 *              truncated name in this case).
4030 *
4031 *              **-EINVAL** if current value was unavailable, e.g. because
4032 *              sysctl is uninitialized and read returns -EIO for it.
4033 *
4034 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4035 *      Description
4036 *              Get new value being written by user space to sysctl (before
4037 *              the actual write happens) and copy it as a string into
4038 *              provided by program buffer *buf* of size *buf_len*.
4039 *
4040 *              User space may write new value at file position > 0.
4041 *
4042 *              The buffer is always NUL terminated, unless it's zero-sized.
4043 *      Return
4044 *              Number of character copied (not including the trailing NUL).
4045 *
4046 *              **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4047 *              truncated name in this case).
4048 *
4049 *              **-EINVAL** if sysctl is being read.
4050 *
4051 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
4052 *      Description
4053 *              Override new value being written by user space to sysctl with
4054 *              value provided by program in buffer *buf* of size *buf_len*.
4055 *
4056 *              *buf* should contain a string in same form as provided by user
4057 *              space on sysctl write.
4058 *
4059 *              User space may write new value at file position > 0. To override
4060 *              the whole sysctl value file position should be set to zero.
4061 *      Return
4062 *              0 on success.
4063 *
4064 *              **-E2BIG** if the *buf_len* is too big.
4065 *
4066 *              **-EINVAL** if sysctl is being read.
4067 *
4068 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
4069 *      Description
4070 *              Convert the initial part of the string from buffer *buf* of
4071 *              size *buf_len* to a long integer according to the given base
4072 *              and save the result in *res*.
4073 *
4074 *              The string may begin with an arbitrary amount of white space
4075 *              (as determined by **isspace**\ (3)) followed by a single
4076 *              optional '**-**' sign.
4077 *
4078 *              Five least significant bits of *flags* encode base, other bits
4079 *              are currently unused.
4080 *
4081 *              Base must be either 8, 10, 16 or 0 to detect it automatically
4082 *              similar to user space **strtol**\ (3).
4083 *      Return
4084 *              Number of characters consumed on success. Must be positive but
4085 *              no more than *buf_len*.
4086 *
4087 *              **-EINVAL** if no valid digits were found or unsupported base
4088 *              was provided.
4089 *
4090 *              **-ERANGE** if resulting value was out of range.
4091 *
4092 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
4093 *      Description
4094 *              Convert the initial part of the string from buffer *buf* of
4095 *              size *buf_len* to an unsigned long integer according to the
4096 *              given base and save the result in *res*.
4097 *
4098 *              The string may begin with an arbitrary amount of white space
4099 *              (as determined by **isspace**\ (3)).
4100 *
4101 *              Five least significant bits of *flags* encode base, other bits
4102 *              are currently unused.
4103 *
4104 *              Base must be either 8, 10, 16 or 0 to detect it automatically
4105 *              similar to user space **strtoul**\ (3).
4106 *      Return
4107 *              Number of characters consumed on success. Must be positive but
4108 *              no more than *buf_len*.
4109 *
4110 *              **-EINVAL** if no valid digits were found or unsupported base
4111 *              was provided.
4112 *
4113 *              **-ERANGE** if resulting value was out of range.
4114 *
4115 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
4116 *      Description
4117 *              Get a bpf-local-storage from a *sk*.
4118 *
4119 *              Logically, it could be thought of getting the value from
4120 *              a *map* with *sk* as the **key**.  From this
4121 *              perspective,  the usage is not much different from
4122 *              **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
4123 *              helper enforces the key must be a full socket and the map must
4124 *              be a **BPF_MAP_TYPE_SK_STORAGE** also.
4125 *
4126 *              Underneath, the value is stored locally at *sk* instead of
4127 *              the *map*.  The *map* is used as the bpf-local-storage
4128 *              "type". The bpf-local-storage "type" (i.e. the *map*) is
4129 *              searched against all bpf-local-storages residing at *sk*.
4130 *
4131 *              *sk* is a kernel **struct sock** pointer for LSM program.
4132 *              *sk* is a **struct bpf_sock** pointer for other program types.
4133 *
4134 *              An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
4135 *              used such that a new bpf-local-storage will be
4136 *              created if one does not exist.  *value* can be used
4137 *              together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
4138 *              the initial value of a bpf-local-storage.  If *value* is
4139 *              **NULL**, the new bpf-local-storage will be zero initialized.
4140 *      Return
4141 *              A bpf-local-storage pointer is returned on success.
4142 *
4143 *              **NULL** if not found or there was an error in adding
4144 *              a new bpf-local-storage.
4145 *
4146 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
4147 *      Description
4148 *              Delete a bpf-local-storage from a *sk*.
4149 *      Return
4150 *              0 on success.
4151 *
4152 *              **-ENOENT** if the bpf-local-storage cannot be found.
4153 *              **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
4154 *
4155 * long bpf_send_signal(u32 sig)
4156 *      Description
4157 *              Send signal *sig* to the process of the current task.
4158 *              The signal may be delivered to any of this process's threads.
4159 *      Return
4160 *              0 on success or successfully queued.
4161 *
4162 *              **-EBUSY** if work queue under nmi is full.
4163 *
4164 *              **-EINVAL** if *sig* is invalid.
4165 *
4166 *              **-EPERM** if no permission to send the *sig*.
4167 *
4168 *              **-EAGAIN** if bpf program can try again.
4169 *
4170 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
4171 *      Description
4172 *              Try to issue a SYN cookie for the packet with corresponding
4173 *              IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
4174 *
4175 *              *iph* points to the start of the IPv4 or IPv6 header, while
4176 *              *iph_len* contains **sizeof**\ (**struct iphdr**) or
4177 *              **sizeof**\ (**struct ipv6hdr**).
4178 *
4179 *              *th* points to the start of the TCP header, while *th_len*
4180 *              contains the length of the TCP header with options (at least
4181 *              **sizeof**\ (**struct tcphdr**)).
4182 *      Return
4183 *              On success, lower 32 bits hold the generated SYN cookie in
4184 *              followed by 16 bits which hold the MSS value for that cookie,
4185 *              and the top 16 bits are unused.
4186 *
4187 *              On failure, the returned value is one of the following:
4188 *
4189 *              **-EINVAL** SYN cookie cannot be issued due to error
4190 *
4191 *              **-ENOENT** SYN cookie should not be issued (no SYN flood)
4192 *
4193 *              **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
4194 *
4195 *              **-EPROTONOSUPPORT** IP packet version is not 4 or 6
4196 *
4197 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4198 *      Description
4199 *              Write raw *data* blob into a special BPF perf event held by
4200 *              *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4201 *              event must have the following attributes: **PERF_SAMPLE_RAW**
4202 *              as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4203 *              **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4204 *
4205 *              The *flags* are used to indicate the index in *map* for which
4206 *              the value must be put, masked with **BPF_F_INDEX_MASK**.
4207 *              Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4208 *              to indicate that the index of the current CPU core should be
4209 *              used.
4210 *
4211 *              The value to write, of *size*, is passed through eBPF stack and
4212 *              pointed by *data*.
4213 *
4214 *              *ctx* is a pointer to in-kernel struct sk_buff.
4215 *
4216 *              This helper is similar to **bpf_perf_event_output**\ () but
4217 *              restricted to raw_tracepoint bpf programs.
4218 *      Return
4219 *              0 on success, or a negative error in case of failure.
4220 *
4221 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
4222 *      Description
4223 *              Safely attempt to read *size* bytes from user space address
4224 *              *unsafe_ptr* and store the data in *dst*.
4225 *      Return
4226 *              0 on success, or a negative error in case of failure.
4227 *
4228 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
4229 *      Description
4230 *              Safely attempt to read *size* bytes from kernel space address
4231 *              *unsafe_ptr* and store the data in *dst*.
4232 *      Return
4233 *              0 on success, or a negative error in case of failure.
4234 *
4235 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
4236 *      Description
4237 *              Copy a NUL terminated string from an unsafe user address
4238 *              *unsafe_ptr* to *dst*. The *size* should include the
4239 *              terminating NUL byte. In case the string length is smaller than
4240 *              *size*, the target is not padded with further NUL bytes. If the
4241 *              string length is larger than *size*, just *size*-1 bytes are
4242 *              copied and the last byte is set to NUL.
4243 *
4244 *              On success, returns the number of bytes that were written,
4245 *              including the terminal NUL. This makes this helper useful in
4246 *              tracing programs for reading strings, and more importantly to
4247 *              get its length at runtime. See the following snippet:
4248 *
4249 *              ::
4250 *
4251 *                      SEC("kprobe/sys_open")
4252 *                      void bpf_sys_open(struct pt_regs *ctx)
4253 *                      {
4254 *                              char buf[PATHLEN]; // PATHLEN is defined to 256
4255 *                              int res = bpf_probe_read_user_str(buf, sizeof(buf),
4256 *                                                                ctx->di);
4257 *
4258 *                              // Consume buf, for example push it to
4259 *                              // userspace via bpf_perf_event_output(); we
4260 *                              // can use res (the string length) as event
4261 *                              // size, after checking its boundaries.
4262 *                      }
4263 *
4264 *              In comparison, using **bpf_probe_read_user**\ () helper here
4265 *              instead to read the string would require to estimate the length
4266 *              at compile time, and would often result in copying more memory
4267 *              than necessary.
4268 *
4269 *              Another useful use case is when parsing individual process
4270 *              arguments or individual environment variables navigating
4271 *              *current*\ **->mm->arg_start** and *current*\
4272 *              **->mm->env_start**: using this helper and the return value,
4273 *              one can quickly iterate at the right offset of the memory area.
4274 *      Return
4275 *              On success, the strictly positive length of the output string,
4276 *              including the trailing NUL character. On error, a negative
4277 *              value.
4278 *
4279 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
4280 *      Description
4281 *              Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
4282 *              to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
4283 *      Return
4284 *              On success, the strictly positive length of the string, including
4285 *              the trailing NUL character. On error, a negative value.
4286 *
4287 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
4288 *      Description
4289 *              Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
4290 *              *rcv_nxt* is the ack_seq to be sent out.
4291 *      Return
4292 *              0 on success, or a negative error in case of failure.
4293 *
4294 * long bpf_send_signal_thread(u32 sig)
4295 *      Description
4296 *              Send signal *sig* to the thread corresponding to the current task.
4297 *      Return
4298 *              0 on success or successfully queued.
4299 *
4300 *              **-EBUSY** if work queue under nmi is full.
4301 *
4302 *              **-EINVAL** if *sig* is invalid.
4303 *
4304 *              **-EPERM** if no permission to send the *sig*.
4305 *
4306 *              **-EAGAIN** if bpf program can try again.
4307 *
4308 * u64 bpf_jiffies64(void)
4309 *      Description
4310 *              Obtain the 64bit jiffies
4311 *      Return
4312 *              The 64 bit jiffies
4313 *
4314 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
4315 *      Description
4316 *              For an eBPF program attached to a perf event, retrieve the
4317 *              branch records (**struct perf_branch_entry**) associated to *ctx*
4318 *              and store it in the buffer pointed by *buf* up to size
4319 *              *size* bytes.
4320 *      Return
4321 *              On success, number of bytes written to *buf*. On error, a
4322 *              negative value.
4323 *
4324 *              The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
4325 *              instead return the number of bytes required to store all the
4326 *              branch entries. If this flag is set, *buf* may be NULL.
4327 *
4328 *              **-EINVAL** if arguments invalid or **size** not a multiple
4329 *              of **sizeof**\ (**struct perf_branch_entry**\ ).
4330 *
4331 *              **-ENOENT** if architecture does not support branch records.
4332 *
4333 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4334 *      Description
4335 *              Returns 0 on success, values for *pid* and *tgid* as seen from the current
4336 *              *namespace* will be returned in *nsdata*.
4337 *      Return
4338 *              0 on success, or one of the following in case of failure:
4339 *
4340 *              **-EINVAL** if dev and inum supplied don't match dev_t and inode number
4341 *              with nsfs of current task, or if dev conversion to dev_t lost high bits.
4342 *
4343 *              **-ENOENT** if pidns does not exists for the current task.
4344 *
4345 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4346 *      Description
4347 *              Write raw *data* blob into a special BPF perf event held by
4348 *              *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4349 *              event must have the following attributes: **PERF_SAMPLE_RAW**
4350 *              as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4351 *              **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4352 *
4353 *              The *flags* are used to indicate the index in *map* for which
4354 *              the value must be put, masked with **BPF_F_INDEX_MASK**.
4355 *              Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4356 *              to indicate that the index of the current CPU core should be
4357 *              used.
4358 *
4359 *              The value to write, of *size*, is passed through eBPF stack and
4360 *              pointed by *data*.
4361 *
4362 *              *ctx* is a pointer to in-kernel struct xdp_buff.
4363 *
4364 *              This helper is similar to **bpf_perf_eventoutput**\ () but
4365 *              restricted to raw_tracepoint bpf programs.
4366 *      Return
4367 *              0 on success, or a negative error in case of failure.
4368 *
4369 * u64 bpf_get_netns_cookie(void *ctx)
4370 *      Description
4371 *              Retrieve the cookie (generated by the kernel) of the network
4372 *              namespace the input *ctx* is associated with. The network
4373 *              namespace cookie remains stable for its lifetime and provides
4374 *              a global identifier that can be assumed unique. If *ctx* is
4375 *              NULL, then the helper returns the cookie for the initial
4376 *              network namespace. The cookie itself is very similar to that
4377 *              of **bpf_get_socket_cookie**\ () helper, but for network
4378 *              namespaces instead of sockets.
4379 *      Return
4380 *              A 8-byte long opaque number.
4381 *
4382 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4383 *      Description
4384 *              Return id of cgroup v2 that is ancestor of the cgroup associated
4385 *              with the current task at the *ancestor_level*. The root cgroup
4386 *              is at *ancestor_level* zero and each step down the hierarchy
4387 *              increments the level. If *ancestor_level* == level of cgroup
4388 *              associated with the current task, then return value will be the
4389 *              same as that of **bpf_get_current_cgroup_id**\ ().
4390 *
4391 *              The helper is useful to implement policies based on cgroups
4392 *              that are upper in hierarchy than immediate cgroup associated
4393 *              with the current task.
4394 *
4395 *              The format of returned id and helper limitations are same as in
4396 *              **bpf_get_current_cgroup_id**\ ().
4397 *      Return
4398 *              The id is returned or 0 in case the id could not be retrieved.
4399 *
4400 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4401 *      Description
4402 *              Helper is overloaded depending on BPF program type. This
4403 *              description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4404 *              **BPF_PROG_TYPE_SCHED_ACT** programs.
4405 *
4406 *              Assign the *sk* to the *skb*. When combined with appropriate
4407 *              routing configuration to receive the packet towards the socket,
4408 *              will cause *skb* to be delivered to the specified socket.
4409 *              Subsequent redirection of *skb* via  **bpf_redirect**\ (),
4410 *              **bpf_clone_redirect**\ () or other methods outside of BPF may
4411 *              interfere with successful delivery to the socket.
4412 *
4413 *              This operation is only valid from TC ingress path.
4414 *
4415 *              The *flags* argument must be zero.
4416 *      Return
4417 *              0 on success, or a negative error in case of failure:
4418 *
4419 *              **-EINVAL** if specified *flags* are not supported.
4420 *
4421 *              **-ENOENT** if the socket is unavailable for assignment.
4422 *
4423 *              **-ENETUNREACH** if the socket is unreachable (wrong netns).
4424 *
4425 *              **-EOPNOTSUPP** if the operation is not supported, for example
4426 *              a call from outside of TC ingress.
4427 *
4428 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4429 *      Description
4430 *              Helper is overloaded depending on BPF program type. This
4431 *              description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4432 *
4433 *              Select the *sk* as a result of a socket lookup.
4434 *
4435 *              For the operation to succeed passed socket must be compatible
4436 *              with the packet description provided by the *ctx* object.
4437 *
4438 *              L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4439 *              be an exact match. While IP family (**AF_INET** or
4440 *              **AF_INET6**) must be compatible, that is IPv6 sockets
4441 *              that are not v6-only can be selected for IPv4 packets.
4442 *
4443 *              Only TCP listeners and UDP unconnected sockets can be
4444 *              selected. *sk* can also be NULL to reset any previous
4445 *              selection.
4446 *
4447 *              *flags* argument can combination of following values:
4448 *
4449 *              * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4450 *                socket selection, potentially done by a BPF program
4451 *                that ran before us.
4452 *
4453 *              * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4454 *                load-balancing within reuseport group for the socket
4455 *                being selected.
4456 *
4457 *              On success *ctx->sk* will point to the selected socket.
4458 *
4459 *      Return
4460 *              0 on success, or a negative errno in case of failure.
4461 *
4462 *              * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4463 *                not compatible with packet family (*ctx->family*).
4464 *
4465 *              * **-EEXIST** if socket has been already selected,
4466 *                potentially by another program, and
4467 *                **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4468 *
4469 *              * **-EINVAL** if unsupported flags were specified.
4470 *
4471 *              * **-EPROTOTYPE** if socket L4 protocol
4472 *                (*sk->protocol*) doesn't match packet protocol
4473 *                (*ctx->protocol*).
4474 *
4475 *              * **-ESOCKTNOSUPPORT** if socket is not in allowed
4476 *                state (TCP listening or UDP unconnected).
4477 *
4478 * u64 bpf_ktime_get_boot_ns(void)
4479 *      Description
4480 *              Return the time elapsed since system boot, in nanoseconds.
4481 *              Does include the time the system was suspended.
4482 *              See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4483 *      Return
4484 *              Current *ktime*.
4485 *
4486 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4487 *      Description
4488 *              **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4489 *              out the format string.
4490 *              The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4491 *              the format string itself. The *data* and *data_len* are format string
4492 *              arguments. The *data* are a **u64** array and corresponding format string
4493 *              values are stored in the array. For strings and pointers where pointees
4494 *              are accessed, only the pointer values are stored in the *data* array.
4495 *              The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4496 *
4497 *              Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4498 *              Reading kernel memory may fail due to either invalid address or
4499 *              valid address but requiring a major memory fault. If reading kernel memory
4500 *              fails, the string for **%s** will be an empty string, and the ip
4501 *              address for **%p{i,I}{4,6}** will be 0. Not returning error to
4502 *              bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4503 *      Return
4504 *              0 on success, or a negative error in case of failure:
4505 *
4506 *              **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4507 *              by returning 1 from bpf program.
4508 *
4509 *              **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4510 *
4511 *              **-E2BIG** if *fmt* contains too many format specifiers.
4512 *
4513 *              **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4514 *
4515 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4516 *      Description
4517 *              **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4518 *              The *m* represents the seq_file. The *data* and *len* represent the
4519 *              data to write in bytes.
4520 *      Return
4521 *              0 on success, or a negative error in case of failure:
4522 *
4523 *              **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4524 *
4525 * u64 bpf_sk_cgroup_id(void *sk)
4526 *      Description
4527 *              Return the cgroup v2 id of the socket *sk*.
4528 *
4529 *              *sk* must be a non-**NULL** pointer to a socket, e.g. one
4530 *              returned from **bpf_sk_lookup_xxx**\ (),
4531 *              **bpf_sk_fullsock**\ (), etc. The format of returned id is
4532 *              same as in **bpf_skb_cgroup_id**\ ().
4533 *
4534 *              This helper is available only if the kernel was compiled with
4535 *              the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4536 *      Return
4537 *              The id is returned or 0 in case the id could not be retrieved.
4538 *
4539 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4540 *      Description
4541 *              Return id of cgroup v2 that is ancestor of cgroup associated
4542 *              with the *sk* at the *ancestor_level*.  The root cgroup is at
4543 *              *ancestor_level* zero and each step down the hierarchy
4544 *              increments the level. If *ancestor_level* == level of cgroup
4545 *              associated with *sk*, then return value will be same as that
4546 *              of **bpf_sk_cgroup_id**\ ().
4547 *
4548 *              The helper is useful to implement policies based on cgroups
4549 *              that are upper in hierarchy than immediate cgroup associated
4550 *              with *sk*.
4551 *
4552 *              The format of returned id and helper limitations are same as in
4553 *              **bpf_sk_cgroup_id**\ ().
4554 *      Return
4555 *              The id is returned or 0 in case the id could not be retrieved.
4556 *
4557 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4558 *      Description
4559 *              Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4560 *              If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4561 *              of new data availability is sent.
4562 *              If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4563 *              of new data availability is sent unconditionally.
4564 *              If **0** is specified in *flags*, an adaptive notification
4565 *              of new data availability is sent.
4566 *
4567 *              An adaptive notification is a notification sent whenever the user-space
4568 *              process has caught up and consumed all available payloads. In case the user-space
4569 *              process is still processing a previous payload, then no notification is needed
4570 *              as it will process the newly added payload automatically.
4571 *      Return
4572 *              0 on success, or a negative error in case of failure.
4573 *
4574 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4575 *      Description
4576 *              Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4577 *              *flags* must be 0.
4578 *      Return
4579 *              Valid pointer with *size* bytes of memory available; NULL,
4580 *              otherwise.
4581 *
4582 * void bpf_ringbuf_submit(void *data, u64 flags)
4583 *      Description
4584 *              Submit reserved ring buffer sample, pointed to by *data*.
4585 *              If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4586 *              of new data availability is sent.
4587 *              If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4588 *              of new data availability is sent unconditionally.
4589 *              If **0** is specified in *flags*, an adaptive notification
4590 *              of new data availability is sent.
4591 *
4592 *              See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4593 *      Return
4594 *              Nothing. Always succeeds.
4595 *
4596 * void bpf_ringbuf_discard(void *data, u64 flags)
4597 *      Description
4598 *              Discard reserved ring buffer sample, pointed to by *data*.
4599 *              If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4600 *              of new data availability is sent.
4601 *              If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4602 *              of new data availability is sent unconditionally.
4603 *              If **0** is specified in *flags*, an adaptive notification
4604 *              of new data availability is sent.
4605 *
4606 *              See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4607 *      Return
4608 *              Nothing. Always succeeds.
4609 *
4610 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4611 *      Description
4612 *              Query various characteristics of provided ring buffer. What
4613 *              exactly is queries is determined by *flags*:
4614 *
4615 *              * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4616 *              * **BPF_RB_RING_SIZE**: The size of ring buffer.
4617 *              * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4618 *              * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4619 *
4620 *              Data returned is just a momentary snapshot of actual values
4621 *              and could be inaccurate, so this facility should be used to
4622 *              power heuristics and for reporting, not to make 100% correct
4623 *              calculation.
4624 *      Return
4625 *              Requested value, or 0, if *flags* are not recognized.
4626 *
4627 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4628 *      Description
4629 *              Change the skbs checksum level by one layer up or down, or
4630 *              reset it entirely to none in order to have the stack perform
4631 *              checksum validation. The level is applicable to the following
4632 *              protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4633 *              | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4634 *              through **bpf_skb_adjust_room**\ () helper with passing in
4635 *              **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4636 *              to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4637 *              the UDP header is removed. Similarly, an encap of the latter
4638 *              into the former could be accompanied by a helper call to
4639 *              **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4640 *              skb is still intended to be processed in higher layers of the
4641 *              stack instead of just egressing at tc.
4642 *
4643 *              There are three supported level settings at this time:
4644 *
4645 *              * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4646 *                with CHECKSUM_UNNECESSARY.
4647 *              * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4648 *                with CHECKSUM_UNNECESSARY.
4649 *              * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4650 *                sets CHECKSUM_NONE to force checksum validation by the stack.
4651 *              * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4652 *                skb->csum_level.
4653 *      Return
4654 *              0 on success, or a negative error in case of failure. In the
4655 *              case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4656 *              is returned or the error code -EACCES in case the skb is not
4657 *              subject to CHECKSUM_UNNECESSARY.
4658 *
4659 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4660 *      Description
4661 *              Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4662 *      Return
4663 *              *sk* if casting is valid, or **NULL** otherwise.
4664 *
4665 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4666 *      Description
4667 *              Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4668 *      Return
4669 *              *sk* if casting is valid, or **NULL** otherwise.
4670 *
4671 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4672 *      Description
4673 *              Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4674 *      Return
4675 *              *sk* if casting is valid, or **NULL** otherwise.
4676 *
4677 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4678 *      Description
4679 *              Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4680 *      Return
4681 *              *sk* if casting is valid, or **NULL** otherwise.
4682 *
4683 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4684 *      Description
4685 *              Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4686 *      Return
4687 *              *sk* if casting is valid, or **NULL** otherwise.
4688 *
4689 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4690 *      Description
4691 *              Return a user or a kernel stack in bpf program provided buffer.
4692 *              Note: the user stack will only be populated if the *task* is
4693 *              the current task; all other tasks will return -EOPNOTSUPP.
4694 *              To achieve this, the helper needs *task*, which is a valid
4695 *              pointer to **struct task_struct**. To store the stacktrace, the
4696 *              bpf program provides *buf* with a nonnegative *size*.
4697 *
4698 *              The last argument, *flags*, holds the number of stack frames to
4699 *              skip (from 0 to 255), masked with
4700 *              **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4701 *              the following flags:
4702 *
4703 *              **BPF_F_USER_STACK**
4704 *                      Collect a user space stack instead of a kernel stack.
4705 *                      The *task* must be the current task.
4706 *              **BPF_F_USER_BUILD_ID**
4707 *                      Collect buildid+offset instead of ips for user stack,
4708 *                      only valid if **BPF_F_USER_STACK** is also specified.
4709 *
4710 *              **bpf_get_task_stack**\ () can collect up to
4711 *              **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4712 *              to sufficient large buffer size. Note that
4713 *              this limit can be controlled with the **sysctl** program, and
4714 *              that it should be manually increased in order to profile long
4715 *              user stacks (such as stacks for Java programs). To do so, use:
4716 *
4717 *              ::
4718 *
4719 *                      # sysctl kernel.perf_event_max_stack=<new value>
4720 *      Return
4721 *              The non-negative copied *buf* length equal to or less than
4722 *              *size* on success, or a negative error in case of failure.
4723 *
4724 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4725 *      Description
4726 *              Load header option.  Support reading a particular TCP header
4727 *              option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4728 *
4729 *              If *flags* is 0, it will search the option from the
4730 *              *skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**
4731 *              has details on what skb_data contains under different
4732 *              *skops*\ **->op**.
4733 *
4734 *              The first byte of the *searchby_res* specifies the
4735 *              kind that it wants to search.
4736 *
4737 *              If the searching kind is an experimental kind
4738 *              (i.e. 253 or 254 according to RFC6994).  It also
4739 *              needs to specify the "magic" which is either
4740 *              2 bytes or 4 bytes.  It then also needs to
4741 *              specify the size of the magic by using
4742 *              the 2nd byte which is "kind-length" of a TCP
4743 *              header option and the "kind-length" also
4744 *              includes the first 2 bytes "kind" and "kind-length"
4745 *              itself as a normal TCP header option also does.
4746 *
4747 *              For example, to search experimental kind 254 with
4748 *              2 byte magic 0xeB9F, the searchby_res should be
4749 *              [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4750 *
4751 *              To search for the standard window scale option (3),
4752 *              the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4753 *              Note, kind-length must be 0 for regular option.
4754 *
4755 *              Searching for No-Op (0) and End-of-Option-List (1) are
4756 *              not supported.
4757 *
4758 *              *len* must be at least 2 bytes which is the minimal size
4759 *              of a header option.
4760 *
4761 *              Supported flags:
4762 *
4763 *              * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4764 *                saved_syn packet or the just-received syn packet.
4765 *
4766 *      Return
4767 *              > 0 when found, the header option is copied to *searchby_res*.
4768 *              The return value is the total length copied. On failure, a
4769 *              negative error code is returned:
4770 *
4771 *              **-EINVAL** if a parameter is invalid.
4772 *
4773 *              **-ENOMSG** if the option is not found.
4774 *
4775 *              **-ENOENT** if no syn packet is available when
4776 *              **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4777 *
4778 *              **-ENOSPC** if there is not enough space.  Only *len* number of
4779 *              bytes are copied.
4780 *
4781 *              **-EFAULT** on failure to parse the header options in the
4782 *              packet.
4783 *
4784 *              **-EPERM** if the helper cannot be used under the current
4785 *              *skops*\ **->op**.
4786 *
4787 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4788 *      Description
4789 *              Store header option.  The data will be copied
4790 *              from buffer *from* with length *len* to the TCP header.
4791 *
4792 *              The buffer *from* should have the whole option that
4793 *              includes the kind, kind-length, and the actual
4794 *              option data.  The *len* must be at least kind-length
4795 *              long.  The kind-length does not have to be 4 byte
4796 *              aligned.  The kernel will take care of the padding
4797 *              and setting the 4 bytes aligned value to th->doff.
4798 *
4799 *              This helper will check for duplicated option
4800 *              by searching the same option in the outgoing skb.
4801 *
4802 *              This helper can only be called during
4803 *              **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4804 *
4805 *      Return
4806 *              0 on success, or negative error in case of failure:
4807 *
4808 *              **-EINVAL** If param is invalid.
4809 *
4810 *              **-ENOSPC** if there is not enough space in the header.
4811 *              Nothing has been written
4812 *
4813 *              **-EEXIST** if the option already exists.
4814 *
4815 *              **-EFAULT** on failure to parse the existing header options.
4816 *
4817 *              **-EPERM** if the helper cannot be used under the current
4818 *              *skops*\ **->op**.
4819 *
4820 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4821 *      Description
4822 *              Reserve *len* bytes for the bpf header option.  The
4823 *              space will be used by **bpf_store_hdr_opt**\ () later in
4824 *              **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4825 *
4826 *              If **bpf_reserve_hdr_opt**\ () is called multiple times,
4827 *              the total number of bytes will be reserved.
4828 *
4829 *              This helper can only be called during
4830 *              **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4831 *
4832 *      Return
4833 *              0 on success, or negative error in case of failure:
4834 *
4835 *              **-EINVAL** if a parameter is invalid.
4836 *
4837 *              **-ENOSPC** if there is not enough space in the header.
4838 *
4839 *              **-EPERM** if the helper cannot be used under the current
4840 *              *skops*\ **->op**.
4841 *
4842 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4843 *      Description
4844 *              Get a bpf_local_storage from an *inode*.
4845 *
4846 *              Logically, it could be thought of as getting the value from
4847 *              a *map* with *inode* as the **key**.  From this
4848 *              perspective,  the usage is not much different from
4849 *              **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4850 *              helper enforces the key must be an inode and the map must also
4851 *              be a **BPF_MAP_TYPE_INODE_STORAGE**.
4852 *
4853 *              Underneath, the value is stored locally at *inode* instead of
4854 *              the *map*.  The *map* is used as the bpf-local-storage
4855 *              "type". The bpf-local-storage "type" (i.e. the *map*) is
4856 *              searched against all bpf_local_storage residing at *inode*.
4857 *
4858 *              An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4859 *              used such that a new bpf_local_storage will be
4860 *              created if one does not exist.  *value* can be used
4861 *              together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4862 *              the initial value of a bpf_local_storage.  If *value* is
4863 *              **NULL**, the new bpf_local_storage will be zero initialized.
4864 *      Return
4865 *              A bpf_local_storage pointer is returned on success.
4866 *
4867 *              **NULL** if not found or there was an error in adding
4868 *              a new bpf_local_storage.
4869 *
4870 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4871 *      Description
4872 *              Delete a bpf_local_storage from an *inode*.
4873 *      Return
4874 *              0 on success.
4875 *
4876 *              **-ENOENT** if the bpf_local_storage cannot be found.
4877 *
4878 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4879 *      Description
4880 *              Return full path for given **struct path** object, which
4881 *              needs to be the kernel BTF *path* object. The path is
4882 *              returned in the provided buffer *buf* of size *sz* and
4883 *              is zero terminated.
4884 *
4885 *      Return
4886 *              On success, the strictly positive length of the string,
4887 *              including the trailing NUL character. On error, a negative
4888 *              value.
4889 *
4890 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4891 *      Description
4892 *              Read *size* bytes from user space address *user_ptr* and store
4893 *              the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4894 *      Return
4895 *              0 on success, or a negative error in case of failure.
4896 *
4897 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4898 *      Description
4899 *              Use BTF to store a string representation of *ptr*->ptr in *str*,
4900 *              using *ptr*->type_id.  This value should specify the type
4901 *              that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4902 *              can be used to look up vmlinux BTF type ids. Traversing the
4903 *              data structure using BTF, the type information and values are
4904 *              stored in the first *str_size* - 1 bytes of *str*.  Safe copy of
4905 *              the pointer data is carried out to avoid kernel crashes during
4906 *              operation.  Smaller types can use string space on the stack;
4907 *              larger programs can use map data to store the string
4908 *              representation.
4909 *
4910 *              The string can be subsequently shared with userspace via
4911 *              bpf_perf_event_output() or ring buffer interfaces.
4912 *              bpf_trace_printk() is to be avoided as it places too small
4913 *              a limit on string size to be useful.
4914 *
4915 *              *flags* is a combination of
4916 *
4917 *              **BTF_F_COMPACT**
4918 *                      no formatting around type information
4919 *              **BTF_F_NONAME**
4920 *                      no struct/union member names/types
4921 *              **BTF_F_PTR_RAW**
4922 *                      show raw (unobfuscated) pointer values;
4923 *                      equivalent to printk specifier %px.
4924 *              **BTF_F_ZERO**
4925 *                      show zero-valued struct/union members; they
4926 *                      are not displayed by default
4927 *
4928 *      Return
4929 *              The number of bytes that were written (or would have been
4930 *              written if output had to be truncated due to string size),
4931 *              or a negative error in cases of failure.
4932 *
4933 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4934 *      Description
4935 *              Use BTF to write to seq_write a string representation of
4936 *              *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4937 *              *flags* are identical to those used for bpf_snprintf_btf.
4938 *      Return
4939 *              0 on success or a negative error in case of failure.
4940 *
4941 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4942 *      Description
4943 *              See **bpf_get_cgroup_classid**\ () for the main description.
4944 *              This helper differs from **bpf_get_cgroup_classid**\ () in that
4945 *              the cgroup v1 net_cls class is retrieved only from the *skb*'s
4946 *              associated socket instead of the current process.
4947 *      Return
4948 *              The id is returned or 0 in case the id could not be retrieved.
4949 *
4950 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4951 *      Description
4952 *              Redirect the packet to another net device of index *ifindex*
4953 *              and fill in L2 addresses from neighboring subsystem. This helper
4954 *              is somewhat similar to **bpf_redirect**\ (), except that it
4955 *              populates L2 addresses as well, meaning, internally, the helper
4956 *              relies on the neighbor lookup for the L2 address of the nexthop.
4957 *
4958 *              The helper will perform a FIB lookup based on the skb's
4959 *              networking header to get the address of the next hop, unless
4960 *              this is supplied by the caller in the *params* argument. The
4961 *              *plen* argument indicates the len of *params* and should be set
4962 *              to 0 if *params* is NULL.
4963 *
4964 *              The *flags* argument is reserved and must be 0. The helper is
4965 *              currently only supported for tc BPF program types, and enabled
4966 *              for IPv4 and IPv6 protocols.
4967 *      Return
4968 *              The helper returns **TC_ACT_REDIRECT** on success or
4969 *              **TC_ACT_SHOT** on error.
4970 *
4971 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4972 *     Description
4973 *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4974 *             pointer to the percpu kernel variable on *cpu*. A ksym is an
4975 *             extern variable decorated with '__ksym'. For ksym, there is a
4976 *             global var (either static or global) defined of the same name
4977 *             in the kernel. The ksym is percpu if the global var is percpu.
4978 *             The returned pointer points to the global percpu var on *cpu*.
4979 *
4980 *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4981 *             kernel, except that bpf_per_cpu_ptr() may return NULL. This
4982 *             happens if *cpu* is larger than nr_cpu_ids. The caller of
4983 *             bpf_per_cpu_ptr() must check the returned value.
4984 *     Return
4985 *             A pointer pointing to the kernel percpu variable on *cpu*, or
4986 *             NULL, if *cpu* is invalid.
4987 *
4988 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4989 *      Description
4990 *              Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4991 *              pointer to the percpu kernel variable on this cpu. See the
4992 *              description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4993 *
4994 *              bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4995 *              the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4996 *              never return NULL.
4997 *      Return
4998 *              A pointer pointing to the kernel percpu variable on this cpu.
4999 *
5000 * long bpf_redirect_peer(u32 ifindex, u64 flags)
5001 *      Description
5002 *              Redirect the packet to another net device of index *ifindex*.
5003 *              This helper is somewhat similar to **bpf_redirect**\ (), except
5004 *              that the redirection happens to the *ifindex*' peer device and
5005 *              the netns switch takes place from ingress to ingress without
5006 *              going through the CPU's backlog queue.
5007 *
5008 *              *skb*\ **->mark** and *skb*\ **->tstamp** are not cleared during
5009 *              the netns switch.
5010 *
5011 *              The *flags* argument is reserved and must be 0. The helper is
5012 *              currently only supported for tc BPF program types at the
5013 *              ingress hook and for veth and netkit target device types. The
5014 *              peer device must reside in a different network namespace.
5015 *      Return
5016 *              The helper returns **TC_ACT_REDIRECT** on success or
5017 *              **TC_ACT_SHOT** on error.
5018 *
5019 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
5020 *      Description
5021 *              Get a bpf_local_storage from the *task*.
5022 *
5023 *              Logically, it could be thought of as getting the value from
5024 *              a *map* with *task* as the **key**.  From this
5025 *              perspective,  the usage is not much different from
5026 *              **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
5027 *              helper enforces the key must be a task_struct and the map must also
5028 *              be a **BPF_MAP_TYPE_TASK_STORAGE**.
5029 *
5030 *              Underneath, the value is stored locally at *task* instead of
5031 *              the *map*.  The *map* is used as the bpf-local-storage
5032 *              "type". The bpf-local-storage "type" (i.e. the *map*) is
5033 *              searched against all bpf_local_storage residing at *task*.
5034 *
5035 *              An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5036 *              used such that a new bpf_local_storage will be
5037 *              created if one does not exist.  *value* can be used
5038 *              together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5039 *              the initial value of a bpf_local_storage.  If *value* is
5040 *              **NULL**, the new bpf_local_storage will be zero initialized.
5041 *      Return
5042 *              A bpf_local_storage pointer is returned on success.
5043 *
5044 *              **NULL** if not found or there was an error in adding
5045 *              a new bpf_local_storage.
5046 *
5047 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
5048 *      Description
5049 *              Delete a bpf_local_storage from a *task*.
5050 *      Return
5051 *              0 on success.
5052 *
5053 *              **-ENOENT** if the bpf_local_storage cannot be found.
5054 *
5055 * struct task_struct *bpf_get_current_task_btf(void)
5056 *      Description
5057 *              Return a BTF pointer to the "current" task.
5058 *              This pointer can also be used in helpers that accept an
5059 *              *ARG_PTR_TO_BTF_ID* of type *task_struct*.
5060 *      Return
5061 *              Pointer to the current task.
5062 *
5063 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
5064 *      Description
5065 *              Set or clear certain options on *bprm*:
5066 *
5067 *              **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
5068 *              which sets the **AT_SECURE** auxv for glibc. The bit
5069 *              is cleared if the flag is not specified.
5070 *      Return
5071 *              **-EINVAL** if invalid *flags* are passed, zero otherwise.
5072 *
5073 * u64 bpf_ktime_get_coarse_ns(void)
5074 *      Description
5075 *              Return a coarse-grained version of the time elapsed since
5076 *              system boot, in nanoseconds. Does not include time the system
5077 *              was suspended.
5078 *
5079 *              See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
5080 *      Return
5081 *              Current *ktime*.
5082 *
5083 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
5084 *      Description
5085 *              Returns the stored IMA hash of the *inode* (if it's available).
5086 *              If the hash is larger than *size*, then only *size*
5087 *              bytes will be copied to *dst*
5088 *      Return
5089 *              The **hash_algo** is returned on success,
5090 *              **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if
5091 *              invalid arguments are passed.
5092 *
5093 * struct socket *bpf_sock_from_file(struct file *file)
5094 *      Description
5095 *              If the given file represents a socket, returns the associated
5096 *              socket.
5097 *      Return
5098 *              A pointer to a struct socket on success or NULL if the file is
5099 *              not a socket.
5100 *
5101 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
5102 *      Description
5103 *              Check packet size against exceeding MTU of net device (based
5104 *              on *ifindex*).  This helper will likely be used in combination
5105 *              with helpers that adjust/change the packet size.
5106 *
5107 *              The argument *len_diff* can be used for querying with a planned
5108 *              size change. This allows to check MTU prior to changing packet
5109 *              ctx. Providing a *len_diff* adjustment that is larger than the
5110 *              actual packet size (resulting in negative packet size) will in
5111 *              principle not exceed the MTU, which is why it is not considered
5112 *              a failure.  Other BPF helpers are needed for performing the
5113 *              planned size change; therefore the responsibility for catching
5114 *              a negative packet size belongs in those helpers.
5115 *
5116 *              Specifying *ifindex* zero means the MTU check is performed
5117 *              against the current net device.  This is practical if this isn't
5118 *              used prior to redirect.
5119 *
5120 *              On input *mtu_len* must be a valid pointer, else verifier will
5121 *              reject BPF program.  If the value *mtu_len* is initialized to
5122 *              zero then the ctx packet size is use.  When value *mtu_len* is
5123 *              provided as input this specify the L3 length that the MTU check
5124 *              is done against. Remember XDP and TC length operate at L2, but
5125 *              this value is L3 as this correlate to MTU and IP-header tot_len
5126 *              values which are L3 (similar behavior as bpf_fib_lookup).
5127 *
5128 *              The Linux kernel route table can configure MTUs on a more
5129 *              specific per route level, which is not provided by this helper.
5130 *              For route level MTU checks use the **bpf_fib_lookup**\ ()
5131 *              helper.
5132 *
5133 *              *ctx* is either **struct xdp_md** for XDP programs or
5134 *              **struct sk_buff** for tc cls_act programs.
5135 *
5136 *              The *flags* argument can be a combination of one or more of the
5137 *              following values:
5138 *
5139 *              **BPF_MTU_CHK_SEGS**
5140 *                      This flag will only works for *ctx* **struct sk_buff**.
5141 *                      If packet context contains extra packet segment buffers
5142 *                      (often knows as GSO skb), then MTU check is harder to
5143 *                      check at this point, because in transmit path it is
5144 *                      possible for the skb packet to get re-segmented
5145 *                      (depending on net device features).  This could still be
5146 *                      a MTU violation, so this flag enables performing MTU
5147 *                      check against segments, with a different violation
5148 *                      return code to tell it apart. Check cannot use len_diff.
5149 *
5150 *              On return *mtu_len* pointer contains the MTU value of the net
5151 *              device.  Remember the net device configured MTU is the L3 size,
5152 *              which is returned here and XDP and TC length operate at L2.
5153 *              Helper take this into account for you, but remember when using
5154 *              MTU value in your BPF-code.
5155 *
5156 *      Return
5157 *              * 0 on success, and populate MTU value in *mtu_len* pointer.
5158 *
5159 *              * < 0 if any input argument is invalid (*mtu_len* not updated)
5160 *
5161 *              MTU violations return positive values, but also populate MTU
5162 *              value in *mtu_len* pointer, as this can be needed for
5163 *              implementing PMTU handing:
5164 *
5165 *              * **BPF_MTU_CHK_RET_FRAG_NEEDED**
5166 *              * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
5167 *
5168 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
5169 *      Description
5170 *              For each element in **map**, call **callback_fn** function with
5171 *              **map**, **callback_ctx** and other map-specific parameters.
5172 *              The **callback_fn** should be a static function and
5173 *              the **callback_ctx** should be a pointer to the stack.
5174 *              The **flags** is used to control certain aspects of the helper.
5175 *              Currently, the **flags** must be 0.
5176 *
5177 *              The following are a list of supported map types and their
5178 *              respective expected callback signatures:
5179 *
5180 *              BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
5181 *              BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
5182 *              BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
5183 *
5184 *              long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
5185 *
5186 *              For per_cpu maps, the map_value is the value on the cpu where the
5187 *              bpf_prog is running.
5188 *
5189 *              If **callback_fn** return 0, the helper will continue to the next
5190 *              element. If return value is 1, the helper will skip the rest of
5191 *              elements and return. Other return values are not used now.
5192 *
5193 *      Return
5194 *              The number of traversed map elements for success, **-EINVAL** for
5195 *              invalid **flags**.
5196 *
5197 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
5198 *      Description
5199 *              Outputs a string into the **str** buffer of size **str_size**
5200 *              based on a format string stored in a read-only map pointed by
5201 *              **fmt**.
5202 *
5203 *              Each format specifier in **fmt** corresponds to one u64 element
5204 *              in the **data** array. For strings and pointers where pointees
5205 *              are accessed, only the pointer values are stored in the *data*
5206 *              array. The *data_len* is the size of *data* in bytes - must be
5207 *              a multiple of 8.
5208 *
5209 *              Formats **%s** and **%p{i,I}{4,6}** require to read kernel
5210 *              memory. Reading kernel memory may fail due to either invalid
5211 *              address or valid address but requiring a major memory fault. If
5212 *              reading kernel memory fails, the string for **%s** will be an
5213 *              empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
5214 *              Not returning error to bpf program is consistent with what
5215 *              **bpf_trace_printk**\ () does for now.
5216 *
5217 *      Return
5218 *              The strictly positive length of the formatted string, including
5219 *              the trailing zero character. If the return value is greater than
5220 *              **str_size**, **str** contains a truncated string, guaranteed to
5221 *              be zero-terminated except when **str_size** is 0.
5222 *
5223 *              Or **-EBUSY** if the per-CPU memory copy buffer is busy.
5224 *
5225 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
5226 *      Description
5227 *              Execute bpf syscall with given arguments.
5228 *      Return
5229 *              A syscall result.
5230 *
5231 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
5232 *      Description
5233 *              Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
5234 *      Return
5235 *              Returns btf_id and btf_obj_fd in lower and upper 32 bits.
5236 *
5237 * long bpf_sys_close(u32 fd)
5238 *      Description
5239 *              Execute close syscall for given FD.
5240 *      Return
5241 *              A syscall result.
5242 *
5243 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
5244 *      Description
5245 *              Initialize the timer.
5246 *              First 4 bits of *flags* specify clockid.
5247 *              Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
5248 *              All other bits of *flags* are reserved.
5249 *              The verifier will reject the program if *timer* is not from
5250 *              the same *map*.
5251 *      Return
5252 *              0 on success.
5253 *              **-EBUSY** if *timer* is already initialized.
5254 *              **-EINVAL** if invalid *flags* are passed.
5255 *              **-EPERM** if *timer* is in a map that doesn't have any user references.
5256 *              The user space should either hold a file descriptor to a map with timers
5257 *              or pin such map in bpffs. When map is unpinned or file descriptor is
5258 *              closed all timers in the map will be cancelled and freed.
5259 *
5260 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
5261 *      Description
5262 *              Configure the timer to call *callback_fn* static function.
5263 *      Return
5264 *              0 on success.
5265 *              **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5266 *              **-EPERM** if *timer* is in a map that doesn't have any user references.
5267 *              The user space should either hold a file descriptor to a map with timers
5268 *              or pin such map in bpffs. When map is unpinned or file descriptor is
5269 *              closed all timers in the map will be cancelled and freed.
5270 *
5271 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
5272 *      Description
5273 *              Set timer expiration N nanoseconds from the current time. The
5274 *              configured callback will be invoked in soft irq context on some cpu
5275 *              and will not repeat unless another bpf_timer_start() is made.
5276 *              In such case the next invocation can migrate to a different cpu.
5277 *              Since struct bpf_timer is a field inside map element the map
5278 *              owns the timer. The bpf_timer_set_callback() will increment refcnt
5279 *              of BPF program to make sure that callback_fn code stays valid.
5280 *              When user space reference to a map reaches zero all timers
5281 *              in a map are cancelled and corresponding program's refcnts are
5282 *              decremented. This is done to make sure that Ctrl-C of a user
5283 *              process doesn't leave any timers running. If map is pinned in
5284 *              bpffs the callback_fn can re-arm itself indefinitely.
5285 *              bpf_map_update/delete_elem() helpers and user space sys_bpf commands
5286 *              cancel and free the timer in the given map element.
5287 *              The map can contain timers that invoke callback_fn-s from different
5288 *              programs. The same callback_fn can serve different timers from
5289 *              different maps if key/value layout matches across maps.
5290 *              Every bpf_timer_set_callback() can have different callback_fn.
5291 *
5292 *              *flags* can be one of:
5293 *
5294 *              **BPF_F_TIMER_ABS**
5295 *                      Start the timer in absolute expire value instead of the
5296 *                      default relative one.
5297 *              **BPF_F_TIMER_CPU_PIN**
5298 *                      Timer will be pinned to the CPU of the caller.
5299 *
5300 *      Return
5301 *              0 on success.
5302 *              **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
5303 *              or invalid *flags* are passed.
5304 *
5305 * long bpf_timer_cancel(struct bpf_timer *timer)
5306 *      Description
5307 *              Cancel the timer and wait for callback_fn to finish if it was running.
5308 *      Return
5309 *              0 if the timer was not active.
5310 *              1 if the timer was active.
5311 *              **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5312 *              **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
5313 *              own timer which would have led to a deadlock otherwise.
5314 *
5315 * u64 bpf_get_func_ip(void *ctx)
5316 *      Description
5317 *              Get address of the traced function (for tracing and kprobe programs).
5318 *
5319 *              When called for kprobe program attached as uprobe it returns
5320 *              probe address for both entry and return uprobe.
5321 *
5322 *      Return
5323 *              Address of the traced function for kprobe.
5324 *              0 for kprobes placed within the function (not at the entry).
5325 *              Address of the probe for uprobe and return uprobe.
5326 *
5327 * u64 bpf_get_attach_cookie(void *ctx)
5328 *      Description
5329 *              Get bpf_cookie value provided (optionally) during the program
5330 *              attachment. It might be different for each individual
5331 *              attachment, even if BPF program itself is the same.
5332 *              Expects BPF program context *ctx* as a first argument.
5333 *
5334 *              Supported for the following program types:
5335 *                      - kprobe/uprobe;
5336 *                      - tracepoint;
5337 *                      - perf_event.
5338 *      Return
5339 *              Value specified by user at BPF link creation/attachment time
5340 *              or 0, if it was not specified.
5341 *
5342 * long bpf_task_pt_regs(struct task_struct *task)
5343 *      Description
5344 *              Get the struct pt_regs associated with **task**.
5345 *      Return
5346 *              A pointer to struct pt_regs.
5347 *
5348 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5349 *      Description
5350 *              Get branch trace from hardware engines like Intel LBR. The
5351 *              hardware engine is stopped shortly after the helper is
5352 *              called. Therefore, the user need to filter branch entries
5353 *              based on the actual use case. To capture branch trace
5354 *              before the trigger point of the BPF program, the helper
5355 *              should be called at the beginning of the BPF program.
5356 *
5357 *              The data is stored as struct perf_branch_entry into output
5358 *              buffer *entries*. *size* is the size of *entries* in bytes.
5359 *              *flags* is reserved for now and must be zero.
5360 *
5361 *      Return
5362 *              On success, number of bytes written to *buf*. On error, a
5363 *              negative value.
5364 *
5365 *              **-EINVAL** if *flags* is not zero.
5366 *
5367 *              **-ENOENT** if architecture does not support branch records.
5368 *
5369 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5370 *      Description
5371 *              Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5372 *              to format and can handle more format args as a result.
5373 *
5374 *              Arguments are to be used as in **bpf_seq_printf**\ () helper.
5375 *      Return
5376 *              The number of bytes written to the buffer, or a negative error
5377 *              in case of failure.
5378 *
5379 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5380 *      Description
5381 *              Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5382 *      Return
5383 *              *sk* if casting is valid, or **NULL** otherwise.
5384 *
5385 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5386 *      Description
5387 *              Get the address of a kernel symbol, returned in *res*. *res* is
5388 *              set to 0 if the symbol is not found.
5389 *      Return
5390 *              On success, zero. On error, a negative value.
5391 *
5392 *              **-EINVAL** if *flags* is not zero.
5393 *
5394 *              **-EINVAL** if string *name* is not the same size as *name_sz*.
5395 *
5396 *              **-ENOENT** if symbol is not found.
5397 *
5398 *              **-EPERM** if caller does not have permission to obtain kernel address.
5399 *
5400 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5401 *      Description
5402 *              Find vma of *task* that contains *addr*, call *callback_fn*
5403 *              function with *task*, *vma*, and *callback_ctx*.
5404 *              The *callback_fn* should be a static function and
5405 *              the *callback_ctx* should be a pointer to the stack.
5406 *              The *flags* is used to control certain aspects of the helper.
5407 *              Currently, the *flags* must be 0.
5408 *
5409 *              The expected callback signature is
5410 *
5411 *              long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5412 *
5413 *      Return
5414 *              0 on success.
5415 *              **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5416 *              **-EBUSY** if failed to try lock mmap_lock.
5417 *              **-EINVAL** for invalid **flags**.
5418 *
5419 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5420 *      Description
5421 *              For **nr_loops**, call **callback_fn** function
5422 *              with **callback_ctx** as the context parameter.
5423 *              The **callback_fn** should be a static function and
5424 *              the **callback_ctx** should be a pointer to the stack.
5425 *              The **flags** is used to control certain aspects of the helper.
5426 *              Currently, the **flags** must be 0. Currently, nr_loops is
5427 *              limited to 1 << 23 (~8 million) loops.
5428 *
5429 *              long (\*callback_fn)(u64 index, void \*ctx);
5430 *
5431 *              where **index** is the current index in the loop. The index
5432 *              is zero-indexed.
5433 *
5434 *              If **callback_fn** returns 0, the helper will continue to the next
5435 *              loop. If return value is 1, the helper will skip the rest of
5436 *              the loops and return. Other return values are not used now,
5437 *              and will be rejected by the verifier.
5438 *
5439 *      Return
5440 *              The number of loops performed, **-EINVAL** for invalid **flags**,
5441 *              **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5442 *
5443 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5444 *      Description
5445 *              Do strncmp() between **s1** and **s2**. **s1** doesn't need
5446 *              to be null-terminated and **s1_sz** is the maximum storage
5447 *              size of **s1**. **s2** must be a read-only string.
5448 *      Return
5449 *              An integer less than, equal to, or greater than zero
5450 *              if the first **s1_sz** bytes of **s1** is found to be
5451 *              less than, to match, or be greater than **s2**.
5452 *
5453 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5454 *      Description
5455 *              Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5456 *              returned in **value**.
5457 *
5458 *      Return
5459 *              0 on success.
5460 *              **-EINVAL** if n >= argument register count of traced function.
5461 *
5462 * long bpf_get_func_ret(void *ctx, u64 *value)
5463 *      Description
5464 *              Get return value of the traced function (for tracing programs)
5465 *              in **value**.
5466 *
5467 *      Return
5468 *              0 on success.
5469 *              **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5470 *
5471 * long bpf_get_func_arg_cnt(void *ctx)
5472 *      Description
5473 *              Get number of registers of the traced function (for tracing programs) where
5474 *              function arguments are stored in these registers.
5475 *
5476 *      Return
5477 *              The number of argument registers of the traced function.
5478 *
5479 * int bpf_get_retval(void)
5480 *      Description
5481 *              Get the BPF program's return value that will be returned to the upper layers.
5482 *
5483 *              This helper is currently supported by cgroup programs and only by the hooks
5484 *              where BPF program's return value is returned to the userspace via errno.
5485 *      Return
5486 *              The BPF program's return value.
5487 *
5488 * int bpf_set_retval(int retval)
5489 *      Description
5490 *              Set the BPF program's return value that will be returned to the upper layers.
5491 *
5492 *              This helper is currently supported by cgroup programs and only by the hooks
5493 *              where BPF program's return value is returned to the userspace via errno.
5494 *
5495 *              Note that there is the following corner case where the program exports an error
5496 *              via bpf_set_retval but signals success via 'return 1':
5497 *
5498 *                      bpf_set_retval(-EPERM);
5499 *                      return 1;
5500 *
5501 *              In this case, the BPF program's return value will use helper's -EPERM. This
5502 *              still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5503 *
5504 *      Return
5505 *              0 on success, or a negative error in case of failure.
5506 *
5507 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5508 *      Description
5509 *              Get the total size of a given xdp buff (linear and paged area)
5510 *      Return
5511 *              The total size of a given xdp buffer.
5512 *
5513 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5514 *      Description
5515 *              This helper is provided as an easy way to load data from a
5516 *              xdp buffer. It can be used to load *len* bytes from *offset* from
5517 *              the frame associated to *xdp_md*, into the buffer pointed by
5518 *              *buf*.
5519 *      Return
5520 *              0 on success, or a negative error in case of failure.
5521 *
5522 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5523 *      Description
5524 *              Store *len* bytes from buffer *buf* into the frame
5525 *              associated to *xdp_md*, at *offset*.
5526 *      Return
5527 *              0 on success, or a negative error in case of failure.
5528 *
5529 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5530 *      Description
5531 *              Read *size* bytes from user space address *user_ptr* in *tsk*'s
5532 *              address space, and stores the data in *dst*. *flags* is not
5533 *              used yet and is provided for future extensibility. This helper
5534 *              can only be used by sleepable programs.
5535 *      Return
5536 *              0 on success, or a negative error in case of failure. On error
5537 *              *dst* buffer is zeroed out.
5538 *
5539 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5540 *      Description
5541 *              Change the __sk_buff->tstamp_type to *tstamp_type*
5542 *              and set *tstamp* to the __sk_buff->tstamp together.
5543 *
5544 *              If there is no need to change the __sk_buff->tstamp_type,
5545 *              the tstamp value can be directly written to __sk_buff->tstamp
5546 *              instead.
5547 *
5548 *              BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5549 *              will be kept during bpf_redirect_*().  A non zero
5550 *              *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5551 *              *tstamp_type*.
5552 *
5553 *              A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5554 *              with a zero *tstamp*.
5555 *
5556 *              Only IPv4 and IPv6 skb->protocol are supported.
5557 *
5558 *              This function is most useful when it needs to set a
5559 *              mono delivery time to __sk_buff->tstamp and then
5560 *              bpf_redirect_*() to the egress of an iface.  For example,
5561 *              changing the (rcv) timestamp in __sk_buff->tstamp at
5562 *              ingress to a mono delivery time and then bpf_redirect_*()
5563 *              to sch_fq@phy-dev.
5564 *      Return
5565 *              0 on success.
5566 *              **-EINVAL** for invalid input
5567 *              **-EOPNOTSUPP** for unsupported protocol
5568 *
5569 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5570 *      Description
5571 *              Returns a calculated IMA hash of the *file*.
5572 *              If the hash is larger than *size*, then only *size*
5573 *              bytes will be copied to *dst*
5574 *      Return
5575 *              The **hash_algo** is returned on success,
5576 *              **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if
5577 *              invalid arguments are passed.
5578 *
5579 * void *bpf_kptr_xchg(void *dst, void *ptr)
5580 *      Description
5581 *              Exchange kptr at pointer *dst* with *ptr*, and return the old value.
5582 *              *dst* can be map value or local kptr. *ptr* can be NULL, otherwise
5583 *              it must be a referenced pointer which will be released when this helper
5584 *              is called.
5585 *      Return
5586 *              The old value of kptr (which can be NULL). The returned pointer
5587 *              if not NULL, is a reference which must be released using its
5588 *              corresponding release function, or moved into a BPF map before
5589 *              program exit.
5590 *
5591 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5592 *      Description
5593 *              Perform a lookup in *percpu map* for an entry associated to
5594 *              *key* on *cpu*.
5595 *      Return
5596 *              Map value associated to *key* on *cpu*, or **NULL** if no entry
5597 *              was found or *cpu* is invalid.
5598 *
5599 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5600 *      Description
5601 *              Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5602 *      Return
5603 *              *sk* if casting is valid, or **NULL** otherwise.
5604 *
5605 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5606 *      Description
5607 *              Get a dynptr to local memory *data*.
5608 *
5609 *              *data* must be a ptr to a map value.
5610 *              The maximum *size* supported is DYNPTR_MAX_SIZE.
5611 *              *flags* is currently unused.
5612 *      Return
5613 *              0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5614 *              -EINVAL if flags is not 0.
5615 *
5616 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5617 *      Description
5618 *              Reserve *size* bytes of payload in a ring buffer *ringbuf*
5619 *              through the dynptr interface. *flags* must be 0.
5620 *
5621 *              Please note that a corresponding bpf_ringbuf_submit_dynptr or
5622 *              bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5623 *              reservation fails. This is enforced by the verifier.
5624 *      Return
5625 *              0 on success, or a negative error in case of failure.
5626 *
5627 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5628 *      Description
5629 *              Submit reserved ring buffer sample, pointed to by *data*,
5630 *              through the dynptr interface. This is a no-op if the dynptr is
5631 *              invalid/null.
5632 *
5633 *              For more information on *flags*, please see
5634 *              'bpf_ringbuf_submit'.
5635 *      Return
5636 *              Nothing. Always succeeds.
5637 *
5638 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5639 *      Description
5640 *              Discard reserved ring buffer sample through the dynptr
5641 *              interface. This is a no-op if the dynptr is invalid/null.
5642 *
5643 *              For more information on *flags*, please see
5644 *              'bpf_ringbuf_discard'.
5645 *      Return
5646 *              Nothing. Always succeeds.
5647 *
5648 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
5649 *      Description
5650 *              Read *len* bytes from *src* into *dst*, starting from *offset*
5651 *              into *src*.
5652 *              *flags* is currently unused.
5653 *      Return
5654 *              0 on success, -E2BIG if *offset* + *len* exceeds the length
5655 *              of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5656 *              *flags* is not 0.
5657 *
5658 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5659 *      Description
5660 *              Write *len* bytes from *src* into *dst*, starting from *offset*
5661 *              into *dst*.
5662 *
5663 *              *flags* must be 0 except for skb-type dynptrs.
5664 *
5665 *              For skb-type dynptrs:
5666 *                  *  All data slices of the dynptr are automatically
5667 *                     invalidated after **bpf_dynptr_write**\ (). This is
5668 *                     because writing may pull the skb and change the
5669 *                     underlying packet buffer.
5670 *
5671 *                  *  For *flags*, please see the flags accepted by
5672 *                     **bpf_skb_store_bytes**\ ().
5673 *      Return
5674 *              0 on success, -E2BIG if *offset* + *len* exceeds the length
5675 *              of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5676 *              is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
5677 *              other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
5678 *
5679 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
5680 *      Description
5681 *              Get a pointer to the underlying dynptr data.
5682 *
5683 *              *len* must be a statically known value. The returned data slice
5684 *              is invalidated whenever the dynptr is invalidated.
5685 *
5686 *              skb and xdp type dynptrs may not use bpf_dynptr_data. They should
5687 *              instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
5688 *      Return
5689 *              Pointer to the underlying dynptr data, NULL if the dynptr is
5690 *              read-only, if the dynptr is invalid, or if the offset and length
5691 *              is out of bounds.
5692 *
5693 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5694 *      Description
5695 *              Try to issue a SYN cookie for the packet with corresponding
5696 *              IPv4/TCP headers, *iph* and *th*, without depending on a
5697 *              listening socket.
5698 *
5699 *              *iph* points to the IPv4 header.
5700 *
5701 *              *th* points to the start of the TCP header, while *th_len*
5702 *              contains the length of the TCP header (at least
5703 *              **sizeof**\ (**struct tcphdr**)).
5704 *      Return
5705 *              On success, lower 32 bits hold the generated SYN cookie in
5706 *              followed by 16 bits which hold the MSS value for that cookie,
5707 *              and the top 16 bits are unused.
5708 *
5709 *              On failure, the returned value is one of the following:
5710 *
5711 *              **-EINVAL** if *th_len* is invalid.
5712 *
5713 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5714 *      Description
5715 *              Try to issue a SYN cookie for the packet with corresponding
5716 *              IPv6/TCP headers, *iph* and *th*, without depending on a
5717 *              listening socket.
5718 *
5719 *              *iph* points to the IPv6 header.
5720 *
5721 *              *th* points to the start of the TCP header, while *th_len*
5722 *              contains the length of the TCP header (at least
5723 *              **sizeof**\ (**struct tcphdr**)).
5724 *      Return
5725 *              On success, lower 32 bits hold the generated SYN cookie in
5726 *              followed by 16 bits which hold the MSS value for that cookie,
5727 *              and the top 16 bits are unused.
5728 *
5729 *              On failure, the returned value is one of the following:
5730 *
5731 *              **-EINVAL** if *th_len* is invalid.
5732 *
5733 *              **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5734 *
5735 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5736 *      Description
5737 *              Check whether *iph* and *th* contain a valid SYN cookie ACK
5738 *              without depending on a listening socket.
5739 *
5740 *              *iph* points to the IPv4 header.
5741 *
5742 *              *th* points to the TCP header.
5743 *      Return
5744 *              0 if *iph* and *th* are a valid SYN cookie ACK.
5745 *
5746 *              On failure, the returned value is one of the following:
5747 *
5748 *              **-EACCES** if the SYN cookie is not valid.
5749 *
5750 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5751 *      Description
5752 *              Check whether *iph* and *th* contain a valid SYN cookie ACK
5753 *              without depending on a listening socket.
5754 *
5755 *              *iph* points to the IPv6 header.
5756 *
5757 *              *th* points to the TCP header.
5758 *      Return
5759 *              0 if *iph* and *th* are a valid SYN cookie ACK.
5760 *
5761 *              On failure, the returned value is one of the following:
5762 *
5763 *              **-EACCES** if the SYN cookie is not valid.
5764 *
5765 *              **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5766 *
5767 * u64 bpf_ktime_get_tai_ns(void)
5768 *      Description
5769 *              A nonsettable system-wide clock derived from wall-clock time but
5770 *              ignoring leap seconds.  This clock does not experience
5771 *              discontinuities and backwards jumps caused by NTP inserting leap
5772 *              seconds as CLOCK_REALTIME does.
5773 *
5774 *              See: **clock_gettime**\ (**CLOCK_TAI**)
5775 *      Return
5776 *              Current *ktime*.
5777 *
5778 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5779 *      Description
5780 *              Drain samples from the specified user ring buffer, and invoke
5781 *              the provided callback for each such sample:
5782 *
5783 *              long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5784 *
5785 *              If **callback_fn** returns 0, the helper will continue to try
5786 *              and drain the next sample, up to a maximum of
5787 *              BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5788 *              the helper will skip the rest of the samples and return. Other
5789 *              return values are not used now, and will be rejected by the
5790 *              verifier.
5791 *      Return
5792 *              The number of drained samples if no error was encountered while
5793 *              draining samples, or 0 if no samples were present in the ring
5794 *              buffer. If a user-space producer was epoll-waiting on this map,
5795 *              and at least one sample was drained, they will receive an event
5796 *              notification notifying them of available space in the ring
5797 *              buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5798 *              function, no wakeup notification will be sent. If the
5799 *              BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5800 *              be sent even if no sample was drained.
5801 *
5802 *              On failure, the returned value is one of the following:
5803 *
5804 *              **-EBUSY** if the ring buffer is contended, and another calling
5805 *              context was concurrently draining the ring buffer.
5806 *
5807 *              **-EINVAL** if user-space is not properly tracking the ring
5808 *              buffer due to the producer position not being aligned to 8
5809 *              bytes, a sample not being aligned to 8 bytes, or the producer
5810 *              position not matching the advertised length of a sample.
5811 *
5812 *              **-E2BIG** if user-space has tried to publish a sample which is
5813 *              larger than the size of the ring buffer, or which cannot fit
5814 *              within a struct bpf_dynptr.
5815 *
5816 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5817 *      Description
5818 *              Get a bpf_local_storage from the *cgroup*.
5819 *
5820 *              Logically, it could be thought of as getting the value from
5821 *              a *map* with *cgroup* as the **key**.  From this
5822 *              perspective,  the usage is not much different from
5823 *              **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5824 *              helper enforces the key must be a cgroup struct and the map must also
5825 *              be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5826 *
5827 *              In reality, the local-storage value is embedded directly inside of the
5828 *              *cgroup* object itself, rather than being located in the
5829 *              **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5830 *              queried for some *map* on a *cgroup* object, the kernel will perform an
5831 *              O(n) iteration over all of the live local-storage values for that
5832 *              *cgroup* object until the local-storage value for the *map* is found.
5833 *
5834 *              An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5835 *              used such that a new bpf_local_storage will be
5836 *              created if one does not exist.  *value* can be used
5837 *              together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5838 *              the initial value of a bpf_local_storage.  If *value* is
5839 *              **NULL**, the new bpf_local_storage will be zero initialized.
5840 *      Return
5841 *              A bpf_local_storage pointer is returned on success.
5842 *
5843 *              **NULL** if not found or there was an error in adding
5844 *              a new bpf_local_storage.
5845 *
5846 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5847 *      Description
5848 *              Delete a bpf_local_storage from a *cgroup*.
5849 *      Return
5850 *              0 on success.
5851 *
5852 *              **-ENOENT** if the bpf_local_storage cannot be found.
5853 */
5854#define ___BPF_FUNC_MAPPER(FN, ctx...)                  \
5855        FN(unspec, 0, ##ctx)                            \
5856        FN(map_lookup_elem, 1, ##ctx)                   \
5857        FN(map_update_elem, 2, ##ctx)                   \
5858        FN(map_delete_elem, 3, ##ctx)                   \
5859        FN(probe_read, 4, ##ctx)                        \
5860        FN(ktime_get_ns, 5, ##ctx)                      \
5861        FN(trace_printk, 6, ##ctx)                      \
5862        FN(get_prandom_u32, 7, ##ctx)                   \
5863        FN(get_smp_processor_id, 8, ##ctx)              \
5864        FN(skb_store_bytes, 9, ##ctx)                   \
5865        FN(l3_csum_replace, 10, ##ctx)                  \
5866        FN(l4_csum_replace, 11, ##ctx)                  \
5867        FN(tail_call, 12, ##ctx)                        \
5868        FN(clone_redirect, 13, ##ctx)                   \
5869        FN(get_current_pid_tgid, 14, ##ctx)             \
5870        FN(get_current_uid_gid, 15, ##ctx)              \
5871        FN(get_current_comm, 16, ##ctx)                 \
5872        FN(get_cgroup_classid, 17, ##ctx)               \
5873        FN(skb_vlan_push, 18, ##ctx)                    \
5874        FN(skb_vlan_pop, 19, ##ctx)                     \
5875        FN(skb_get_tunnel_key, 20, ##ctx)               \
5876        FN(skb_set_tunnel_key, 21, ##ctx)               \
5877        FN(perf_event_read, 22, ##ctx)                  \
5878        FN(redirect, 23, ##ctx)                         \
5879        FN(get_route_realm, 24, ##ctx)                  \
5880        FN(perf_event_output, 25, ##ctx)                \
5881        FN(skb_load_bytes, 26, ##ctx)                   \
5882        FN(get_stackid, 27, ##ctx)                      \
5883        FN(csum_diff, 28, ##ctx)                        \
5884        FN(skb_get_tunnel_opt, 29, ##ctx)               \
5885        FN(skb_set_tunnel_opt, 30, ##ctx)               \
5886        FN(skb_change_proto, 31, ##ctx)                 \
5887        FN(skb_change_type, 32, ##ctx)                  \
5888        FN(skb_under_cgroup, 33, ##ctx)                 \
5889        FN(get_hash_recalc, 34, ##ctx)                  \
5890        FN(get_current_task, 35, ##ctx)                 \
5891        FN(probe_write_user, 36, ##ctx)                 \
5892        FN(current_task_under_cgroup, 37, ##ctx)        \
5893        FN(skb_change_tail, 38, ##ctx)                  \
5894        FN(skb_pull_data, 39, ##ctx)                    \
5895        FN(csum_update, 40, ##ctx)                      \
5896        FN(set_hash_invalid, 41, ##ctx)                 \
5897        FN(get_numa_node_id, 42, ##ctx)                 \
5898        FN(skb_change_head, 43, ##ctx)                  \
5899        FN(xdp_adjust_head, 44, ##ctx)                  \
5900        FN(probe_read_str, 45, ##ctx)                   \
5901        FN(get_socket_cookie, 46, ##ctx)                \
5902        FN(get_socket_uid, 47, ##ctx)                   \
5903        FN(set_hash, 48, ##ctx)                         \
5904        FN(setsockopt, 49, ##ctx)                       \
5905        FN(skb_adjust_room, 50, ##ctx)                  \
5906        FN(redirect_map, 51, ##ctx)                     \
5907        FN(sk_redirect_map, 52, ##ctx)                  \
5908        FN(sock_map_update, 53, ##ctx)                  \
5909        FN(xdp_adjust_meta, 54, ##ctx)                  \
5910        FN(perf_event_read_value, 55, ##ctx)            \
5911        FN(perf_prog_read_value, 56, ##ctx)             \
5912        FN(getsockopt, 57, ##ctx)                       \
5913        FN(override_return, 58, ##ctx)                  \
5914        FN(sock_ops_cb_flags_set, 59, ##ctx)            \
5915        FN(msg_redirect_map, 60, ##ctx)                 \
5916        FN(msg_apply_bytes, 61, ##ctx)                  \
5917        FN(msg_cork_bytes, 62, ##ctx)                   \
5918        FN(msg_pull_data, 63, ##ctx)                    \
5919        FN(bind, 64, ##ctx)                             \
5920        FN(xdp_adjust_tail, 65, ##ctx)                  \
5921        FN(skb_get_xfrm_state, 66, ##ctx)               \
5922        FN(get_stack, 67, ##ctx)                        \
5923        FN(skb_load_bytes_relative, 68, ##ctx)          \
5924        FN(fib_lookup, 69, ##ctx)                       \
5925        FN(sock_hash_update, 70, ##ctx)                 \
5926        FN(msg_redirect_hash, 71, ##ctx)                \
5927        FN(sk_redirect_hash, 72, ##ctx)                 \
5928        FN(lwt_push_encap, 73, ##ctx)                   \
5929        FN(lwt_seg6_store_bytes, 74, ##ctx)             \
5930        FN(lwt_seg6_adjust_srh, 75, ##ctx)              \
5931        FN(lwt_seg6_action, 76, ##ctx)                  \
5932        FN(rc_repeat, 77, ##ctx)                        \
5933        FN(rc_keydown, 78, ##ctx)                       \
5934        FN(skb_cgroup_id, 79, ##ctx)                    \
5935        FN(get_current_cgroup_id, 80, ##ctx)            \
5936        FN(get_local_storage, 81, ##ctx)                \
5937        FN(sk_select_reuseport, 82, ##ctx)              \
5938        FN(skb_ancestor_cgroup_id, 83, ##ctx)           \
5939        FN(sk_lookup_tcp, 84, ##ctx)                    \
5940        FN(sk_lookup_udp, 85, ##ctx)                    \
5941        FN(sk_release, 86, ##ctx)                       \
5942        FN(map_push_elem, 87, ##ctx)                    \
5943        FN(map_pop_elem, 88, ##ctx)                     \
5944        FN(map_peek_elem, 89, ##ctx)                    \
5945        FN(msg_push_data, 90, ##ctx)                    \
5946        FN(msg_pop_data, 91, ##ctx)                     \
5947        FN(rc_pointer_rel, 92, ##ctx)                   \
5948        FN(spin_lock, 93, ##ctx)                        \
5949        FN(spin_unlock, 94, ##ctx)                      \
5950        FN(sk_fullsock, 95, ##ctx)                      \
5951        FN(tcp_sock, 96, ##ctx)                         \
5952        FN(skb_ecn_set_ce, 97, ##ctx)                   \
5953        FN(get_listener_sock, 98, ##ctx)                \
5954        FN(skc_lookup_tcp, 99, ##ctx)                   \
5955        FN(tcp_check_syncookie, 100, ##ctx)             \
5956        FN(sysctl_get_name, 101, ##ctx)                 \
5957        FN(sysctl_get_current_value, 102, ##ctx)        \
5958        FN(sysctl_get_new_value, 103, ##ctx)            \
5959        FN(sysctl_set_new_value, 104, ##ctx)            \
5960        FN(strtol, 105, ##ctx)                          \
5961        FN(strtoul, 106, ##ctx)                         \
5962        FN(sk_storage_get, 107, ##ctx)                  \
5963        FN(sk_storage_delete, 108, ##ctx)               \
5964        FN(send_signal, 109, ##ctx)                     \
5965        FN(tcp_gen_syncookie, 110, ##ctx)               \
5966        FN(skb_output, 111, ##ctx)                      \
5967        FN(probe_read_user, 112, ##ctx)                 \
5968        FN(probe_read_kernel, 113, ##ctx)               \
5969        FN(probe_read_user_str, 114, ##ctx)             \
5970        FN(probe_read_kernel_str, 115, ##ctx)           \
5971        FN(tcp_send_ack, 116, ##ctx)                    \
5972        FN(send_signal_thread, 117, ##ctx)              \
5973        FN(jiffies64, 118, ##ctx)                       \
5974        FN(read_branch_records, 119, ##ctx)             \
5975        FN(get_ns_current_pid_tgid, 120, ##ctx)         \
5976        FN(xdp_output, 121, ##ctx)                      \
5977        FN(get_netns_cookie, 122, ##ctx)                \
5978        FN(get_current_ancestor_cgroup_id, 123, ##ctx)  \
5979        FN(sk_assign, 124, ##ctx)                       \
5980        FN(ktime_get_boot_ns, 125, ##ctx)               \
5981        FN(seq_printf, 126, ##ctx)                      \
5982        FN(seq_write, 127, ##ctx)                       \
5983        FN(sk_cgroup_id, 128, ##ctx)                    \
5984        FN(sk_ancestor_cgroup_id, 129, ##ctx)           \
5985        FN(ringbuf_output, 130, ##ctx)                  \
5986        FN(ringbuf_reserve, 131, ##ctx)                 \
5987        FN(ringbuf_submit, 132, ##ctx)                  \
5988        FN(ringbuf_discard, 133, ##ctx)                 \
5989        FN(ringbuf_query, 134, ##ctx)                   \
5990        FN(csum_level, 135, ##ctx)                      \
5991        FN(skc_to_tcp6_sock, 136, ##ctx)                \
5992        FN(skc_to_tcp_sock, 137, ##ctx)                 \
5993        FN(skc_to_tcp_timewait_sock, 138, ##ctx)        \
5994        FN(skc_to_tcp_request_sock, 139, ##ctx)         \
5995        FN(skc_to_udp6_sock, 140, ##ctx)                \
5996        FN(get_task_stack, 141, ##ctx)                  \
5997        FN(load_hdr_opt, 142, ##ctx)                    \
5998        FN(store_hdr_opt, 143, ##ctx)                   \
5999        FN(reserve_hdr_opt, 144, ##ctx)                 \
6000        FN(inode_storage_get, 145, ##ctx)               \
6001        FN(inode_storage_delete, 146, ##ctx)            \
6002        FN(d_path, 147, ##ctx)                          \
6003        FN(copy_from_user, 148, ##ctx)                  \
6004        FN(snprintf_btf, 149, ##ctx)                    \
6005        FN(seq_printf_btf, 150, ##ctx)                  \
6006        FN(skb_cgroup_classid, 151, ##ctx)              \
6007        FN(redirect_neigh, 152, ##ctx)                  \
6008        FN(per_cpu_ptr, 153, ##ctx)                     \
6009        FN(this_cpu_ptr, 154, ##ctx)                    \
6010        FN(redirect_peer, 155, ##ctx)                   \
6011        FN(task_storage_get, 156, ##ctx)                \
6012        FN(task_storage_delete, 157, ##ctx)             \
6013        FN(get_current_task_btf, 158, ##ctx)            \
6014        FN(bprm_opts_set, 159, ##ctx)                   \
6015        FN(ktime_get_coarse_ns, 160, ##ctx)             \
6016        FN(ima_inode_hash, 161, ##ctx)                  \
6017        FN(sock_from_file, 162, ##ctx)                  \
6018        FN(check_mtu, 163, ##ctx)                       \
6019        FN(for_each_map_elem, 164, ##ctx)               \
6020        FN(snprintf, 165, ##ctx)                        \
6021        FN(sys_bpf, 166, ##ctx)                         \
6022        FN(btf_find_by_name_kind, 167, ##ctx)           \
6023        FN(sys_close, 168, ##ctx)                       \
6024        FN(timer_init, 169, ##ctx)                      \
6025        FN(timer_set_callback, 170, ##ctx)              \
6026        FN(timer_start, 171, ##ctx)                     \
6027        FN(timer_cancel, 172, ##ctx)                    \
6028        FN(get_func_ip, 173, ##ctx)                     \
6029        FN(get_attach_cookie, 174, ##ctx)               \
6030        FN(task_pt_regs, 175, ##ctx)                    \
6031        FN(get_branch_snapshot, 176, ##ctx)             \
6032        FN(trace_vprintk, 177, ##ctx)                   \
6033        FN(skc_to_unix_sock, 178, ##ctx)                \
6034        FN(kallsyms_lookup_name, 179, ##ctx)            \
6035        FN(find_vma, 180, ##ctx)                        \
6036        FN(loop, 181, ##ctx)                            \
6037        FN(strncmp, 182, ##ctx)                         \
6038        FN(get_func_arg, 183, ##ctx)                    \
6039        FN(get_func_ret, 184, ##ctx)                    \
6040        FN(get_func_arg_cnt, 185, ##ctx)                \
6041        FN(get_retval, 186, ##ctx)                      \
6042        FN(set_retval, 187, ##ctx)                      \
6043        FN(xdp_get_buff_len, 188, ##ctx)                \
6044        FN(xdp_load_bytes, 189, ##ctx)                  \
6045        FN(xdp_store_bytes, 190, ##ctx)                 \
6046        FN(copy_from_user_task, 191, ##ctx)             \
6047        FN(skb_set_tstamp, 192, ##ctx)                  \
6048        FN(ima_file_hash, 193, ##ctx)                   \
6049        FN(kptr_xchg, 194, ##ctx)                       \
6050        FN(map_lookup_percpu_elem, 195, ##ctx)          \
6051        FN(skc_to_mptcp_sock, 196, ##ctx)               \
6052        FN(dynptr_from_mem, 197, ##ctx)                 \
6053        FN(ringbuf_reserve_dynptr, 198, ##ctx)          \
6054        FN(ringbuf_submit_dynptr, 199, ##ctx)           \
6055        FN(ringbuf_discard_dynptr, 200, ##ctx)          \
6056        FN(dynptr_read, 201, ##ctx)                     \
6057        FN(dynptr_write, 202, ##ctx)                    \
6058        FN(dynptr_data, 203, ##ctx)                     \
6059        FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx)      \
6060        FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx)      \
6061        FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx)    \
6062        FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx)    \
6063        FN(ktime_get_tai_ns, 208, ##ctx)                \
6064        FN(user_ringbuf_drain, 209, ##ctx)              \
6065        FN(cgrp_storage_get, 210, ##ctx)                \
6066        FN(cgrp_storage_delete, 211, ##ctx)             \
6067        /* This helper list is effectively frozen. If you are trying to \
6068         * add a new helper, you should add a kfunc instead which has   \
6069         * less stability guarantees. See Documentation/bpf/kfuncs.rst  \
6070         */
6071
6072/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
6073 * know or care about integer value that is now passed as second argument
6074 */
6075#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
6076#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
6077
6078/* integer value in 'imm' field of BPF_CALL instruction selects which helper
6079 * function eBPF program intends to call
6080 */
6081#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
6082enum bpf_func_id {
6083        ___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
6084        __BPF_FUNC_MAX_ID,
6085};
6086#undef __BPF_ENUM_FN
6087
6088/* All flags used by eBPF helper functions, placed here. */
6089
6090/* BPF_FUNC_skb_store_bytes flags. */
6091enum {
6092        BPF_F_RECOMPUTE_CSUM            = (1ULL << 0),
6093        BPF_F_INVALIDATE_HASH           = (1ULL << 1),
6094};
6095
6096/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
6097 * First 4 bits are for passing the header field size.
6098 */
6099enum {
6100        BPF_F_HDR_FIELD_MASK            = 0xfULL,
6101};
6102
6103/* BPF_FUNC_l4_csum_replace flags. */
6104enum {
6105        BPF_F_PSEUDO_HDR                = (1ULL << 4),
6106        BPF_F_MARK_MANGLED_0            = (1ULL << 5),
6107        BPF_F_MARK_ENFORCE              = (1ULL << 6),
6108        BPF_F_IPV6                      = (1ULL << 7),
6109};
6110
6111/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
6112enum {
6113        BPF_F_TUNINFO_IPV6              = (1ULL << 0),
6114};
6115
6116/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
6117enum {
6118        BPF_F_SKIP_FIELD_MASK           = 0xffULL,
6119        BPF_F_USER_STACK                = (1ULL << 8),
6120/* flags used by BPF_FUNC_get_stackid only. */
6121        BPF_F_FAST_STACK_CMP            = (1ULL << 9),
6122        BPF_F_REUSE_STACKID             = (1ULL << 10),
6123/* flags used by BPF_FUNC_get_stack only. */
6124        BPF_F_USER_BUILD_ID             = (1ULL << 11),
6125};
6126
6127/* BPF_FUNC_skb_set_tunnel_key flags. */
6128enum {
6129        BPF_F_ZERO_CSUM_TX              = (1ULL << 1),
6130        BPF_F_DONT_FRAGMENT             = (1ULL << 2),
6131        BPF_F_SEQ_NUMBER                = (1ULL << 3),
6132        BPF_F_NO_TUNNEL_KEY             = (1ULL << 4),
6133};
6134
6135/* BPF_FUNC_skb_get_tunnel_key flags. */
6136enum {
6137        BPF_F_TUNINFO_FLAGS             = (1ULL << 4),
6138};
6139
6140/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
6141 * BPF_FUNC_perf_event_read_value flags.
6142 */
6143enum {
6144        BPF_F_INDEX_MASK                = 0xffffffffULL,
6145        BPF_F_CURRENT_CPU               = BPF_F_INDEX_MASK,
6146/* BPF_FUNC_perf_event_output for sk_buff input context. */
6147        BPF_F_CTXLEN_MASK               = (0xfffffULL << 32),
6148};
6149
6150/* Current network namespace */
6151enum {
6152        BPF_F_CURRENT_NETNS             = (-1L),
6153};
6154
6155/* BPF_FUNC_csum_level level values. */
6156enum {
6157        BPF_CSUM_LEVEL_QUERY,
6158        BPF_CSUM_LEVEL_INC,
6159        BPF_CSUM_LEVEL_DEC,
6160        BPF_CSUM_LEVEL_RESET,
6161};
6162
6163/* BPF_FUNC_skb_adjust_room flags. */
6164enum {
6165        BPF_F_ADJ_ROOM_FIXED_GSO        = (1ULL << 0),
6166        BPF_F_ADJ_ROOM_ENCAP_L3_IPV4    = (1ULL << 1),
6167        BPF_F_ADJ_ROOM_ENCAP_L3_IPV6    = (1ULL << 2),
6168        BPF_F_ADJ_ROOM_ENCAP_L4_GRE     = (1ULL << 3),
6169        BPF_F_ADJ_ROOM_ENCAP_L4_UDP     = (1ULL << 4),
6170        BPF_F_ADJ_ROOM_NO_CSUM_RESET    = (1ULL << 5),
6171        BPF_F_ADJ_ROOM_ENCAP_L2_ETH     = (1ULL << 6),
6172        BPF_F_ADJ_ROOM_DECAP_L3_IPV4    = (1ULL << 7),
6173        BPF_F_ADJ_ROOM_DECAP_L3_IPV6    = (1ULL << 8),
6174};
6175
6176enum {
6177        BPF_ADJ_ROOM_ENCAP_L2_MASK      = 0xff,
6178        BPF_ADJ_ROOM_ENCAP_L2_SHIFT     = 56,
6179};
6180
6181#define BPF_F_ADJ_ROOM_ENCAP_L2(len)    (((__u64)len & \
6182                                          BPF_ADJ_ROOM_ENCAP_L2_MASK) \
6183                                         << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
6184
6185/* BPF_FUNC_sysctl_get_name flags. */
6186enum {
6187        BPF_F_SYSCTL_BASE_NAME          = (1ULL << 0),
6188};
6189
6190/* BPF_FUNC_<kernel_obj>_storage_get flags */
6191enum {
6192        BPF_LOCAL_STORAGE_GET_F_CREATE  = (1ULL << 0),
6193        /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
6194         * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
6195         */
6196        BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,
6197};
6198
6199/* BPF_FUNC_read_branch_records flags. */
6200enum {
6201        BPF_F_GET_BRANCH_RECORDS_SIZE   = (1ULL << 0),
6202};
6203
6204/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
6205 * BPF_FUNC_bpf_ringbuf_output flags.
6206 */
6207enum {
6208        BPF_RB_NO_WAKEUP                = (1ULL << 0),
6209        BPF_RB_FORCE_WAKEUP             = (1ULL << 1),
6210};
6211
6212/* BPF_FUNC_bpf_ringbuf_query flags */
6213enum {
6214        BPF_RB_AVAIL_DATA = 0,
6215        BPF_RB_RING_SIZE = 1,
6216        BPF_RB_CONS_POS = 2,
6217        BPF_RB_PROD_POS = 3,
6218};
6219
6220/* BPF ring buffer constants */
6221enum {
6222        BPF_RINGBUF_BUSY_BIT            = (1U << 31),
6223        BPF_RINGBUF_DISCARD_BIT         = (1U << 30),
6224        BPF_RINGBUF_HDR_SZ              = 8,
6225};
6226
6227/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
6228enum {
6229        BPF_SK_LOOKUP_F_REPLACE         = (1ULL << 0),
6230        BPF_SK_LOOKUP_F_NO_REUSEPORT    = (1ULL << 1),
6231};
6232
6233/* Mode for BPF_FUNC_skb_adjust_room helper. */
6234enum bpf_adj_room_mode {
6235        BPF_ADJ_ROOM_NET,
6236        BPF_ADJ_ROOM_MAC,
6237};
6238
6239/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
6240enum bpf_hdr_start_off {
6241        BPF_HDR_START_MAC,
6242        BPF_HDR_START_NET,
6243};
6244
6245/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
6246enum bpf_lwt_encap_mode {
6247        BPF_LWT_ENCAP_SEG6,
6248        BPF_LWT_ENCAP_SEG6_INLINE,
6249        BPF_LWT_ENCAP_IP,
6250};
6251
6252/* Flags for bpf_bprm_opts_set helper */
6253enum {
6254        BPF_F_BPRM_SECUREEXEC   = (1ULL << 0),
6255};
6256
6257/* Flags for bpf_redirect and bpf_redirect_map helpers */
6258enum {
6259        BPF_F_INGRESS           = (1ULL << 0), /* used for skb path */
6260        BPF_F_BROADCAST         = (1ULL << 3), /* used for XDP path */
6261        BPF_F_EXCLUDE_INGRESS   = (1ULL << 4), /* used for XDP path */
6262#define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
6263};
6264
6265#define __bpf_md_ptr(type, name)        \
6266union {                                 \
6267        type name;                      \
6268        __u64 :64;                      \
6269} __attribute__((aligned(8)))
6270
6271/* The enum used in skb->tstamp_type. It specifies the clock type
6272 * of the time stored in the skb->tstamp.
6273 */
6274enum {
6275        BPF_SKB_TSTAMP_UNSPEC = 0,              /* DEPRECATED */
6276        BPF_SKB_TSTAMP_DELIVERY_MONO = 1,       /* DEPRECATED */
6277        BPF_SKB_CLOCK_REALTIME = 0,
6278        BPF_SKB_CLOCK_MONOTONIC = 1,
6279        BPF_SKB_CLOCK_TAI = 2,
6280        /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle,
6281         * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid.
6282         */
6283};
6284
6285/* user accessible mirror of in-kernel sk_buff.
6286 * new fields can only be added to the end of this structure
6287 */
6288struct __sk_buff {
6289        __u32 len;
6290        __u32 pkt_type;
6291        __u32 mark;
6292        __u32 queue_mapping;
6293        __u32 protocol;
6294        __u32 vlan_present;
6295        __u32 vlan_tci;
6296        __u32 vlan_proto;
6297        __u32 priority;
6298        __u32 ingress_ifindex;
6299        __u32 ifindex;
6300        __u32 tc_index;
6301        __u32 cb[5];
6302        __u32 hash;
6303        __u32 tc_classid;
6304        __u32 data;
6305        __u32 data_end;
6306        __u32 napi_id;
6307
6308        /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
6309        __u32 family;
6310        __u32 remote_ip4;       /* Stored in network byte order */
6311        __u32 local_ip4;        /* Stored in network byte order */
6312        __u32 remote_ip6[4];    /* Stored in network byte order */
6313        __u32 local_ip6[4];     /* Stored in network byte order */
6314        __u32 remote_port;      /* Stored in network byte order */
6315        __u32 local_port;       /* stored in host byte order */
6316        /* ... here. */
6317
6318        __u32 data_meta;
6319        __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
6320        __u64 tstamp;
6321        __u32 wire_len;
6322        __u32 gso_segs;
6323        __bpf_md_ptr(struct bpf_sock *, sk);
6324        __u32 gso_size;
6325        __u8  tstamp_type;
6326        __u32 :24;              /* Padding, future use. */
6327        __u64 hwtstamp;
6328};
6329
6330struct bpf_tunnel_key {
6331        __u32 tunnel_id;
6332        union {
6333                __u32 remote_ipv4;
6334                __u32 remote_ipv6[4];
6335        };
6336        __u8 tunnel_tos;
6337        __u8 tunnel_ttl;
6338        union {
6339                __u16 tunnel_ext;       /* compat */
6340                __be16 tunnel_flags;
6341        };
6342        __u32 tunnel_label;
6343        union {
6344                __u32 local_ipv4;
6345                __u32 local_ipv6[4];
6346        };
6347};
6348
6349/* user accessible mirror of in-kernel xfrm_state.
6350 * new fields can only be added to the end of this structure
6351 */
6352struct bpf_xfrm_state {
6353        __u32 reqid;
6354        __u32 spi;      /* Stored in network byte order */
6355        __u16 family;
6356        __u16 ext;      /* Padding, future use. */
6357        union {
6358                __u32 remote_ipv4;      /* Stored in network byte order */
6359                __u32 remote_ipv6[4];   /* Stored in network byte order */
6360        };
6361};
6362
6363/* Generic BPF return codes which all BPF program types may support.
6364 * The values are binary compatible with their TC_ACT_* counter-part to
6365 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
6366 * programs.
6367 *
6368 * XDP is handled seprately, see XDP_*.
6369 */
6370enum bpf_ret_code {
6371        BPF_OK = 0,
6372        /* 1 reserved */
6373        BPF_DROP = 2,
6374        /* 3-6 reserved */
6375        BPF_REDIRECT = 7,
6376        /* >127 are reserved for prog type specific return codes.
6377         *
6378         * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6379         *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6380         *    changed and should be routed based on its new L3 header.
6381         *    (This is an L3 redirect, as opposed to L2 redirect
6382         *    represented by BPF_REDIRECT above).
6383         */
6384        BPF_LWT_REROUTE = 128,
6385        /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6386         *   to indicate that no custom dissection was performed, and
6387         *   fallback to standard dissector is requested.
6388         */
6389        BPF_FLOW_DISSECTOR_CONTINUE = 129,
6390};
6391
6392struct bpf_sock {
6393        __u32 bound_dev_if;
6394        __u32 family;
6395        __u32 type;
6396        __u32 protocol;
6397        __u32 mark;
6398        __u32 priority;
6399        /* IP address also allows 1 and 2 bytes access */
6400        __u32 src_ip4;
6401        __u32 src_ip6[4];
6402        __u32 src_port;         /* host byte order */
6403        __be16 dst_port;        /* network byte order */
6404        __u16 :16;              /* zero padding */
6405        __u32 dst_ip4;
6406        __u32 dst_ip6[4];
6407        __u32 state;
6408        __s32 rx_queue_mapping;
6409};
6410
6411struct bpf_tcp_sock {
6412        __u32 snd_cwnd;         /* Sending congestion window            */
6413        __u32 srtt_us;          /* smoothed round trip time << 3 in usecs */
6414        __u32 rtt_min;
6415        __u32 snd_ssthresh;     /* Slow start size threshold            */
6416        __u32 rcv_nxt;          /* What we want to receive next         */
6417        __u32 snd_nxt;          /* Next sequence we send                */
6418        __u32 snd_una;          /* First byte we want an ack for        */
6419        __u32 mss_cache;        /* Cached effective mss, not including SACKS */
6420        __u32 ecn_flags;        /* ECN status bits.                     */
6421        __u32 rate_delivered;   /* saved rate sample: packets delivered */
6422        __u32 rate_interval_us; /* saved rate sample: time elapsed */
6423        __u32 packets_out;      /* Packets which are "in flight"        */
6424        __u32 retrans_out;      /* Retransmitted packets out            */
6425        __u32 total_retrans;    /* Total retransmits for entire connection */
6426        __u32 segs_in;          /* RFC4898 tcpEStatsPerfSegsIn
6427                                 * total number of segments in.
6428                                 */
6429        __u32 data_segs_in;     /* RFC4898 tcpEStatsPerfDataSegsIn
6430                                 * total number of data segments in.
6431                                 */
6432        __u32 segs_out;         /* RFC4898 tcpEStatsPerfSegsOut
6433                                 * The total number of segments sent.
6434                                 */
6435        __u32 data_segs_out;    /* RFC4898 tcpEStatsPerfDataSegsOut
6436                                 * total number of data segments sent.
6437                                 */
6438        __u32 lost_out;         /* Lost packets                 */
6439        __u32 sacked_out;       /* SACK'd packets                       */
6440        __u64 bytes_received;   /* RFC4898 tcpEStatsAppHCThruOctetsReceived
6441                                 * sum(delta(rcv_nxt)), or how many bytes
6442                                 * were acked.
6443                                 */
6444        __u64 bytes_acked;      /* RFC4898 tcpEStatsAppHCThruOctetsAcked
6445                                 * sum(delta(snd_una)), or how many bytes
6446                                 * were acked.
6447                                 */
6448        __u32 dsack_dups;       /* RFC4898 tcpEStatsStackDSACKDups
6449                                 * total number of DSACK blocks received
6450                                 */
6451        __u32 delivered;        /* Total data packets delivered incl. rexmits */
6452        __u32 delivered_ce;     /* Like the above but only ECE marked packets */
6453        __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
6454};
6455
6456struct bpf_sock_tuple {
6457        union {
6458                struct {
6459                        __be32 saddr;
6460                        __be32 daddr;
6461                        __be16 sport;
6462                        __be16 dport;
6463                } ipv4;
6464                struct {
6465                        __be32 saddr[4];
6466                        __be32 daddr[4];
6467                        __be16 sport;
6468                        __be16 dport;
6469                } ipv6;
6470        };
6471};
6472
6473/* (Simplified) user return codes for tcx prog type.
6474 * A valid tcx program must return one of these defined values. All other
6475 * return codes are reserved for future use. Must remain compatible with
6476 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
6477 * return codes are mapped to TCX_NEXT.
6478 */
6479enum tcx_action_base {
6480        TCX_NEXT        = -1,
6481        TCX_PASS        = 0,
6482        TCX_DROP        = 2,
6483        TCX_REDIRECT    = 7,
6484};
6485
6486struct bpf_xdp_sock {
6487        __u32 queue_id;
6488};
6489
6490#define XDP_PACKET_HEADROOM 256
6491
6492/* User return codes for XDP prog type.
6493 * A valid XDP program must return one of these defined values. All other
6494 * return codes are reserved for future use. Unknown return codes will
6495 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6496 */
6497enum xdp_action {
6498        XDP_ABORTED = 0,
6499        XDP_DROP,
6500        XDP_PASS,
6501        XDP_TX,
6502        XDP_REDIRECT,
6503};
6504
6505/* user accessible metadata for XDP packet hook
6506 * new fields must be added to the end of this structure
6507 */
6508struct xdp_md {
6509        __u32 data;
6510        __u32 data_end;
6511        __u32 data_meta;
6512        /* Below access go through struct xdp_rxq_info */
6513        __u32 ingress_ifindex; /* rxq->dev->ifindex */
6514        __u32 rx_queue_index;  /* rxq->queue_index  */
6515
6516        __u32 egress_ifindex;  /* txq->dev->ifindex */
6517};
6518
6519/* DEVMAP map-value layout
6520 *
6521 * The struct data-layout of map-value is a configuration interface.
6522 * New members can only be added to the end of this structure.
6523 */
6524struct bpf_devmap_val {
6525        __u32 ifindex;   /* device index */
6526        union {
6527                int   fd;  /* prog fd on map write */
6528                __u32 id;  /* prog id on map read */
6529        } bpf_prog;
6530};
6531
6532/* CPUMAP map-value layout
6533 *
6534 * The struct data-layout of map-value is a configuration interface.
6535 * New members can only be added to the end of this structure.
6536 */
6537struct bpf_cpumap_val {
6538        __u32 qsize;    /* queue size to remote target CPU */
6539        union {
6540                int   fd;       /* prog fd on map write */
6541                __u32 id;       /* prog id on map read */
6542        } bpf_prog;
6543};
6544
6545enum sk_action {
6546        SK_DROP = 0,
6547        SK_PASS,
6548};
6549
6550/* user accessible metadata for SK_MSG packet hook, new fields must
6551 * be added to the end of this structure
6552 */
6553struct sk_msg_md {
6554        __bpf_md_ptr(void *, data);
6555        __bpf_md_ptr(void *, data_end);
6556
6557        __u32 family;
6558        __u32 remote_ip4;       /* Stored in network byte order */
6559        __u32 local_ip4;        /* Stored in network byte order */
6560        __u32 remote_ip6[4];    /* Stored in network byte order */
6561        __u32 local_ip6[4];     /* Stored in network byte order */
6562        __u32 remote_port;      /* Stored in network byte order */
6563        __u32 local_port;       /* stored in host byte order */
6564        __u32 size;             /* Total size of sk_msg */
6565
6566        __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6567};
6568
6569struct sk_reuseport_md {
6570        /*
6571         * Start of directly accessible data. It begins from
6572         * the tcp/udp header.
6573         */
6574        __bpf_md_ptr(void *, data);
6575        /* End of directly accessible data */
6576        __bpf_md_ptr(void *, data_end);
6577        /*
6578         * Total length of packet (starting from the tcp/udp header).
6579         * Note that the directly accessible bytes (data_end - data)
6580         * could be less than this "len".  Those bytes could be
6581         * indirectly read by a helper "bpf_skb_load_bytes()".
6582         */
6583        __u32 len;
6584        /*
6585         * Eth protocol in the mac header (network byte order). e.g.
6586         * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6587         */
6588        __u32 eth_protocol;
6589        __u32 ip_protocol;      /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6590        __u32 bind_inany;       /* Is sock bound to an INANY address? */
6591        __u32 hash;             /* A hash of the packet 4 tuples */
6592        /* When reuse->migrating_sk is NULL, it is selecting a sk for the
6593         * new incoming connection request (e.g. selecting a listen sk for
6594         * the received SYN in the TCP case).  reuse->sk is one of the sk
6595         * in the reuseport group. The bpf prog can use reuse->sk to learn
6596         * the local listening ip/port without looking into the skb.
6597         *
6598         * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6599         * reuse->migrating_sk is the socket that needs to be migrated
6600         * to another listening socket.  migrating_sk could be a fullsock
6601         * sk that is fully established or a reqsk that is in-the-middle
6602         * of 3-way handshake.
6603         */
6604        __bpf_md_ptr(struct bpf_sock *, sk);
6605        __bpf_md_ptr(struct bpf_sock *, migrating_sk);
6606};
6607
6608#define BPF_TAG_SIZE    8
6609
6610struct bpf_prog_info {
6611        __u32 type;
6612        __u32 id;
6613        __u8  tag[BPF_TAG_SIZE];
6614        __u32 jited_prog_len;
6615        __u32 xlated_prog_len;
6616        __aligned_u64 jited_prog_insns;
6617        __aligned_u64 xlated_prog_insns;
6618        __u64 load_time;        /* ns since boottime */
6619        __u32 created_by_uid;
6620        __u32 nr_map_ids;
6621        __aligned_u64 map_ids;
6622        char name[BPF_OBJ_NAME_LEN];
6623        __u32 ifindex;
6624        __u32 gpl_compatible:1;
6625        __u32 :31; /* alignment pad */
6626        __u64 netns_dev;
6627        __u64 netns_ino;
6628        __u32 nr_jited_ksyms;
6629        __u32 nr_jited_func_lens;
6630        __aligned_u64 jited_ksyms;
6631        __aligned_u64 jited_func_lens;
6632        __u32 btf_id;
6633        __u32 func_info_rec_size;
6634        __aligned_u64 func_info;
6635        __u32 nr_func_info;
6636        __u32 nr_line_info;
6637        __aligned_u64 line_info;
6638        __aligned_u64 jited_line_info;
6639        __u32 nr_jited_line_info;
6640        __u32 line_info_rec_size;
6641        __u32 jited_line_info_rec_size;
6642        __u32 nr_prog_tags;
6643        __aligned_u64 prog_tags;
6644        __u64 run_time_ns;
6645        __u64 run_cnt;
6646        __u64 recursion_misses;
6647        __u32 verified_insns;
6648        __u32 attach_btf_obj_id;
6649        __u32 attach_btf_id;
6650} __attribute__((aligned(8)));
6651
6652struct bpf_map_info {
6653        __u32 type;
6654        __u32 id;
6655        __u32 key_size;
6656        __u32 value_size;
6657        __u32 max_entries;
6658        __u32 map_flags;
6659        char  name[BPF_OBJ_NAME_LEN];
6660        __u32 ifindex;
6661        __u32 btf_vmlinux_value_type_id;
6662        __u64 netns_dev;
6663        __u64 netns_ino;
6664        __u32 btf_id;
6665        __u32 btf_key_type_id;
6666        __u32 btf_value_type_id;
6667        __u32 btf_vmlinux_id;
6668        __u64 map_extra;
6669} __attribute__((aligned(8)));
6670
6671struct bpf_btf_info {
6672        __aligned_u64 btf;
6673        __u32 btf_size;
6674        __u32 id;
6675        __aligned_u64 name;
6676        __u32 name_len;
6677        __u32 kernel_btf;
6678} __attribute__((aligned(8)));
6679
6680struct bpf_link_info {
6681        __u32 type;
6682        __u32 id;
6683        __u32 prog_id;
6684        union {
6685                struct {
6686                        __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6687                        __u32 tp_name_len;     /* in/out: tp_name buffer len */
6688                        __u32 :32;
6689                        __u64 cookie;
6690                } raw_tracepoint;
6691                struct {
6692                        __u32 attach_type;
6693                        __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6694                        __u32 target_btf_id; /* BTF type id inside the object */
6695                        __u32 :32;
6696                        __u64 cookie;
6697                } tracing;
6698                struct {
6699                        __u64 cgroup_id;
6700                        __u32 attach_type;
6701                } cgroup;
6702                struct {
6703                        __aligned_u64 target_name; /* in/out: target_name buffer ptr */
6704                        __u32 target_name_len;     /* in/out: target_name buffer len */
6705
6706                        /* If the iter specific field is 32 bits, it can be put
6707                         * in the first or second union. Otherwise it should be
6708                         * put in the second union.
6709                         */
6710                        union {
6711                                struct {
6712                                        __u32 map_id;
6713                                } map;
6714                        };
6715                        union {
6716                                struct {
6717                                        __u64 cgroup_id;
6718                                        __u32 order;
6719                                } cgroup;
6720                                struct {
6721                                        __u32 tid;
6722                                        __u32 pid;
6723                                } task;
6724                        };
6725                } iter;
6726                struct  {
6727                        __u32 netns_ino;
6728                        __u32 attach_type;
6729                } netns;
6730                struct {
6731                        __u32 ifindex;
6732                } xdp;
6733                struct {
6734                        __u32 map_id;
6735                } struct_ops;
6736                struct {
6737                        __u32 pf;
6738                        __u32 hooknum;
6739                        __s32 priority;
6740                        __u32 flags;
6741                } netfilter;
6742                struct {
6743                        __aligned_u64 addrs;
6744                        __u32 count; /* in/out: kprobe_multi function count */
6745                        __u32 flags;
6746                        __u64 missed;
6747                        __aligned_u64 cookies;
6748                } kprobe_multi;
6749                struct {
6750                        __aligned_u64 path;
6751                        __aligned_u64 offsets;
6752                        __aligned_u64 ref_ctr_offsets;
6753                        __aligned_u64 cookies;
6754                        __u32 path_size; /* in/out: real path size on success, including zero byte */
6755                        __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */
6756                        __u32 flags;
6757                        __u32 pid;
6758                } uprobe_multi;
6759                struct {
6760                        __u32 type; /* enum bpf_perf_event_type */
6761                        __u32 :32;
6762                        union {
6763                                struct {
6764                                        __aligned_u64 file_name; /* in/out */
6765                                        __u32 name_len;
6766                                        __u32 offset; /* offset from file_name */
6767                                        __u64 cookie;
6768                                        __u64 ref_ctr_offset;
6769                                } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
6770                                struct {
6771                                        __aligned_u64 func_name; /* in/out */
6772                                        __u32 name_len;
6773                                        __u32 offset; /* offset from func_name */
6774                                        __u64 addr;
6775                                        __u64 missed;
6776                                        __u64 cookie;
6777                                } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
6778                                struct {
6779                                        __aligned_u64 tp_name;   /* in/out */
6780                                        __u32 name_len;
6781                                        __u32 :32;
6782                                        __u64 cookie;
6783                                } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
6784                                struct {
6785                                        __u64 config;
6786                                        __u32 type;
6787                                        __u32 :32;
6788                                        __u64 cookie;
6789                                } event; /* BPF_PERF_EVENT_EVENT */
6790                        };
6791                } perf_event;
6792                struct {
6793                        __u32 ifindex;
6794                        __u32 attach_type;
6795                } tcx;
6796                struct {
6797                        __u32 ifindex;
6798                        __u32 attach_type;
6799                } netkit;
6800                struct {
6801                        __u32 map_id;
6802                        __u32 attach_type;
6803                } sockmap;
6804        };
6805} __attribute__((aligned(8)));
6806
6807struct bpf_token_info {
6808        __u64 allowed_cmds;
6809        __u64 allowed_maps;
6810        __u64 allowed_progs;
6811        __u64 allowed_attachs;
6812} __attribute__((aligned(8)));
6813
6814/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6815 * by user and intended to be used by socket (e.g. to bind to, depends on
6816 * attach type).
6817 */
6818struct bpf_sock_addr {
6819        __u32 user_family;      /* Allows 4-byte read, but no write. */
6820        __u32 user_ip4;         /* Allows 1,2,4-byte read and 4-byte write.
6821                                 * Stored in network byte order.
6822                                 */
6823        __u32 user_ip6[4];      /* Allows 1,2,4,8-byte read and 4,8-byte write.
6824                                 * Stored in network byte order.
6825                                 */
6826        __u32 user_port;        /* Allows 1,2,4-byte read and 4-byte write.
6827                                 * Stored in network byte order
6828                                 */
6829        __u32 family;           /* Allows 4-byte read, but no write */
6830        __u32 type;             /* Allows 4-byte read, but no write */
6831        __u32 protocol;         /* Allows 4-byte read, but no write */
6832        __u32 msg_src_ip4;      /* Allows 1,2,4-byte read and 4-byte write.
6833                                 * Stored in network byte order.
6834                                 */
6835        __u32 msg_src_ip6[4];   /* Allows 1,2,4,8-byte read and 4,8-byte write.
6836                                 * Stored in network byte order.
6837                                 */
6838        __bpf_md_ptr(struct bpf_sock *, sk);
6839};
6840
6841/* User bpf_sock_ops struct to access socket values and specify request ops
6842 * and their replies.
6843 * Some of this fields are in network (bigendian) byte order and may need
6844 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6845 * New fields can only be added at the end of this structure
6846 */
6847struct bpf_sock_ops {
6848        __u32 op;
6849        union {
6850                __u32 args[4];          /* Optionally passed to bpf program */
6851                __u32 reply;            /* Returned by bpf program          */
6852                __u32 replylong[4];     /* Optionally returned by bpf prog  */
6853        };
6854        __u32 family;
6855        __u32 remote_ip4;       /* Stored in network byte order */
6856        __u32 local_ip4;        /* Stored in network byte order */
6857        __u32 remote_ip6[4];    /* Stored in network byte order */
6858        __u32 local_ip6[4];     /* Stored in network byte order */
6859        __u32 remote_port;      /* Stored in network byte order */
6860        __u32 local_port;       /* stored in host byte order */
6861        __u32 is_fullsock;      /* Some TCP fields are only valid if
6862                                 * there is a full socket. If not, the
6863                                 * fields read as zero.
6864                                 */
6865        __u32 snd_cwnd;
6866        __u32 srtt_us;          /* Averaged RTT << 3 in usecs */
6867        __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6868        __u32 state;
6869        __u32 rtt_min;
6870        __u32 snd_ssthresh;
6871        __u32 rcv_nxt;
6872        __u32 snd_nxt;
6873        __u32 snd_una;
6874        __u32 mss_cache;
6875        __u32 ecn_flags;
6876        __u32 rate_delivered;
6877        __u32 rate_interval_us;
6878        __u32 packets_out;
6879        __u32 retrans_out;
6880        __u32 total_retrans;
6881        __u32 segs_in;
6882        __u32 data_segs_in;
6883        __u32 segs_out;
6884        __u32 data_segs_out;
6885        __u32 lost_out;
6886        __u32 sacked_out;
6887        __u32 sk_txhash;
6888        __u64 bytes_received;
6889        __u64 bytes_acked;
6890        __bpf_md_ptr(struct bpf_sock *, sk);
6891        /* [skb_data, skb_data_end) covers the whole TCP header.
6892         *
6893         * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6894         * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the
6895         *                                header has not been written.
6896         * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6897         *                                been written so far.
6898         * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes
6899         *                                      the 3WHS.
6900         * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6901         *                                      the 3WHS.
6902         *
6903         * bpf_load_hdr_opt() can also be used to read a particular option.
6904         */
6905        __bpf_md_ptr(void *, skb_data);
6906        __bpf_md_ptr(void *, skb_data_end);
6907        __u32 skb_len;          /* The total length of a packet.
6908                                 * It includes the header, options,
6909                                 * and payload.
6910                                 */
6911        __u32 skb_tcp_flags;    /* tcp_flags of the header.  It provides
6912                                 * an easy way to check for tcp_flags
6913                                 * without parsing skb_data.
6914                                 *
6915                                 * In particular, the skb_tcp_flags
6916                                 * will still be available in
6917                                 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6918                                 * the outgoing header has not
6919                                 * been written yet.
6920                                 */
6921        __u64 skb_hwtstamp;
6922};
6923
6924/* Definitions for bpf_sock_ops_cb_flags */
6925enum {
6926        BPF_SOCK_OPS_RTO_CB_FLAG        = (1<<0),
6927        BPF_SOCK_OPS_RETRANS_CB_FLAG    = (1<<1),
6928        BPF_SOCK_OPS_STATE_CB_FLAG      = (1<<2),
6929        BPF_SOCK_OPS_RTT_CB_FLAG        = (1<<3),
6930        /* Call bpf for all received TCP headers.  The bpf prog will be
6931         * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6932         *
6933         * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6934         * for the header option related helpers that will be useful
6935         * to the bpf programs.
6936         *
6937         * It could be used at the client/active side (i.e. connect() side)
6938         * when the server told it that the server was in syncookie
6939         * mode and required the active side to resend the bpf-written
6940         * options.  The active side can keep writing the bpf-options until
6941         * it received a valid packet from the server side to confirm
6942         * the earlier packet (and options) has been received.  The later
6943         * example patch is using it like this at the active side when the
6944         * server is in syncookie mode.
6945         *
6946         * The bpf prog will usually turn this off in the common cases.
6947         */
6948        BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG  = (1<<4),
6949        /* Call bpf when kernel has received a header option that
6950         * the kernel cannot handle.  The bpf prog will be called under
6951         * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6952         *
6953         * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6954         * for the header option related helpers that will be useful
6955         * to the bpf programs.
6956         */
6957        BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6958        /* Call bpf when the kernel is writing header options for the
6959         * outgoing packet.  The bpf prog will first be called
6960         * to reserve space in a skb under
6961         * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then
6962         * the bpf prog will be called to write the header option(s)
6963         * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6964         *
6965         * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6966         * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6967         * related helpers that will be useful to the bpf programs.
6968         *
6969         * The kernel gets its chance to reserve space and write
6970         * options first before the BPF program does.
6971         */
6972        BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6973/* Mask of all currently supported cb flags */
6974        BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,
6975};
6976
6977enum {
6978        SK_BPF_CB_TX_TIMESTAMPING       = 1<<0,
6979        SK_BPF_CB_MASK                  = (SK_BPF_CB_TX_TIMESTAMPING - 1) |
6980                                           SK_BPF_CB_TX_TIMESTAMPING
6981};
6982
6983/* List of known BPF sock_ops operators.
6984 * New entries can only be added at the end
6985 */
6986enum {
6987        BPF_SOCK_OPS_VOID,
6988        BPF_SOCK_OPS_TIMEOUT_INIT,      /* Should return SYN-RTO value to use or
6989                                         * -1 if default value should be used
6990                                         */
6991        BPF_SOCK_OPS_RWND_INIT,         /* Should return initial advertized
6992                                         * window (in packets) or -1 if default
6993                                         * value should be used
6994                                         */
6995        BPF_SOCK_OPS_TCP_CONNECT_CB,    /* Calls BPF program right before an
6996                                         * active connection is initialized
6997                                         */
6998        BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,     /* Calls BPF program when an
6999                                                 * active connection is
7000                                                 * established
7001                                                 */
7002        BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,    /* Calls BPF program when a
7003                                                 * passive connection is
7004                                                 * established
7005                                                 */
7006        BPF_SOCK_OPS_NEEDS_ECN,         /* If connection's congestion control
7007                                         * needs ECN
7008                                         */
7009        BPF_SOCK_OPS_BASE_RTT,          /* Get base RTT. The correct value is
7010                                         * based on the path and may be
7011                                         * dependent on the congestion control
7012                                         * algorithm. In general it indicates
7013                                         * a congestion threshold. RTTs above
7014                                         * this indicate congestion
7015                                         */
7016        BPF_SOCK_OPS_RTO_CB,            /* Called when an RTO has triggered.
7017                                         * Arg1: value of icsk_retransmits
7018                                         * Arg2: value of icsk_rto
7019                                         * Arg3: whether RTO has expired
7020                                         */
7021        BPF_SOCK_OPS_RETRANS_CB,        /* Called when skb is retransmitted.
7022                                         * Arg1: sequence number of 1st byte
7023                                         * Arg2: # segments
7024                                         * Arg3: return value of
7025                                         *       tcp_transmit_skb (0 => success)
7026                                         */
7027        BPF_SOCK_OPS_STATE_CB,          /* Called when TCP changes state.
7028                                         * Arg1: old_state
7029                                         * Arg2: new_state
7030                                         */
7031        BPF_SOCK_OPS_TCP_LISTEN_CB,     /* Called on listen(2), right after
7032                                         * socket transition to LISTEN state.
7033                                         */
7034        BPF_SOCK_OPS_RTT_CB,            /* Called on every RTT.
7035                                         * Arg1: measured RTT input (mrtt)
7036                                         * Arg2: updated srtt
7037                                         */
7038        BPF_SOCK_OPS_PARSE_HDR_OPT_CB,  /* Parse the header option.
7039                                         * It will be called to handle
7040                                         * the packets received at
7041                                         * an already established
7042                                         * connection.
7043                                         *
7044                                         * sock_ops->skb_data:
7045                                         * Referring to the received skb.
7046                                         * It covers the TCP header only.
7047                                         *
7048                                         * bpf_load_hdr_opt() can also
7049                                         * be used to search for a
7050                                         * particular option.
7051                                         */
7052        BPF_SOCK_OPS_HDR_OPT_LEN_CB,    /* Reserve space for writing the
7053                                         * header option later in
7054                                         * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7055                                         * Arg1: bool want_cookie. (in
7056                                         *       writing SYNACK only)
7057                                         *
7058                                         * sock_ops->skb_data:
7059                                         * Not available because no header has
7060                                         * been written yet.
7061                                         *
7062                                         * sock_ops->skb_tcp_flags:
7063                                         * The tcp_flags of the
7064                                         * outgoing skb. (e.g. SYN, ACK, FIN).
7065                                         *
7066                                         * bpf_reserve_hdr_opt() should
7067                                         * be used to reserve space.
7068                                         */
7069        BPF_SOCK_OPS_WRITE_HDR_OPT_CB,  /* Write the header options
7070                                         * Arg1: bool want_cookie. (in
7071                                         *       writing SYNACK only)
7072                                         *
7073                                         * sock_ops->skb_data:
7074                                         * Referring to the outgoing skb.
7075                                         * It covers the TCP header
7076                                         * that has already been written
7077                                         * by the kernel and the
7078                                         * earlier bpf-progs.
7079                                         *
7080                                         * sock_ops->skb_tcp_flags:
7081                                         * The tcp_flags of the outgoing
7082                                         * skb. (e.g. SYN, ACK, FIN).
7083                                         *
7084                                         * bpf_store_hdr_opt() should
7085                                         * be used to write the
7086                                         * option.
7087                                         *
7088                                         * bpf_load_hdr_opt() can also
7089                                         * be used to search for a
7090                                         * particular option that
7091                                         * has already been written
7092                                         * by the kernel or the
7093                                         * earlier bpf-progs.
7094                                         */
7095        BPF_SOCK_OPS_TSTAMP_SCHED_CB,   /* Called when skb is passing
7096                                         * through dev layer when
7097                                         * SK_BPF_CB_TX_TIMESTAMPING
7098                                         * feature is on.
7099                                         */
7100        BPF_SOCK_OPS_TSTAMP_SND_SW_CB,  /* Called when skb is about to send
7101                                         * to the nic when SK_BPF_CB_TX_TIMESTAMPING
7102                                         * feature is on.
7103                                         */
7104        BPF_SOCK_OPS_TSTAMP_SND_HW_CB,  /* Called in hardware phase when
7105                                         * SK_BPF_CB_TX_TIMESTAMPING feature
7106                                         * is on.
7107                                         */
7108        BPF_SOCK_OPS_TSTAMP_ACK_CB,     /* Called when all the skbs in the
7109                                         * same sendmsg call are acked
7110                                         * when SK_BPF_CB_TX_TIMESTAMPING
7111                                         * feature is on.
7112                                         */
7113        BPF_SOCK_OPS_TSTAMP_SENDMSG_CB, /* Called when every sendmsg syscall
7114                                         * is triggered. It's used to correlate
7115                                         * sendmsg timestamp with corresponding
7116                                         * tskey.
7117                                         */
7118};
7119
7120/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
7121 * changes between the TCP and BPF versions. Ideally this should never happen.
7122 * If it does, we need to add code to convert them before calling
7123 * the BPF sock_ops function.
7124 */
7125enum {
7126        BPF_TCP_ESTABLISHED = 1,
7127        BPF_TCP_SYN_SENT,
7128        BPF_TCP_SYN_RECV,
7129        BPF_TCP_FIN_WAIT1,
7130        BPF_TCP_FIN_WAIT2,
7131        BPF_TCP_TIME_WAIT,
7132        BPF_TCP_CLOSE,
7133        BPF_TCP_CLOSE_WAIT,
7134        BPF_TCP_LAST_ACK,
7135        BPF_TCP_LISTEN,
7136        BPF_TCP_CLOSING,        /* Now a valid state */
7137        BPF_TCP_NEW_SYN_RECV,
7138        BPF_TCP_BOUND_INACTIVE,
7139
7140        BPF_TCP_MAX_STATES      /* Leave at the end! */
7141};
7142
7143enum {
7144        TCP_BPF_IW              = 1001, /* Set TCP initial congestion window */
7145        TCP_BPF_SNDCWND_CLAMP   = 1002, /* Set sndcwnd_clamp */
7146        TCP_BPF_DELACK_MAX      = 1003, /* Max delay ack in usecs */
7147        TCP_BPF_RTO_MIN         = 1004, /* Min delay ack in usecs */
7148        /* Copy the SYN pkt to optval
7149         *
7150         * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the
7151         * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
7152         * to only getting from the saved_syn.  It can either get the
7153         * syn packet from:
7154         *
7155         * 1. the just-received SYN packet (only available when writing the
7156         *    SYNACK).  It will be useful when it is not necessary to
7157         *    save the SYN packet for latter use.  It is also the only way
7158         *    to get the SYN during syncookie mode because the syn
7159         *    packet cannot be saved during syncookie.
7160         *
7161         * OR
7162         *
7163         * 2. the earlier saved syn which was done by
7164         *    bpf_setsockopt(TCP_SAVE_SYN).
7165         *
7166         * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
7167         * SYN packet is obtained.
7168         *
7169         * If the bpf-prog does not need the IP[46] header,  the
7170         * bpf-prog can avoid parsing the IP header by using
7171         * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both
7172         * IP[46] and TCP header by using TCP_BPF_SYN_IP.
7173         *
7174         *      >0: Total number of bytes copied
7175         * -ENOSPC: Not enough space in optval. Only optlen number of
7176         *          bytes is copied.
7177         * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
7178         *          is not saved by setsockopt(TCP_SAVE_SYN).
7179         */
7180        TCP_BPF_SYN             = 1005, /* Copy the TCP header */
7181        TCP_BPF_SYN_IP          = 1006, /* Copy the IP[46] and TCP header */
7182        TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */
7183        TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */
7184        SK_BPF_CB_FLAGS         = 1009, /* Get or set sock ops flags in socket */
7185};
7186
7187enum {
7188        BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
7189};
7190
7191/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
7192 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7193 */
7194enum {
7195        BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,      /* Kernel is finding the
7196                                                 * total option spaces
7197                                                 * required for an established
7198                                                 * sk in order to calculate the
7199                                                 * MSS.  No skb is actually
7200                                                 * sent.
7201                                                 */
7202        BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,    /* Kernel is in syncookie mode
7203                                                 * when sending a SYN.
7204                                                 */
7205};
7206
7207struct bpf_perf_event_value {
7208        __u64 counter;
7209        __u64 enabled;
7210        __u64 running;
7211};
7212
7213enum {
7214        BPF_DEVCG_ACC_MKNOD     = (1ULL << 0),
7215        BPF_DEVCG_ACC_READ      = (1ULL << 1),
7216        BPF_DEVCG_ACC_WRITE     = (1ULL << 2),
7217};
7218
7219enum {
7220        BPF_DEVCG_DEV_BLOCK     = (1ULL << 0),
7221        BPF_DEVCG_DEV_CHAR      = (1ULL << 1),
7222};
7223
7224struct bpf_cgroup_dev_ctx {
7225        /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
7226        __u32 access_type;
7227        __u32 major;
7228        __u32 minor;
7229};
7230
7231struct bpf_raw_tracepoint_args {
7232        __u64 args[0];
7233};
7234
7235/* DIRECT:  Skip the FIB rules and go to FIB table associated with device
7236 * OUTPUT:  Do lookup from egress perspective; default is ingress
7237 */
7238enum {
7239        BPF_FIB_LOOKUP_DIRECT  = (1U << 0),
7240        BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),
7241        BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
7242        BPF_FIB_LOOKUP_TBID    = (1U << 3),
7243        BPF_FIB_LOOKUP_SRC     = (1U << 4),
7244        BPF_FIB_LOOKUP_MARK    = (1U << 5),
7245};
7246
7247enum {
7248        BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */
7249        BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */
7250        BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */
7251        BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */
7252        BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */
7253        BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
7254        BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */
7255        BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */
7256        BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
7257        BPF_FIB_LKUP_RET_NO_SRC_ADDR,  /* failed to derive IP src addr */
7258};
7259
7260struct bpf_fib_lookup {
7261        /* input:  network family for lookup (AF_INET, AF_INET6)
7262         * output: network family of egress nexthop
7263         */
7264        __u8    family;
7265
7266        /* set if lookup is to consider L4 data - e.g., FIB rules */
7267        __u8    l4_protocol;
7268        __be16  sport;
7269        __be16  dport;
7270
7271        union { /* used for MTU check */
7272                /* input to lookup */
7273                __u16   tot_len; /* L3 length from network hdr (iph->tot_len) */
7274
7275                /* output: MTU value */
7276                __u16   mtu_result;
7277        } __attribute__((packed, aligned(2)));
7278        /* input: L3 device index for lookup
7279         * output: device index from FIB lookup
7280         */
7281        __u32   ifindex;
7282
7283        union {
7284                /* inputs to lookup */
7285                __u8    tos;            /* AF_INET  */
7286                __be32  flowinfo;       /* AF_INET6, flow_label + priority */
7287
7288                /* output: metric of fib result (IPv4/IPv6 only) */
7289                __u32   rt_metric;
7290        };
7291
7292        /* input: source address to consider for lookup
7293         * output: source address result from lookup
7294         */
7295        union {
7296                __be32          ipv4_src;
7297                __u32           ipv6_src[4];  /* in6_addr; network order */
7298        };
7299
7300        /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
7301         * network header. output: bpf_fib_lookup sets to gateway address
7302         * if FIB lookup returns gateway route
7303         */
7304        union {
7305                __be32          ipv4_dst;
7306                __u32           ipv6_dst[4];  /* in6_addr; network order */
7307        };
7308
7309        union {
7310                struct {
7311                        /* output */
7312                        __be16  h_vlan_proto;
7313                        __be16  h_vlan_TCI;
7314                };
7315                /* input: when accompanied with the
7316                 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
7317                 * specific routing table to use for the fib lookup.
7318                 */
7319                __u32   tbid;
7320        };
7321
7322        union {
7323                /* input */
7324                struct {
7325                        __u32   mark;   /* policy routing */
7326                        /* 2 4-byte holes for input */
7327                };
7328
7329                /* output: source and dest mac */
7330                struct {
7331                        __u8    smac[6];        /* ETH_ALEN */
7332                        __u8    dmac[6];        /* ETH_ALEN */
7333                };
7334        };
7335};
7336
7337struct bpf_redir_neigh {
7338        /* network family for lookup (AF_INET, AF_INET6) */
7339        __u32 nh_family;
7340        /* network address of nexthop; skips fib lookup to find gateway */
7341        union {
7342                __be32          ipv4_nh;
7343                __u32           ipv6_nh[4];  /* in6_addr; network order */
7344        };
7345};
7346
7347/* bpf_check_mtu flags*/
7348enum  bpf_check_mtu_flags {
7349        BPF_MTU_CHK_SEGS  = (1U << 0),
7350};
7351
7352enum bpf_check_mtu_ret {
7353        BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */
7354        BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
7355        BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */
7356};
7357
7358enum bpf_task_fd_type {
7359        BPF_FD_TYPE_RAW_TRACEPOINT,     /* tp name */
7360        BPF_FD_TYPE_TRACEPOINT,         /* tp name */
7361        BPF_FD_TYPE_KPROBE,             /* (symbol + offset) or addr */
7362        BPF_FD_TYPE_KRETPROBE,          /* (symbol + offset) or addr */
7363        BPF_FD_TYPE_UPROBE,             /* filename + offset */
7364        BPF_FD_TYPE_URETPROBE,          /* filename + offset */
7365};
7366
7367enum {
7368        BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG             = (1U << 0),
7369        BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL         = (1U << 1),
7370        BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP              = (1U << 2),
7371};
7372
7373struct bpf_flow_keys {
7374        __u16   nhoff;
7375        __u16   thoff;
7376        __u16   addr_proto;                     /* ETH_P_* of valid addrs */
7377        __u8    is_frag;
7378        __u8    is_first_frag;
7379        __u8    is_encap;
7380        __u8    ip_proto;
7381        __be16  n_proto;
7382        __be16  sport;
7383        __be16  dport;
7384        union {
7385                struct {
7386                        __be32  ipv4_src;
7387                        __be32  ipv4_dst;
7388                };
7389                struct {
7390                        __u32   ipv6_src[4];    /* in6_addr; network order */
7391                        __u32   ipv6_dst[4];    /* in6_addr; network order */
7392                };
7393        };
7394        __u32   flags;
7395        __be32  flow_label;
7396};
7397
7398struct bpf_func_info {
7399        __u32   insn_off;
7400        __u32   type_id;
7401};
7402
7403#define BPF_LINE_INFO_LINE_NUM(line_col)        ((line_col) >> 10)
7404#define BPF_LINE_INFO_LINE_COL(line_col)        ((line_col) & 0x3ff)
7405
7406struct bpf_line_info {
7407        __u32   insn_off;
7408        __u32   file_name_off;
7409        __u32   line_off;
7410        __u32   line_col;
7411};
7412
7413struct bpf_spin_lock {
7414        __u32   val;
7415};
7416
7417struct bpf_timer {
7418        __u64 __opaque[2];
7419} __attribute__((aligned(8)));
7420
7421struct bpf_wq {
7422        __u64 __opaque[2];
7423} __attribute__((aligned(8)));
7424
7425struct bpf_dynptr {
7426        __u64 __opaque[2];
7427} __attribute__((aligned(8)));
7428
7429struct bpf_list_head {
7430        __u64 __opaque[2];
7431} __attribute__((aligned(8)));
7432
7433struct bpf_list_node {
7434        __u64 __opaque[3];
7435} __attribute__((aligned(8)));
7436
7437struct bpf_rb_root {
7438        __u64 __opaque[2];
7439} __attribute__((aligned(8)));
7440
7441struct bpf_rb_node {
7442        __u64 __opaque[4];
7443} __attribute__((aligned(8)));
7444
7445struct bpf_refcount {
7446        __u32 __opaque[1];
7447} __attribute__((aligned(4)));
7448
7449struct bpf_sysctl {
7450        __u32   write;          /* Sysctl is being read (= 0) or written (= 1).
7451                                 * Allows 1,2,4-byte read, but no write.
7452                                 */
7453        __u32   file_pos;       /* Sysctl file position to read from, write to.
7454                                 * Allows 1,2,4-byte read an 4-byte write.
7455                                 */
7456};
7457
7458struct bpf_sockopt {
7459        __bpf_md_ptr(struct bpf_sock *, sk);
7460        __bpf_md_ptr(void *, optval);
7461        __bpf_md_ptr(void *, optval_end);
7462
7463        __s32   level;
7464        __s32   optname;
7465        __s32   optlen;
7466        __s32   retval;
7467};
7468
7469struct bpf_pidns_info {
7470        __u32 pid;
7471        __u32 tgid;
7472};
7473
7474/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
7475struct bpf_sk_lookup {
7476        union {
7477                __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
7478                __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
7479        };
7480
7481        __u32 family;           /* Protocol family (AF_INET, AF_INET6) */
7482        __u32 protocol;         /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
7483        __u32 remote_ip4;       /* Network byte order */
7484        __u32 remote_ip6[4];    /* Network byte order */
7485        __be16 remote_port;     /* Network byte order */
7486        __u16 :16;              /* Zero padding */
7487        __u32 local_ip4;        /* Network byte order */
7488        __u32 local_ip6[4];     /* Network byte order */
7489        __u32 local_port;       /* Host byte order */
7490        __u32 ingress_ifindex;          /* The arriving interface. Determined by inet_iif. */
7491};
7492
7493/*
7494 * struct btf_ptr is used for typed pointer representation; the
7495 * type id is used to render the pointer data as the appropriate type
7496 * via the bpf_snprintf_btf() helper described above.  A flags field -
7497 * potentially to specify additional details about the BTF pointer
7498 * (rather than its mode of display) - is included for future use.
7499 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
7500 */
7501struct btf_ptr {
7502        void *ptr;
7503        __u32 type_id;
7504        __u32 flags;            /* BTF ptr flags; unused at present. */
7505};
7506
7507/*
7508 * Flags to control bpf_snprintf_btf() behaviour.
7509 *     - BTF_F_COMPACT: no formatting around type information
7510 *     - BTF_F_NONAME: no struct/union member names/types
7511 *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
7512 *       equivalent to %px.
7513 *     - BTF_F_ZERO: show zero-valued struct/union members; they
7514 *       are not displayed by default
7515 */
7516enum {
7517        BTF_F_COMPACT   =       (1ULL << 0),
7518        BTF_F_NONAME    =       (1ULL << 1),
7519        BTF_F_PTR_RAW   =       (1ULL << 2),
7520        BTF_F_ZERO      =       (1ULL << 3),
7521};
7522
7523/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
7524 * has to be adjusted by relocations. It is emitted by llvm and passed to
7525 * libbpf and later to the kernel.
7526 */
7527enum bpf_core_relo_kind {
7528        BPF_CORE_FIELD_BYTE_OFFSET = 0,      /* field byte offset */
7529        BPF_CORE_FIELD_BYTE_SIZE = 1,        /* field size in bytes */
7530        BPF_CORE_FIELD_EXISTS = 2,           /* field existence in target kernel */
7531        BPF_CORE_FIELD_SIGNED = 3,           /* field signedness (0 - unsigned, 1 - signed) */
7532        BPF_CORE_FIELD_LSHIFT_U64 = 4,       /* bitfield-specific left bitshift */
7533        BPF_CORE_FIELD_RSHIFT_U64 = 5,       /* bitfield-specific right bitshift */
7534        BPF_CORE_TYPE_ID_LOCAL = 6,          /* type ID in local BPF object */
7535        BPF_CORE_TYPE_ID_TARGET = 7,         /* type ID in target kernel */
7536        BPF_CORE_TYPE_EXISTS = 8,            /* type existence in target kernel */
7537        BPF_CORE_TYPE_SIZE = 9,              /* type size in bytes */
7538        BPF_CORE_ENUMVAL_EXISTS = 10,        /* enum value existence in target kernel */
7539        BPF_CORE_ENUMVAL_VALUE = 11,         /* enum value integer value */
7540        BPF_CORE_TYPE_MATCHES = 12,          /* type match in target kernel */
7541};
7542
7543/*
7544 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7545 * and from libbpf to the kernel.
7546 *
7547 * CO-RE relocation captures the following data:
7548 * - insn_off - instruction offset (in bytes) within a BPF program that needs
7549 *   its insn->imm field to be relocated with actual field info;
7550 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7551 *   type or field;
7552 * - access_str_off - offset into corresponding .BTF string section. String
7553 *   interpretation depends on specific relocation kind:
7554 *     - for field-based relocations, string encodes an accessed field using
7555 *       a sequence of field and array indices, separated by colon (:). It's
7556 *       conceptually very close to LLVM's getelementptr ([0]) instruction's
7557 *       arguments for identifying offset to a field.
7558 *     - for type-based relocations, strings is expected to be just "0";
7559 *     - for enum value-based relocations, string contains an index of enum
7560 *       value within its enum type;
7561 * - kind - one of enum bpf_core_relo_kind;
7562 *
7563 * Example:
7564 *   struct sample {
7565 *       int a;
7566 *       struct {
7567 *           int b[10];
7568 *       };
7569 *   };
7570 *
7571 *   struct sample *s = ...;
7572 *   int *x = &s->a;     // encoded as "0:0" (a is field #0)
7573 *   int *y = &s->b[5];  // encoded as "0:1:0:5" (anon struct is field #1,
7574 *                       // b is field #0 inside anon struct, accessing elem #5)
7575 *   int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7576 *
7577 * type_id for all relocs in this example will capture BTF type id of
7578 * `struct sample`.
7579 *
7580 * Such relocation is emitted when using __builtin_preserve_access_index()
7581 * Clang built-in, passing expression that captures field address, e.g.:
7582 *
7583 * bpf_probe_read(&dst, sizeof(dst),
7584 *                __builtin_preserve_access_index(&src->a.b.c));
7585 *
7586 * In this case Clang will emit field relocation recording necessary data to
7587 * be able to find offset of embedded `a.b.c` field within `src` struct.
7588 *
7589 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7590 */
7591struct bpf_core_relo {
7592        __u32 insn_off;
7593        __u32 type_id;
7594        __u32 access_str_off;
7595        enum bpf_core_relo_kind kind;
7596};
7597
7598/*
7599 * Flags to control bpf_timer_start() behaviour.
7600 *     - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
7601 *       relative to current time.
7602 *     - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.
7603 */
7604enum {
7605        BPF_F_TIMER_ABS = (1ULL << 0),
7606        BPF_F_TIMER_CPU_PIN = (1ULL << 1),
7607};
7608
7609/* BPF numbers iterator state */
7610struct bpf_iter_num {
7611        /* opaque iterator state; having __u64 here allows to preserve correct
7612         * alignment requirements in vmlinux.h, generated from BTF
7613         */
7614        __u64 __opaque[1];
7615} __attribute__((aligned(8)));
7616
7617/*
7618 * Flags to control BPF kfunc behaviour.
7619 *     - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective
7620 *       helper documentation for details.)
7621 */
7622enum bpf_kfunc_flags {
7623        BPF_F_PAD_ZEROS = (1ULL << 0),
7624};
7625
7626#endif /* __LINUX_BPF_H__ */
7627