1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_MEMSX 0x80 /* load with sign extension */ 23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 25 26 /* alu/jmp fields */ 27 #define BPF_MOV 0xb0 /* mov reg to reg */ 28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 29 30 /* change endianness of a register */ 31 #define BPF_END 0xd0 /* flags for endianness conversion: */ 32 #define BPF_TO_LE 0x00 /* convert to little-endian */ 33 #define BPF_TO_BE 0x08 /* convert to big-endian */ 34 #define BPF_FROM_LE BPF_TO_LE 35 #define BPF_FROM_BE BPF_TO_BE 36 37 /* jmp encodings */ 38 #define BPF_JNE 0x50 /* jump != */ 39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 45 #define BPF_JCOND 0xe0 /* conditional pseudo jumps: may_goto, goto_or_nop */ 46 #define BPF_CALL 0x80 /* function call */ 47 #define BPF_EXIT 0x90 /* function return */ 48 49 /* atomic op type fields (stored in immediate) */ 50 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 51 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 52 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 53 54 enum bpf_cond_pseudo_jmp { 55 BPF_MAY_GOTO = 0, 56 }; 57 58 /* Register numbers */ 59 enum { 60 BPF_REG_0 = 0, 61 BPF_REG_1, 62 BPF_REG_2, 63 BPF_REG_3, 64 BPF_REG_4, 65 BPF_REG_5, 66 BPF_REG_6, 67 BPF_REG_7, 68 BPF_REG_8, 69 BPF_REG_9, 70 BPF_REG_10, 71 __MAX_BPF_REG, 72 }; 73 74 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 75 #define MAX_BPF_REG __MAX_BPF_REG 76 77 struct bpf_insn { 78 __u8 code; /* opcode */ 79 __u8 dst_reg:4; /* dest register */ 80 __u8 src_reg:4; /* source register */ 81 __s16 off; /* signed offset */ 82 __s32 imm; /* signed immediate constant */ 83 }; 84 85 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 86 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 87 * the trailing flexible array member) instead. 88 */ 89 struct bpf_lpm_trie_key { 90 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 91 __u8 data[0]; /* Arbitrary size */ 92 }; 93 94 /* Header for bpf_lpm_trie_key structs */ 95 struct bpf_lpm_trie_key_hdr { 96 __u32 prefixlen; 97 }; 98 99 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 100 struct bpf_lpm_trie_key_u8 { 101 union { 102 struct bpf_lpm_trie_key_hdr hdr; 103 __u32 prefixlen; 104 }; 105 __u8 data[]; /* Arbitrary size */ 106 }; 107 108 struct bpf_cgroup_storage_key { 109 __u64 cgroup_inode_id; /* cgroup inode id */ 110 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 111 }; 112 113 enum bpf_cgroup_iter_order { 114 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 115 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 116 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 117 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 118 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 119 }; 120 121 union bpf_iter_link_info { 122 struct { 123 __u32 map_fd; 124 } map; 125 struct { 126 enum bpf_cgroup_iter_order order; 127 128 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 129 * both are zero, the walk starts from the default cgroup v2 130 * root. For walking v1 hierarchy, one should always explicitly 131 * specify cgroup_fd. 132 */ 133 __u32 cgroup_fd; 134 __u64 cgroup_id; 135 } cgroup; 136 /* Parameters of task iterators. */ 137 struct { 138 __u32 tid; 139 __u32 pid; 140 __u32 pid_fd; 141 } task; 142 }; 143 144 /* BPF syscall commands, see bpf(2) man-page for more details. */ 145 /** 146 * DOC: eBPF Syscall Preamble 147 * 148 * The operation to be performed by the **bpf**\ () system call is determined 149 * by the *cmd* argument. Each operation takes an accompanying argument, 150 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 151 * below). The size argument is the size of the union pointed to by *attr*. 152 */ 153 /** 154 * DOC: eBPF Syscall Commands 155 * 156 * BPF_MAP_CREATE 157 * Description 158 * Create a map and return a file descriptor that refers to the 159 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 160 * is automatically enabled for the new file descriptor. 161 * 162 * Applying **close**\ (2) to the file descriptor returned by 163 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 164 * 165 * Return 166 * A new file descriptor (a nonnegative integer), or -1 if an 167 * error occurred (in which case, *errno* is set appropriately). 168 * 169 * BPF_MAP_LOOKUP_ELEM 170 * Description 171 * Look up an element with a given *key* in the map referred to 172 * by the file descriptor *map_fd*. 173 * 174 * The *flags* argument may be specified as one of the 175 * following: 176 * 177 * **BPF_F_LOCK** 178 * Look up the value of a spin-locked map without 179 * returning the lock. This must be specified if the 180 * elements contain a spinlock. 181 * 182 * Return 183 * Returns zero on success. On error, -1 is returned and *errno* 184 * is set appropriately. 185 * 186 * BPF_MAP_UPDATE_ELEM 187 * Description 188 * Create or update an element (key/value pair) in a specified map. 189 * 190 * The *flags* argument should be specified as one of the 191 * following: 192 * 193 * **BPF_ANY** 194 * Create a new element or update an existing element. 195 * **BPF_NOEXIST** 196 * Create a new element only if it did not exist. 197 * **BPF_EXIST** 198 * Update an existing element. 199 * **BPF_F_LOCK** 200 * Update a spin_lock-ed map element. 201 * 202 * Return 203 * Returns zero on success. On error, -1 is returned and *errno* 204 * is set appropriately. 205 * 206 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 207 * **E2BIG**, **EEXIST**, or **ENOENT**. 208 * 209 * **E2BIG** 210 * The number of elements in the map reached the 211 * *max_entries* limit specified at map creation time. 212 * **EEXIST** 213 * If *flags* specifies **BPF_NOEXIST** and the element 214 * with *key* already exists in the map. 215 * **ENOENT** 216 * If *flags* specifies **BPF_EXIST** and the element with 217 * *key* does not exist in the map. 218 * 219 * BPF_MAP_DELETE_ELEM 220 * Description 221 * Look up and delete an element by key in a specified map. 222 * 223 * Return 224 * Returns zero on success. On error, -1 is returned and *errno* 225 * is set appropriately. 226 * 227 * BPF_MAP_GET_NEXT_KEY 228 * Description 229 * Look up an element by key in a specified map and return the key 230 * of the next element. Can be used to iterate over all elements 231 * in the map. 232 * 233 * Return 234 * Returns zero on success. On error, -1 is returned and *errno* 235 * is set appropriately. 236 * 237 * The following cases can be used to iterate over all elements of 238 * the map: 239 * 240 * * If *key* is not found, the operation returns zero and sets 241 * the *next_key* pointer to the key of the first element. 242 * * If *key* is found, the operation returns zero and sets the 243 * *next_key* pointer to the key of the next element. 244 * * If *key* is the last element, returns -1 and *errno* is set 245 * to **ENOENT**. 246 * 247 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 248 * **EINVAL** on error. 249 * 250 * BPF_PROG_LOAD 251 * Description 252 * Verify and load an eBPF program, returning a new file 253 * descriptor associated with the program. 254 * 255 * Applying **close**\ (2) to the file descriptor returned by 256 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 257 * 258 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 259 * automatically enabled for the new file descriptor. 260 * 261 * Return 262 * A new file descriptor (a nonnegative integer), or -1 if an 263 * error occurred (in which case, *errno* is set appropriately). 264 * 265 * BPF_OBJ_PIN 266 * Description 267 * Pin an eBPF program or map referred by the specified *bpf_fd* 268 * to the provided *pathname* on the filesystem. 269 * 270 * The *pathname* argument must not contain a dot ("."). 271 * 272 * On success, *pathname* retains a reference to the eBPF object, 273 * preventing deallocation of the object when the original 274 * *bpf_fd* is closed. This allow the eBPF object to live beyond 275 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 276 * process. 277 * 278 * Applying **unlink**\ (2) or similar calls to the *pathname* 279 * unpins the object from the filesystem, removing the reference. 280 * If no other file descriptors or filesystem nodes refer to the 281 * same object, it will be deallocated (see NOTES). 282 * 283 * The filesystem type for the parent directory of *pathname* must 284 * be **BPF_FS_MAGIC**. 285 * 286 * Return 287 * Returns zero on success. On error, -1 is returned and *errno* 288 * is set appropriately. 289 * 290 * BPF_OBJ_GET 291 * Description 292 * Open a file descriptor for the eBPF object pinned to the 293 * specified *pathname*. 294 * 295 * Return 296 * A new file descriptor (a nonnegative integer), or -1 if an 297 * error occurred (in which case, *errno* is set appropriately). 298 * 299 * BPF_PROG_ATTACH 300 * Description 301 * Attach an eBPF program to a *target_fd* at the specified 302 * *attach_type* hook. 303 * 304 * The *attach_type* specifies the eBPF attachment point to 305 * attach the program to, and must be one of *bpf_attach_type* 306 * (see below). 307 * 308 * The *attach_bpf_fd* must be a valid file descriptor for a 309 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 310 * or sock_ops type corresponding to the specified *attach_type*. 311 * 312 * The *target_fd* must be a valid file descriptor for a kernel 313 * object which depends on the attach type of *attach_bpf_fd*: 314 * 315 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 316 * **BPF_PROG_TYPE_CGROUP_SKB**, 317 * **BPF_PROG_TYPE_CGROUP_SOCK**, 318 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 319 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 320 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 321 * **BPF_PROG_TYPE_SOCK_OPS** 322 * 323 * Control Group v2 hierarchy with the eBPF controller 324 * enabled. Requires the kernel to be compiled with 325 * **CONFIG_CGROUP_BPF**. 326 * 327 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 328 * 329 * Network namespace (eg /proc/self/ns/net). 330 * 331 * **BPF_PROG_TYPE_LIRC_MODE2** 332 * 333 * LIRC device path (eg /dev/lircN). Requires the kernel 334 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 335 * 336 * **BPF_PROG_TYPE_SK_SKB**, 337 * **BPF_PROG_TYPE_SK_MSG** 338 * 339 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 340 * 341 * Return 342 * Returns zero on success. On error, -1 is returned and *errno* 343 * is set appropriately. 344 * 345 * BPF_PROG_DETACH 346 * Description 347 * Detach the eBPF program associated with the *target_fd* at the 348 * hook specified by *attach_type*. The program must have been 349 * previously attached using **BPF_PROG_ATTACH**. 350 * 351 * Return 352 * Returns zero on success. On error, -1 is returned and *errno* 353 * is set appropriately. 354 * 355 * BPF_PROG_TEST_RUN 356 * Description 357 * Run the eBPF program associated with the *prog_fd* a *repeat* 358 * number of times against a provided program context *ctx_in* and 359 * data *data_in*, and return the modified program context 360 * *ctx_out*, *data_out* (for example, packet data), result of the 361 * execution *retval*, and *duration* of the test run. 362 * 363 * The sizes of the buffers provided as input and output 364 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 365 * be provided in the corresponding variables *ctx_size_in*, 366 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 367 * of these parameters are not provided (ie set to NULL), the 368 * corresponding size field must be zero. 369 * 370 * Some program types have particular requirements: 371 * 372 * **BPF_PROG_TYPE_SK_LOOKUP** 373 * *data_in* and *data_out* must be NULL. 374 * 375 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 376 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 377 * 378 * *ctx_out*, *data_in* and *data_out* must be NULL. 379 * *repeat* must be zero. 380 * 381 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 382 * 383 * Return 384 * Returns zero on success. On error, -1 is returned and *errno* 385 * is set appropriately. 386 * 387 * **ENOSPC** 388 * Either *data_size_out* or *ctx_size_out* is too small. 389 * **ENOTSUPP** 390 * This command is not supported by the program type of 391 * the program referred to by *prog_fd*. 392 * 393 * BPF_PROG_GET_NEXT_ID 394 * Description 395 * Fetch the next eBPF program currently loaded into the kernel. 396 * 397 * Looks for the eBPF program with an id greater than *start_id* 398 * and updates *next_id* on success. If no other eBPF programs 399 * remain with ids higher than *start_id*, returns -1 and sets 400 * *errno* to **ENOENT**. 401 * 402 * Return 403 * Returns zero on success. On error, or when no id remains, -1 404 * is returned and *errno* is set appropriately. 405 * 406 * BPF_MAP_GET_NEXT_ID 407 * Description 408 * Fetch the next eBPF map currently loaded into the kernel. 409 * 410 * Looks for the eBPF map with an id greater than *start_id* 411 * and updates *next_id* on success. If no other eBPF maps 412 * remain with ids higher than *start_id*, returns -1 and sets 413 * *errno* to **ENOENT**. 414 * 415 * Return 416 * Returns zero on success. On error, or when no id remains, -1 417 * is returned and *errno* is set appropriately. 418 * 419 * BPF_PROG_GET_FD_BY_ID 420 * Description 421 * Open a file descriptor for the eBPF program corresponding to 422 * *prog_id*. 423 * 424 * Return 425 * A new file descriptor (a nonnegative integer), or -1 if an 426 * error occurred (in which case, *errno* is set appropriately). 427 * 428 * BPF_MAP_GET_FD_BY_ID 429 * Description 430 * Open a file descriptor for the eBPF map corresponding to 431 * *map_id*. 432 * 433 * Return 434 * A new file descriptor (a nonnegative integer), or -1 if an 435 * error occurred (in which case, *errno* is set appropriately). 436 * 437 * BPF_OBJ_GET_INFO_BY_FD 438 * Description 439 * Obtain information about the eBPF object corresponding to 440 * *bpf_fd*. 441 * 442 * Populates up to *info_len* bytes of *info*, which will be in 443 * one of the following formats depending on the eBPF object type 444 * of *bpf_fd*: 445 * 446 * * **struct bpf_prog_info** 447 * * **struct bpf_map_info** 448 * * **struct bpf_btf_info** 449 * * **struct bpf_link_info** 450 * 451 * Return 452 * Returns zero on success. On error, -1 is returned and *errno* 453 * is set appropriately. 454 * 455 * BPF_PROG_QUERY 456 * Description 457 * Obtain information about eBPF programs associated with the 458 * specified *attach_type* hook. 459 * 460 * The *target_fd* must be a valid file descriptor for a kernel 461 * object which depends on the attach type of *attach_bpf_fd*: 462 * 463 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 464 * **BPF_PROG_TYPE_CGROUP_SKB**, 465 * **BPF_PROG_TYPE_CGROUP_SOCK**, 466 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 467 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 468 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 469 * **BPF_PROG_TYPE_SOCK_OPS** 470 * 471 * Control Group v2 hierarchy with the eBPF controller 472 * enabled. Requires the kernel to be compiled with 473 * **CONFIG_CGROUP_BPF**. 474 * 475 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 476 * 477 * Network namespace (eg /proc/self/ns/net). 478 * 479 * **BPF_PROG_TYPE_LIRC_MODE2** 480 * 481 * LIRC device path (eg /dev/lircN). Requires the kernel 482 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 483 * 484 * **BPF_PROG_QUERY** always fetches the number of programs 485 * attached and the *attach_flags* which were used to attach those 486 * programs. Additionally, if *prog_ids* is nonzero and the number 487 * of attached programs is less than *prog_cnt*, populates 488 * *prog_ids* with the eBPF program ids of the programs attached 489 * at *target_fd*. 490 * 491 * The following flags may alter the result: 492 * 493 * **BPF_F_QUERY_EFFECTIVE** 494 * Only return information regarding programs which are 495 * currently effective at the specified *target_fd*. 496 * 497 * Return 498 * Returns zero on success. On error, -1 is returned and *errno* 499 * is set appropriately. 500 * 501 * BPF_RAW_TRACEPOINT_OPEN 502 * Description 503 * Attach an eBPF program to a tracepoint *name* to access kernel 504 * internal arguments of the tracepoint in their raw form. 505 * 506 * The *prog_fd* must be a valid file descriptor associated with 507 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 508 * 509 * No ABI guarantees are made about the content of tracepoint 510 * arguments exposed to the corresponding eBPF program. 511 * 512 * Applying **close**\ (2) to the file descriptor returned by 513 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 514 * 515 * Return 516 * A new file descriptor (a nonnegative integer), or -1 if an 517 * error occurred (in which case, *errno* is set appropriately). 518 * 519 * BPF_BTF_LOAD 520 * Description 521 * Verify and load BPF Type Format (BTF) metadata into the kernel, 522 * returning a new file descriptor associated with the metadata. 523 * BTF is described in more detail at 524 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 525 * 526 * The *btf* parameter must point to valid memory providing 527 * *btf_size* bytes of BTF binary metadata. 528 * 529 * The returned file descriptor can be passed to other **bpf**\ () 530 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 531 * associate the BTF with those objects. 532 * 533 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 534 * parameters to specify a *btf_log_buf*, *btf_log_size* and 535 * *btf_log_level* which allow the kernel to return freeform log 536 * output regarding the BTF verification process. 537 * 538 * Return 539 * A new file descriptor (a nonnegative integer), or -1 if an 540 * error occurred (in which case, *errno* is set appropriately). 541 * 542 * BPF_BTF_GET_FD_BY_ID 543 * Description 544 * Open a file descriptor for the BPF Type Format (BTF) 545 * corresponding to *btf_id*. 546 * 547 * Return 548 * A new file descriptor (a nonnegative integer), or -1 if an 549 * error occurred (in which case, *errno* is set appropriately). 550 * 551 * BPF_TASK_FD_QUERY 552 * Description 553 * Obtain information about eBPF programs associated with the 554 * target process identified by *pid* and *fd*. 555 * 556 * If the *pid* and *fd* are associated with a tracepoint, kprobe 557 * or uprobe perf event, then the *prog_id* and *fd_type* will 558 * be populated with the eBPF program id and file descriptor type 559 * of type **bpf_task_fd_type**. If associated with a kprobe or 560 * uprobe, the *probe_offset* and *probe_addr* will also be 561 * populated. Optionally, if *buf* is provided, then up to 562 * *buf_len* bytes of *buf* will be populated with the name of 563 * the tracepoint, kprobe or uprobe. 564 * 565 * The resulting *prog_id* may be introspected in deeper detail 566 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 567 * 568 * Return 569 * Returns zero on success. On error, -1 is returned and *errno* 570 * is set appropriately. 571 * 572 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 573 * Description 574 * Look up an element with the given *key* in the map referred to 575 * by the file descriptor *fd*, and if found, delete the element. 576 * 577 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 578 * types, the *flags* argument needs to be set to 0, but for other 579 * map types, it may be specified as: 580 * 581 * **BPF_F_LOCK** 582 * Look up and delete the value of a spin-locked map 583 * without returning the lock. This must be specified if 584 * the elements contain a spinlock. 585 * 586 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 587 * implement this command as a "pop" operation, deleting the top 588 * element rather than one corresponding to *key*. 589 * The *key* and *key_len* parameters should be zeroed when 590 * issuing this operation for these map types. 591 * 592 * This command is only valid for the following map types: 593 * * **BPF_MAP_TYPE_QUEUE** 594 * * **BPF_MAP_TYPE_STACK** 595 * * **BPF_MAP_TYPE_HASH** 596 * * **BPF_MAP_TYPE_PERCPU_HASH** 597 * * **BPF_MAP_TYPE_LRU_HASH** 598 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 599 * 600 * Return 601 * Returns zero on success. On error, -1 is returned and *errno* 602 * is set appropriately. 603 * 604 * BPF_MAP_FREEZE 605 * Description 606 * Freeze the permissions of the specified map. 607 * 608 * Write permissions may be frozen by passing zero *flags*. 609 * Upon success, no future syscall invocations may alter the 610 * map state of *map_fd*. Write operations from eBPF programs 611 * are still possible for a frozen map. 612 * 613 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 614 * 615 * Return 616 * Returns zero on success. On error, -1 is returned and *errno* 617 * is set appropriately. 618 * 619 * BPF_BTF_GET_NEXT_ID 620 * Description 621 * Fetch the next BPF Type Format (BTF) object currently loaded 622 * into the kernel. 623 * 624 * Looks for the BTF object with an id greater than *start_id* 625 * and updates *next_id* on success. If no other BTF objects 626 * remain with ids higher than *start_id*, returns -1 and sets 627 * *errno* to **ENOENT**. 628 * 629 * Return 630 * Returns zero on success. On error, or when no id remains, -1 631 * is returned and *errno* is set appropriately. 632 * 633 * BPF_MAP_LOOKUP_BATCH 634 * Description 635 * Iterate and fetch multiple elements in a map. 636 * 637 * Two opaque values are used to manage batch operations, 638 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 639 * to NULL to begin the batched operation. After each subsequent 640 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 641 * *out_batch* as the *in_batch* for the next operation to 642 * continue iteration from the current point. Both *in_batch* and 643 * *out_batch* must point to memory large enough to hold a key, 644 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH, 645 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters 646 * must be at least 4 bytes wide regardless of key size. 647 * 648 * The *keys* and *values* are output parameters which must point 649 * to memory large enough to hold *count* items based on the key 650 * and value size of the map *map_fd*. The *keys* buffer must be 651 * of *key_size* * *count*. The *values* buffer must be of 652 * *value_size* * *count*. 653 * 654 * The *elem_flags* argument may be specified as one of the 655 * following: 656 * 657 * **BPF_F_LOCK** 658 * Look up the value of a spin-locked map without 659 * returning the lock. This must be specified if the 660 * elements contain a spinlock. 661 * 662 * On success, *count* elements from the map are copied into the 663 * user buffer, with the keys copied into *keys* and the values 664 * copied into the corresponding indices in *values*. 665 * 666 * If an error is returned and *errno* is not **EFAULT**, *count* 667 * is set to the number of successfully processed elements. 668 * 669 * Return 670 * Returns zero on success. On error, -1 is returned and *errno* 671 * is set appropriately. 672 * 673 * May set *errno* to **ENOSPC** to indicate that *keys* or 674 * *values* is too small to dump an entire bucket during 675 * iteration of a hash-based map type. 676 * 677 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 678 * Description 679 * Iterate and delete all elements in a map. 680 * 681 * This operation has the same behavior as 682 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 683 * 684 * * Every element that is successfully returned is also deleted 685 * from the map. This is at least *count* elements. Note that 686 * *count* is both an input and an output parameter. 687 * * Upon returning with *errno* set to **EFAULT**, up to 688 * *count* elements may be deleted without returning the keys 689 * and values of the deleted elements. 690 * 691 * Return 692 * Returns zero on success. On error, -1 is returned and *errno* 693 * is set appropriately. 694 * 695 * BPF_MAP_UPDATE_BATCH 696 * Description 697 * Update multiple elements in a map by *key*. 698 * 699 * The *keys* and *values* are input parameters which must point 700 * to memory large enough to hold *count* items based on the key 701 * and value size of the map *map_fd*. The *keys* buffer must be 702 * of *key_size* * *count*. The *values* buffer must be of 703 * *value_size* * *count*. 704 * 705 * Each element specified in *keys* is sequentially updated to the 706 * value in the corresponding index in *values*. The *in_batch* 707 * and *out_batch* parameters are ignored and should be zeroed. 708 * 709 * The *elem_flags* argument should be specified as one of the 710 * following: 711 * 712 * **BPF_ANY** 713 * Create new elements or update a existing elements. 714 * **BPF_NOEXIST** 715 * Create new elements only if they do not exist. 716 * **BPF_EXIST** 717 * Update existing elements. 718 * **BPF_F_LOCK** 719 * Update spin_lock-ed map elements. This must be 720 * specified if the map value contains a spinlock. 721 * 722 * On success, *count* elements from the map are updated. 723 * 724 * If an error is returned and *errno* is not **EFAULT**, *count* 725 * is set to the number of successfully processed elements. 726 * 727 * Return 728 * Returns zero on success. On error, -1 is returned and *errno* 729 * is set appropriately. 730 * 731 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 732 * **E2BIG**. **E2BIG** indicates that the number of elements in 733 * the map reached the *max_entries* limit specified at map 734 * creation time. 735 * 736 * May set *errno* to one of the following error codes under 737 * specific circumstances: 738 * 739 * **EEXIST** 740 * If *flags* specifies **BPF_NOEXIST** and the element 741 * with *key* already exists in the map. 742 * **ENOENT** 743 * If *flags* specifies **BPF_EXIST** and the element with 744 * *key* does not exist in the map. 745 * 746 * BPF_MAP_DELETE_BATCH 747 * Description 748 * Delete multiple elements in a map by *key*. 749 * 750 * The *keys* parameter is an input parameter which must point 751 * to memory large enough to hold *count* items based on the key 752 * size of the map *map_fd*, that is, *key_size* * *count*. 753 * 754 * Each element specified in *keys* is sequentially deleted. The 755 * *in_batch*, *out_batch*, and *values* parameters are ignored 756 * and should be zeroed. 757 * 758 * The *elem_flags* argument may be specified as one of the 759 * following: 760 * 761 * **BPF_F_LOCK** 762 * Look up the value of a spin-locked map without 763 * returning the lock. This must be specified if the 764 * elements contain a spinlock. 765 * 766 * On success, *count* elements from the map are updated. 767 * 768 * If an error is returned and *errno* is not **EFAULT**, *count* 769 * is set to the number of successfully processed elements. If 770 * *errno* is **EFAULT**, up to *count* elements may be been 771 * deleted. 772 * 773 * Return 774 * Returns zero on success. On error, -1 is returned and *errno* 775 * is set appropriately. 776 * 777 * BPF_LINK_CREATE 778 * Description 779 * Attach an eBPF program to a *target_fd* at the specified 780 * *attach_type* hook and return a file descriptor handle for 781 * managing the link. 782 * 783 * Return 784 * A new file descriptor (a nonnegative integer), or -1 if an 785 * error occurred (in which case, *errno* is set appropriately). 786 * 787 * BPF_LINK_UPDATE 788 * Description 789 * Update the eBPF program in the specified *link_fd* to 790 * *new_prog_fd*. 791 * 792 * Return 793 * Returns zero on success. On error, -1 is returned and *errno* 794 * is set appropriately. 795 * 796 * BPF_LINK_GET_FD_BY_ID 797 * Description 798 * Open a file descriptor for the eBPF Link corresponding to 799 * *link_id*. 800 * 801 * Return 802 * A new file descriptor (a nonnegative integer), or -1 if an 803 * error occurred (in which case, *errno* is set appropriately). 804 * 805 * BPF_LINK_GET_NEXT_ID 806 * Description 807 * Fetch the next eBPF link currently loaded into the kernel. 808 * 809 * Looks for the eBPF link with an id greater than *start_id* 810 * and updates *next_id* on success. If no other eBPF links 811 * remain with ids higher than *start_id*, returns -1 and sets 812 * *errno* to **ENOENT**. 813 * 814 * Return 815 * Returns zero on success. On error, or when no id remains, -1 816 * is returned and *errno* is set appropriately. 817 * 818 * BPF_ENABLE_STATS 819 * Description 820 * Enable eBPF runtime statistics gathering. 821 * 822 * Runtime statistics gathering for the eBPF runtime is disabled 823 * by default to minimize the corresponding performance overhead. 824 * This command enables statistics globally. 825 * 826 * Multiple programs may independently enable statistics. 827 * After gathering the desired statistics, eBPF runtime statistics 828 * may be disabled again by calling **close**\ (2) for the file 829 * descriptor returned by this function. Statistics will only be 830 * disabled system-wide when all outstanding file descriptors 831 * returned by prior calls for this subcommand are closed. 832 * 833 * Return 834 * A new file descriptor (a nonnegative integer), or -1 if an 835 * error occurred (in which case, *errno* is set appropriately). 836 * 837 * BPF_ITER_CREATE 838 * Description 839 * Create an iterator on top of the specified *link_fd* (as 840 * previously created using **BPF_LINK_CREATE**) and return a 841 * file descriptor that can be used to trigger the iteration. 842 * 843 * If the resulting file descriptor is pinned to the filesystem 844 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 845 * for that path will trigger the iterator to read kernel state 846 * using the eBPF program attached to *link_fd*. 847 * 848 * Return 849 * A new file descriptor (a nonnegative integer), or -1 if an 850 * error occurred (in which case, *errno* is set appropriately). 851 * 852 * BPF_LINK_DETACH 853 * Description 854 * Forcefully detach the specified *link_fd* from its 855 * corresponding attachment point. 856 * 857 * Return 858 * Returns zero on success. On error, -1 is returned and *errno* 859 * is set appropriately. 860 * 861 * BPF_PROG_BIND_MAP 862 * Description 863 * Bind a map to the lifetime of an eBPF program. 864 * 865 * The map identified by *map_fd* is bound to the program 866 * identified by *prog_fd* and only released when *prog_fd* is 867 * released. This may be used in cases where metadata should be 868 * associated with a program which otherwise does not contain any 869 * references to the map (for example, embedded in the eBPF 870 * program instructions). 871 * 872 * Return 873 * Returns zero on success. On error, -1 is returned and *errno* 874 * is set appropriately. 875 * 876 * BPF_TOKEN_CREATE 877 * Description 878 * Create BPF token with embedded information about what 879 * BPF-related functionality it allows: 880 * - a set of allowed bpf() syscall commands; 881 * - a set of allowed BPF map types to be created with 882 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 883 * - a set of allowed BPF program types and BPF program attach 884 * types to be loaded with BPF_PROG_LOAD command, if 885 * BPF_PROG_LOAD itself is allowed. 886 * 887 * BPF token is created (derived) from an instance of BPF FS, 888 * assuming it has necessary delegation mount options specified. 889 * This BPF token can be passed as an extra parameter to various 890 * bpf() syscall commands to grant BPF subsystem functionality to 891 * unprivileged processes. 892 * 893 * When created, BPF token is "associated" with the owning 894 * user namespace of BPF FS instance (super block) that it was 895 * derived from, and subsequent BPF operations performed with 896 * BPF token would be performing capabilities checks (i.e., 897 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 898 * that user namespace. Without BPF token, such capabilities 899 * have to be granted in init user namespace, making bpf() 900 * syscall incompatible with user namespace, for the most part. 901 * 902 * Return 903 * A new file descriptor (a nonnegative integer), or -1 if an 904 * error occurred (in which case, *errno* is set appropriately). 905 * 906 * NOTES 907 * eBPF objects (maps and programs) can be shared between processes. 908 * 909 * * After **fork**\ (2), the child inherits file descriptors 910 * referring to the same eBPF objects. 911 * * File descriptors referring to eBPF objects can be transferred over 912 * **unix**\ (7) domain sockets. 913 * * File descriptors referring to eBPF objects can be duplicated in the 914 * usual way, using **dup**\ (2) and similar calls. 915 * * File descriptors referring to eBPF objects can be pinned to the 916 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 917 * 918 * An eBPF object is deallocated only after all file descriptors referring 919 * to the object have been closed and no references remain pinned to the 920 * filesystem or attached (for example, bound to a program or device). 921 */ 922 enum bpf_cmd { 923 BPF_MAP_CREATE, 924 BPF_MAP_LOOKUP_ELEM, 925 BPF_MAP_UPDATE_ELEM, 926 BPF_MAP_DELETE_ELEM, 927 BPF_MAP_GET_NEXT_KEY, 928 BPF_PROG_LOAD, 929 BPF_OBJ_PIN, 930 BPF_OBJ_GET, 931 BPF_PROG_ATTACH, 932 BPF_PROG_DETACH, 933 BPF_PROG_TEST_RUN, 934 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 935 BPF_PROG_GET_NEXT_ID, 936 BPF_MAP_GET_NEXT_ID, 937 BPF_PROG_GET_FD_BY_ID, 938 BPF_MAP_GET_FD_BY_ID, 939 BPF_OBJ_GET_INFO_BY_FD, 940 BPF_PROG_QUERY, 941 BPF_RAW_TRACEPOINT_OPEN, 942 BPF_BTF_LOAD, 943 BPF_BTF_GET_FD_BY_ID, 944 BPF_TASK_FD_QUERY, 945 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 946 BPF_MAP_FREEZE, 947 BPF_BTF_GET_NEXT_ID, 948 BPF_MAP_LOOKUP_BATCH, 949 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 950 BPF_MAP_UPDATE_BATCH, 951 BPF_MAP_DELETE_BATCH, 952 BPF_LINK_CREATE, 953 BPF_LINK_UPDATE, 954 BPF_LINK_GET_FD_BY_ID, 955 BPF_LINK_GET_NEXT_ID, 956 BPF_ENABLE_STATS, 957 BPF_ITER_CREATE, 958 BPF_LINK_DETACH, 959 BPF_PROG_BIND_MAP, 960 BPF_TOKEN_CREATE, 961 __MAX_BPF_CMD, 962 }; 963 964 enum bpf_map_type { 965 BPF_MAP_TYPE_UNSPEC, 966 BPF_MAP_TYPE_HASH, 967 BPF_MAP_TYPE_ARRAY, 968 BPF_MAP_TYPE_PROG_ARRAY, 969 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 970 BPF_MAP_TYPE_PERCPU_HASH, 971 BPF_MAP_TYPE_PERCPU_ARRAY, 972 BPF_MAP_TYPE_STACK_TRACE, 973 BPF_MAP_TYPE_CGROUP_ARRAY, 974 BPF_MAP_TYPE_LRU_HASH, 975 BPF_MAP_TYPE_LRU_PERCPU_HASH, 976 BPF_MAP_TYPE_LPM_TRIE, 977 BPF_MAP_TYPE_ARRAY_OF_MAPS, 978 BPF_MAP_TYPE_HASH_OF_MAPS, 979 BPF_MAP_TYPE_DEVMAP, 980 BPF_MAP_TYPE_SOCKMAP, 981 BPF_MAP_TYPE_CPUMAP, 982 BPF_MAP_TYPE_XSKMAP, 983 BPF_MAP_TYPE_SOCKHASH, 984 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 985 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 986 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 987 * both cgroup-attached and other progs and supports all functionality 988 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 989 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 990 */ 991 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 992 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 993 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 994 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 995 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 996 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 997 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 998 * deprecated. 999 */ 1000 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 1001 BPF_MAP_TYPE_QUEUE, 1002 BPF_MAP_TYPE_STACK, 1003 BPF_MAP_TYPE_SK_STORAGE, 1004 BPF_MAP_TYPE_DEVMAP_HASH, 1005 BPF_MAP_TYPE_STRUCT_OPS, 1006 BPF_MAP_TYPE_RINGBUF, 1007 BPF_MAP_TYPE_INODE_STORAGE, 1008 BPF_MAP_TYPE_TASK_STORAGE, 1009 BPF_MAP_TYPE_BLOOM_FILTER, 1010 BPF_MAP_TYPE_USER_RINGBUF, 1011 BPF_MAP_TYPE_CGRP_STORAGE, 1012 BPF_MAP_TYPE_ARENA, 1013 __MAX_BPF_MAP_TYPE 1014 }; 1015 1016 /* Note that tracing related programs such as 1017 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1018 * are not subject to a stable API since kernel internal data 1019 * structures can change from release to release and may 1020 * therefore break existing tracing BPF programs. Tracing BPF 1021 * programs correspond to /a/ specific kernel which is to be 1022 * analyzed, and not /a/ specific kernel /and/ all future ones. 1023 */ 1024 enum bpf_prog_type { 1025 BPF_PROG_TYPE_UNSPEC, 1026 BPF_PROG_TYPE_SOCKET_FILTER, 1027 BPF_PROG_TYPE_KPROBE, 1028 BPF_PROG_TYPE_SCHED_CLS, 1029 BPF_PROG_TYPE_SCHED_ACT, 1030 BPF_PROG_TYPE_TRACEPOINT, 1031 BPF_PROG_TYPE_XDP, 1032 BPF_PROG_TYPE_PERF_EVENT, 1033 BPF_PROG_TYPE_CGROUP_SKB, 1034 BPF_PROG_TYPE_CGROUP_SOCK, 1035 BPF_PROG_TYPE_LWT_IN, 1036 BPF_PROG_TYPE_LWT_OUT, 1037 BPF_PROG_TYPE_LWT_XMIT, 1038 BPF_PROG_TYPE_SOCK_OPS, 1039 BPF_PROG_TYPE_SK_SKB, 1040 BPF_PROG_TYPE_CGROUP_DEVICE, 1041 BPF_PROG_TYPE_SK_MSG, 1042 BPF_PROG_TYPE_RAW_TRACEPOINT, 1043 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1044 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1045 BPF_PROG_TYPE_LIRC_MODE2, 1046 BPF_PROG_TYPE_SK_REUSEPORT, 1047 BPF_PROG_TYPE_FLOW_DISSECTOR, 1048 BPF_PROG_TYPE_CGROUP_SYSCTL, 1049 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1050 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1051 BPF_PROG_TYPE_TRACING, 1052 BPF_PROG_TYPE_STRUCT_OPS, 1053 BPF_PROG_TYPE_EXT, 1054 BPF_PROG_TYPE_LSM, 1055 BPF_PROG_TYPE_SK_LOOKUP, 1056 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1057 BPF_PROG_TYPE_NETFILTER, 1058 1059 /* 1060 * Until fuse-bpf is upstreamed, this value must be at the end to allow for 1061 * other recently-added upstreamed values to be correct. 1062 * This works because no one should use this value directly, rather they must 1063 * read the value from /sys/fs/fuse/bpf_prog_type_fuse 1064 * Please maintain this value at the end of the list until fuse-bpf is 1065 * upstreamed. 1066 */ 1067 BPF_PROG_TYPE_FUSE, 1068 __MAX_BPF_PROG_TYPE 1069 }; 1070 1071 enum bpf_attach_type { 1072 BPF_CGROUP_INET_INGRESS, 1073 BPF_CGROUP_INET_EGRESS, 1074 BPF_CGROUP_INET_SOCK_CREATE, 1075 BPF_CGROUP_SOCK_OPS, 1076 BPF_SK_SKB_STREAM_PARSER, 1077 BPF_SK_SKB_STREAM_VERDICT, 1078 BPF_CGROUP_DEVICE, 1079 BPF_SK_MSG_VERDICT, 1080 BPF_CGROUP_INET4_BIND, 1081 BPF_CGROUP_INET6_BIND, 1082 BPF_CGROUP_INET4_CONNECT, 1083 BPF_CGROUP_INET6_CONNECT, 1084 BPF_CGROUP_INET4_POST_BIND, 1085 BPF_CGROUP_INET6_POST_BIND, 1086 BPF_CGROUP_UDP4_SENDMSG, 1087 BPF_CGROUP_UDP6_SENDMSG, 1088 BPF_LIRC_MODE2, 1089 BPF_FLOW_DISSECTOR, 1090 BPF_CGROUP_SYSCTL, 1091 BPF_CGROUP_UDP4_RECVMSG, 1092 BPF_CGROUP_UDP6_RECVMSG, 1093 BPF_CGROUP_GETSOCKOPT, 1094 BPF_CGROUP_SETSOCKOPT, 1095 BPF_TRACE_RAW_TP, 1096 BPF_TRACE_FENTRY, 1097 BPF_TRACE_FEXIT, 1098 BPF_MODIFY_RETURN, 1099 BPF_LSM_MAC, 1100 BPF_TRACE_ITER, 1101 BPF_CGROUP_INET4_GETPEERNAME, 1102 BPF_CGROUP_INET6_GETPEERNAME, 1103 BPF_CGROUP_INET4_GETSOCKNAME, 1104 BPF_CGROUP_INET6_GETSOCKNAME, 1105 BPF_XDP_DEVMAP, 1106 BPF_CGROUP_INET_SOCK_RELEASE, 1107 BPF_XDP_CPUMAP, 1108 BPF_SK_LOOKUP, 1109 BPF_XDP, 1110 BPF_SK_SKB_VERDICT, 1111 BPF_SK_REUSEPORT_SELECT, 1112 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1113 BPF_PERF_EVENT, 1114 BPF_TRACE_KPROBE_MULTI, 1115 BPF_LSM_CGROUP, 1116 BPF_STRUCT_OPS, 1117 BPF_NETFILTER, 1118 BPF_TCX_INGRESS, 1119 BPF_TCX_EGRESS, 1120 BPF_TRACE_UPROBE_MULTI, 1121 BPF_CGROUP_UNIX_CONNECT, 1122 BPF_CGROUP_UNIX_SENDMSG, 1123 BPF_CGROUP_UNIX_RECVMSG, 1124 BPF_CGROUP_UNIX_GETPEERNAME, 1125 BPF_CGROUP_UNIX_GETSOCKNAME, 1126 BPF_NETKIT_PRIMARY, 1127 BPF_NETKIT_PEER, 1128 BPF_TRACE_KPROBE_SESSION, 1129 __MAX_BPF_ATTACH_TYPE 1130 }; 1131 1132 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1133 1134 /* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[] 1135 * in sync with the definitions below. 1136 */ 1137 enum bpf_link_type { 1138 BPF_LINK_TYPE_UNSPEC = 0, 1139 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1140 BPF_LINK_TYPE_TRACING = 2, 1141 BPF_LINK_TYPE_CGROUP = 3, 1142 BPF_LINK_TYPE_ITER = 4, 1143 BPF_LINK_TYPE_NETNS = 5, 1144 BPF_LINK_TYPE_XDP = 6, 1145 BPF_LINK_TYPE_PERF_EVENT = 7, 1146 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1147 BPF_LINK_TYPE_STRUCT_OPS = 9, 1148 BPF_LINK_TYPE_NETFILTER = 10, 1149 BPF_LINK_TYPE_TCX = 11, 1150 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1151 BPF_LINK_TYPE_NETKIT = 13, 1152 BPF_LINK_TYPE_SOCKMAP = 14, 1153 __MAX_BPF_LINK_TYPE, 1154 }; 1155 1156 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1157 1158 enum bpf_perf_event_type { 1159 BPF_PERF_EVENT_UNSPEC = 0, 1160 BPF_PERF_EVENT_UPROBE = 1, 1161 BPF_PERF_EVENT_URETPROBE = 2, 1162 BPF_PERF_EVENT_KPROBE = 3, 1163 BPF_PERF_EVENT_KRETPROBE = 4, 1164 BPF_PERF_EVENT_TRACEPOINT = 5, 1165 BPF_PERF_EVENT_EVENT = 6, 1166 }; 1167 1168 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1169 * 1170 * NONE(default): No further bpf programs allowed in the subtree. 1171 * 1172 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1173 * the program in this cgroup yields to sub-cgroup program. 1174 * 1175 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1176 * that cgroup program gets run in addition to the program in this cgroup. 1177 * 1178 * Only one program is allowed to be attached to a cgroup with 1179 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1180 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1181 * release old program and attach the new one. Attach flags has to match. 1182 * 1183 * Multiple programs are allowed to be attached to a cgroup with 1184 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1185 * (those that were attached first, run first) 1186 * The programs of sub-cgroup are executed first, then programs of 1187 * this cgroup and then programs of parent cgroup. 1188 * When children program makes decision (like picking TCP CA or sock bind) 1189 * parent program has a chance to override it. 1190 * 1191 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1192 * programs for a cgroup. Though it's possible to replace an old program at 1193 * any position by also specifying BPF_F_REPLACE flag and position itself in 1194 * replace_bpf_fd attribute. Old program at this position will be released. 1195 * 1196 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1197 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1198 * Ex1: 1199 * cgrp1 (MULTI progs A, B) -> 1200 * cgrp2 (OVERRIDE prog C) -> 1201 * cgrp3 (MULTI prog D) -> 1202 * cgrp4 (OVERRIDE prog E) -> 1203 * cgrp5 (NONE prog F) 1204 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1205 * if prog F is detached, the execution is E,D,A,B 1206 * if prog F and D are detached, the execution is E,A,B 1207 * if prog F, E and D are detached, the execution is C,A,B 1208 * 1209 * All eligible programs are executed regardless of return code from 1210 * earlier programs. 1211 */ 1212 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1213 #define BPF_F_ALLOW_MULTI (1U << 1) 1214 /* Generic attachment flags. */ 1215 #define BPF_F_REPLACE (1U << 2) 1216 #define BPF_F_BEFORE (1U << 3) 1217 #define BPF_F_AFTER (1U << 4) 1218 #define BPF_F_ID (1U << 5) 1219 #define BPF_F_PREORDER (1U << 6) 1220 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1221 1222 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1223 * verifier will perform strict alignment checking as if the kernel 1224 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1225 * and NET_IP_ALIGN defined to 2. 1226 */ 1227 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1228 1229 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1230 * verifier will allow any alignment whatsoever. On platforms 1231 * with strict alignment requirements for loads ands stores (such 1232 * as sparc and mips) the verifier validates that all loads and 1233 * stores provably follow this requirement. This flag turns that 1234 * checking and enforcement off. 1235 * 1236 * It is mostly used for testing when we want to validate the 1237 * context and memory access aspects of the verifier, but because 1238 * of an unaligned access the alignment check would trigger before 1239 * the one we are interested in. 1240 */ 1241 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1242 1243 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1244 * Verifier does sub-register def/use analysis and identifies instructions whose 1245 * def only matters for low 32-bit, high 32-bit is never referenced later 1246 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1247 * that it is safe to ignore clearing high 32-bit for these instructions. This 1248 * saves some back-ends a lot of code-gen. However such optimization is not 1249 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1250 * hence hasn't used verifier's analysis result. But, we really want to have a 1251 * way to be able to verify the correctness of the described optimization on 1252 * x86_64 on which testsuites are frequently exercised. 1253 * 1254 * So, this flag is introduced. Once it is set, verifier will randomize high 1255 * 32-bit for those instructions who has been identified as safe to ignore them. 1256 * Then, if verifier is not doing correct analysis, such randomization will 1257 * regress tests to expose bugs. 1258 */ 1259 #define BPF_F_TEST_RND_HI32 (1U << 2) 1260 1261 /* The verifier internal test flag. Behavior is undefined */ 1262 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1263 1264 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1265 * restrict map and helper usage for such programs. Sleepable BPF programs can 1266 * only be attached to hooks where kernel execution context allows sleeping. 1267 * Such programs are allowed to use helpers that may sleep like 1268 * bpf_copy_from_user(). 1269 */ 1270 #define BPF_F_SLEEPABLE (1U << 4) 1271 1272 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1273 * fully support xdp frags. 1274 */ 1275 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1276 1277 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1278 * program becomes device-bound but can access XDP metadata. 1279 */ 1280 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1281 1282 /* The verifier internal test flag. Behavior is undefined */ 1283 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1284 1285 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1286 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1287 */ 1288 enum { 1289 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1290 }; 1291 1292 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1293 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1294 */ 1295 enum { 1296 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1297 }; 1298 1299 /* link_create.netfilter.flags used in LINK_CREATE command for 1300 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1301 */ 1302 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1303 1304 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1305 * the following extensions: 1306 * 1307 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1308 * insn[0].imm: map fd or fd_idx 1309 * insn[1].imm: 0 1310 * insn[0].off: 0 1311 * insn[1].off: 0 1312 * ldimm64 rewrite: address of map 1313 * verifier type: CONST_PTR_TO_MAP 1314 */ 1315 #define BPF_PSEUDO_MAP_FD 1 1316 #define BPF_PSEUDO_MAP_IDX 5 1317 1318 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1319 * insn[0].imm: map fd or fd_idx 1320 * insn[1].imm: offset into value 1321 * insn[0].off: 0 1322 * insn[1].off: 0 1323 * ldimm64 rewrite: address of map[0]+offset 1324 * verifier type: PTR_TO_MAP_VALUE 1325 */ 1326 #define BPF_PSEUDO_MAP_VALUE 2 1327 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1328 1329 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1330 * insn[0].imm: kernel btd id of VAR 1331 * insn[1].imm: 0 1332 * insn[0].off: 0 1333 * insn[1].off: 0 1334 * ldimm64 rewrite: address of the kernel variable 1335 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1336 * is struct/union. 1337 */ 1338 #define BPF_PSEUDO_BTF_ID 3 1339 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1340 * insn[0].imm: insn offset to the func 1341 * insn[1].imm: 0 1342 * insn[0].off: 0 1343 * insn[1].off: 0 1344 * ldimm64 rewrite: address of the function 1345 * verifier type: PTR_TO_FUNC. 1346 */ 1347 #define BPF_PSEUDO_FUNC 4 1348 1349 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1350 * offset to another bpf function 1351 */ 1352 #define BPF_PSEUDO_CALL 1 1353 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1354 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1355 */ 1356 #define BPF_PSEUDO_KFUNC_CALL 2 1357 1358 enum bpf_addr_space_cast { 1359 BPF_ADDR_SPACE_CAST = 1, 1360 }; 1361 1362 /* flags for BPF_MAP_UPDATE_ELEM command */ 1363 enum { 1364 BPF_ANY = 0, /* create new element or update existing */ 1365 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1366 BPF_EXIST = 2, /* update existing element */ 1367 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1368 }; 1369 1370 /* flags for BPF_MAP_CREATE command */ 1371 enum { 1372 BPF_F_NO_PREALLOC = (1U << 0), 1373 /* Instead of having one common LRU list in the 1374 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1375 * which can scale and perform better. 1376 * Note, the LRU nodes (including free nodes) cannot be moved 1377 * across different LRU lists. 1378 */ 1379 BPF_F_NO_COMMON_LRU = (1U << 1), 1380 /* Specify numa node during map creation */ 1381 BPF_F_NUMA_NODE = (1U << 2), 1382 1383 /* Flags for accessing BPF object from syscall side. */ 1384 BPF_F_RDONLY = (1U << 3), 1385 BPF_F_WRONLY = (1U << 4), 1386 1387 /* Flag for stack_map, store build_id+offset instead of pointer */ 1388 BPF_F_STACK_BUILD_ID = (1U << 5), 1389 1390 /* Zero-initialize hash function seed. This should only be used for testing. */ 1391 BPF_F_ZERO_SEED = (1U << 6), 1392 1393 /* Flags for accessing BPF object from program side. */ 1394 BPF_F_RDONLY_PROG = (1U << 7), 1395 BPF_F_WRONLY_PROG = (1U << 8), 1396 1397 /* Clone map from listener for newly accepted socket */ 1398 BPF_F_CLONE = (1U << 9), 1399 1400 /* Enable memory-mapping BPF map */ 1401 BPF_F_MMAPABLE = (1U << 10), 1402 1403 /* Share perf_event among processes */ 1404 BPF_F_PRESERVE_ELEMS = (1U << 11), 1405 1406 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1407 BPF_F_INNER_MAP = (1U << 12), 1408 1409 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1410 BPF_F_LINK = (1U << 13), 1411 1412 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1413 BPF_F_PATH_FD = (1U << 14), 1414 1415 /* Flag for value_type_btf_obj_fd, the fd is available */ 1416 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1417 1418 /* BPF token FD is passed in a corresponding command's token_fd field */ 1419 BPF_F_TOKEN_FD = (1U << 16), 1420 1421 /* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */ 1422 BPF_F_SEGV_ON_FAULT = (1U << 17), 1423 1424 /* Do not translate kernel bpf_arena pointers to user pointers */ 1425 BPF_F_NO_USER_CONV = (1U << 18), 1426 }; 1427 1428 /* Flags for BPF_PROG_QUERY. */ 1429 1430 /* Query effective (directly attached + inherited from ancestor cgroups) 1431 * programs that will be executed for events within a cgroup. 1432 * attach_flags with this flag are always returned 0. 1433 */ 1434 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1435 1436 /* Flags for BPF_PROG_TEST_RUN */ 1437 1438 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1439 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1440 /* If set, XDP frames will be transmitted after processing */ 1441 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1442 /* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */ 1443 #define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2) 1444 1445 /* type for BPF_ENABLE_STATS */ 1446 enum bpf_stats_type { 1447 /* enabled run_time_ns and run_cnt */ 1448 BPF_STATS_RUN_TIME = 0, 1449 }; 1450 1451 enum bpf_stack_build_id_status { 1452 /* user space need an empty entry to identify end of a trace */ 1453 BPF_STACK_BUILD_ID_EMPTY = 0, 1454 /* with valid build_id and offset */ 1455 BPF_STACK_BUILD_ID_VALID = 1, 1456 /* couldn't get build_id, fallback to ip */ 1457 BPF_STACK_BUILD_ID_IP = 2, 1458 }; 1459 1460 #define BPF_BUILD_ID_SIZE 20 1461 struct bpf_stack_build_id { 1462 __s32 status; 1463 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1464 union { 1465 __u64 offset; 1466 __u64 ip; 1467 }; 1468 }; 1469 1470 #define BPF_OBJ_NAME_LEN 16U 1471 1472 union bpf_attr { 1473 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1474 __u32 map_type; /* one of enum bpf_map_type */ 1475 __u32 key_size; /* size of key in bytes */ 1476 __u32 value_size; /* size of value in bytes */ 1477 __u32 max_entries; /* max number of entries in a map */ 1478 __u32 map_flags; /* BPF_MAP_CREATE related 1479 * flags defined above. 1480 */ 1481 __u32 inner_map_fd; /* fd pointing to the inner map */ 1482 __u32 numa_node; /* numa node (effective only if 1483 * BPF_F_NUMA_NODE is set). 1484 */ 1485 char map_name[BPF_OBJ_NAME_LEN]; 1486 __u32 map_ifindex; /* ifindex of netdev to create on */ 1487 __u32 btf_fd; /* fd pointing to a BTF type data */ 1488 __u32 btf_key_type_id; /* BTF type_id of the key */ 1489 __u32 btf_value_type_id; /* BTF type_id of the value */ 1490 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1491 * struct stored as the 1492 * map value 1493 */ 1494 /* Any per-map-type extra fields 1495 * 1496 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1497 * number of hash functions (if 0, the bloom filter will default 1498 * to using 5 hash functions). 1499 * 1500 * BPF_MAP_TYPE_ARENA - contains the address where user space 1501 * is going to mmap() the arena. It has to be page aligned. 1502 */ 1503 __u64 map_extra; 1504 1505 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1506 * type data for 1507 * btf_vmlinux_value_type_id. 1508 */ 1509 /* BPF token FD to use with BPF_MAP_CREATE operation. 1510 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1511 */ 1512 __s32 map_token_fd; 1513 }; 1514 1515 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1516 __u32 map_fd; 1517 __aligned_u64 key; 1518 union { 1519 __aligned_u64 value; 1520 __aligned_u64 next_key; 1521 }; 1522 __u64 flags; 1523 }; 1524 1525 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1526 __aligned_u64 in_batch; /* start batch, 1527 * NULL to start from beginning 1528 */ 1529 __aligned_u64 out_batch; /* output: next start batch */ 1530 __aligned_u64 keys; 1531 __aligned_u64 values; 1532 __u32 count; /* input/output: 1533 * input: # of key/value 1534 * elements 1535 * output: # of filled elements 1536 */ 1537 __u32 map_fd; 1538 __u64 elem_flags; 1539 __u64 flags; 1540 } batch; 1541 1542 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1543 __u32 prog_type; /* one of enum bpf_prog_type */ 1544 __u32 insn_cnt; 1545 __aligned_u64 insns; 1546 __aligned_u64 license; 1547 __u32 log_level; /* verbosity level of verifier */ 1548 __u32 log_size; /* size of user buffer */ 1549 __aligned_u64 log_buf; /* user supplied buffer */ 1550 __u32 kern_version; /* not used */ 1551 __u32 prog_flags; 1552 char prog_name[BPF_OBJ_NAME_LEN]; 1553 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1554 /* For some prog types expected attach type must be known at 1555 * load time to verify attach type specific parts of prog 1556 * (context accesses, allowed helpers, etc). 1557 */ 1558 __u32 expected_attach_type; 1559 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1560 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1561 __aligned_u64 func_info; /* func info */ 1562 __u32 func_info_cnt; /* number of bpf_func_info records */ 1563 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1564 __aligned_u64 line_info; /* line info */ 1565 __u32 line_info_cnt; /* number of bpf_line_info records */ 1566 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1567 union { 1568 /* valid prog_fd to attach to bpf prog */ 1569 __u32 attach_prog_fd; 1570 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1571 __u32 attach_btf_obj_fd; 1572 }; 1573 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1574 __aligned_u64 fd_array; /* array of FDs */ 1575 __aligned_u64 core_relos; 1576 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1577 /* output: actual total log contents size (including termintaing zero). 1578 * It could be both larger than original log_size (if log was 1579 * truncated), or smaller (if log buffer wasn't filled completely). 1580 */ 1581 __u32 log_true_size; 1582 /* BPF token FD to use with BPF_PROG_LOAD operation. 1583 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1584 */ 1585 __s32 prog_token_fd; 1586 }; 1587 1588 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1589 __aligned_u64 pathname; 1590 __u32 bpf_fd; 1591 __u32 file_flags; 1592 /* Same as dirfd in openat() syscall; see openat(2) 1593 * manpage for details of path FD and pathname semantics; 1594 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1595 * file_flags field, otherwise it should be set to zero; 1596 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1597 */ 1598 __s32 path_fd; 1599 }; 1600 1601 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1602 union { 1603 __u32 target_fd; /* target object to attach to or ... */ 1604 __u32 target_ifindex; /* target ifindex */ 1605 }; 1606 __u32 attach_bpf_fd; 1607 __u32 attach_type; 1608 __u32 attach_flags; 1609 __u32 replace_bpf_fd; 1610 union { 1611 __u32 relative_fd; 1612 __u32 relative_id; 1613 }; 1614 __u64 expected_revision; 1615 }; 1616 1617 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1618 __u32 prog_fd; 1619 __u32 retval; 1620 __u32 data_size_in; /* input: len of data_in */ 1621 __u32 data_size_out; /* input/output: len of data_out 1622 * returns ENOSPC if data_out 1623 * is too small. 1624 */ 1625 __aligned_u64 data_in; 1626 __aligned_u64 data_out; 1627 __u32 repeat; 1628 __u32 duration; 1629 __u32 ctx_size_in; /* input: len of ctx_in */ 1630 __u32 ctx_size_out; /* input/output: len of ctx_out 1631 * returns ENOSPC if ctx_out 1632 * is too small. 1633 */ 1634 __aligned_u64 ctx_in; 1635 __aligned_u64 ctx_out; 1636 __u32 flags; 1637 __u32 cpu; 1638 __u32 batch_size; 1639 } test; 1640 1641 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1642 union { 1643 __u32 start_id; 1644 __u32 prog_id; 1645 __u32 map_id; 1646 __u32 btf_id; 1647 __u32 link_id; 1648 }; 1649 __u32 next_id; 1650 __u32 open_flags; 1651 }; 1652 1653 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1654 __u32 bpf_fd; 1655 __u32 info_len; 1656 __aligned_u64 info; 1657 } info; 1658 1659 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1660 union { 1661 __u32 target_fd; /* target object to query or ... */ 1662 __u32 target_ifindex; /* target ifindex */ 1663 }; 1664 __u32 attach_type; 1665 __u32 query_flags; 1666 __u32 attach_flags; 1667 __aligned_u64 prog_ids; 1668 union { 1669 __u32 prog_cnt; 1670 __u32 count; 1671 }; 1672 __u32 :32; 1673 /* output: per-program attach_flags. 1674 * not allowed to be set during effective query. 1675 */ 1676 __aligned_u64 prog_attach_flags; 1677 __aligned_u64 link_ids; 1678 __aligned_u64 link_attach_flags; 1679 __u64 revision; 1680 } query; 1681 1682 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1683 __u64 name; 1684 __u32 prog_fd; 1685 __u32 :32; 1686 __aligned_u64 cookie; 1687 } raw_tracepoint; 1688 1689 struct { /* anonymous struct for BPF_BTF_LOAD */ 1690 __aligned_u64 btf; 1691 __aligned_u64 btf_log_buf; 1692 __u32 btf_size; 1693 __u32 btf_log_size; 1694 __u32 btf_log_level; 1695 /* output: actual total log contents size (including termintaing zero). 1696 * It could be both larger than original log_size (if log was 1697 * truncated), or smaller (if log buffer wasn't filled completely). 1698 */ 1699 __u32 btf_log_true_size; 1700 __u32 btf_flags; 1701 /* BPF token FD to use with BPF_BTF_LOAD operation. 1702 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1703 */ 1704 __s32 btf_token_fd; 1705 }; 1706 1707 struct { 1708 __u32 pid; /* input: pid */ 1709 __u32 fd; /* input: fd */ 1710 __u32 flags; /* input: flags */ 1711 __u32 buf_len; /* input/output: buf len */ 1712 __aligned_u64 buf; /* input/output: 1713 * tp_name for tracepoint 1714 * symbol for kprobe 1715 * filename for uprobe 1716 */ 1717 __u32 prog_id; /* output: prod_id */ 1718 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1719 __u64 probe_offset; /* output: probe_offset */ 1720 __u64 probe_addr; /* output: probe_addr */ 1721 } task_fd_query; 1722 1723 struct { /* struct used by BPF_LINK_CREATE command */ 1724 union { 1725 __u32 prog_fd; /* eBPF program to attach */ 1726 __u32 map_fd; /* struct_ops to attach */ 1727 }; 1728 union { 1729 __u32 target_fd; /* target object to attach to or ... */ 1730 __u32 target_ifindex; /* target ifindex */ 1731 }; 1732 __u32 attach_type; /* attach type */ 1733 __u32 flags; /* extra flags */ 1734 union { 1735 __u32 target_btf_id; /* btf_id of target to attach to */ 1736 struct { 1737 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1738 __u32 iter_info_len; /* iter_info length */ 1739 }; 1740 struct { 1741 /* black box user-provided value passed through 1742 * to BPF program at the execution time and 1743 * accessible through bpf_get_attach_cookie() BPF helper 1744 */ 1745 __u64 bpf_cookie; 1746 } perf_event; 1747 struct { 1748 __u32 flags; 1749 __u32 cnt; 1750 __aligned_u64 syms; 1751 __aligned_u64 addrs; 1752 __aligned_u64 cookies; 1753 } kprobe_multi; 1754 struct { 1755 /* this is overlaid with the target_btf_id above. */ 1756 __u32 target_btf_id; 1757 /* black box user-provided value passed through 1758 * to BPF program at the execution time and 1759 * accessible through bpf_get_attach_cookie() BPF helper 1760 */ 1761 __u64 cookie; 1762 } tracing; 1763 struct { 1764 __u32 pf; 1765 __u32 hooknum; 1766 __s32 priority; 1767 __u32 flags; 1768 } netfilter; 1769 struct { 1770 union { 1771 __u32 relative_fd; 1772 __u32 relative_id; 1773 }; 1774 __u64 expected_revision; 1775 } tcx; 1776 struct { 1777 __aligned_u64 path; 1778 __aligned_u64 offsets; 1779 __aligned_u64 ref_ctr_offsets; 1780 __aligned_u64 cookies; 1781 __u32 cnt; 1782 __u32 flags; 1783 __u32 pid; 1784 } uprobe_multi; 1785 struct { 1786 union { 1787 __u32 relative_fd; 1788 __u32 relative_id; 1789 }; 1790 __u64 expected_revision; 1791 } netkit; 1792 }; 1793 } link_create; 1794 1795 struct { /* struct used by BPF_LINK_UPDATE command */ 1796 __u32 link_fd; /* link fd */ 1797 union { 1798 /* new program fd to update link with */ 1799 __u32 new_prog_fd; 1800 /* new struct_ops map fd to update link with */ 1801 __u32 new_map_fd; 1802 }; 1803 __u32 flags; /* extra flags */ 1804 union { 1805 /* expected link's program fd; is specified only if 1806 * BPF_F_REPLACE flag is set in flags. 1807 */ 1808 __u32 old_prog_fd; 1809 /* expected link's map fd; is specified only 1810 * if BPF_F_REPLACE flag is set. 1811 */ 1812 __u32 old_map_fd; 1813 }; 1814 } link_update; 1815 1816 struct { 1817 __u32 link_fd; 1818 } link_detach; 1819 1820 struct { /* struct used by BPF_ENABLE_STATS command */ 1821 __u32 type; 1822 } enable_stats; 1823 1824 struct { /* struct used by BPF_ITER_CREATE command */ 1825 __u32 link_fd; 1826 __u32 flags; 1827 } iter_create; 1828 1829 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1830 __u32 prog_fd; 1831 __u32 map_fd; 1832 __u32 flags; /* extra flags */ 1833 } prog_bind_map; 1834 1835 struct { /* struct used by BPF_TOKEN_CREATE command */ 1836 __u32 flags; 1837 __u32 bpffs_fd; 1838 } token_create; 1839 1840 } __attribute__((aligned(8))); 1841 1842 /* The description below is an attempt at providing documentation to eBPF 1843 * developers about the multiple available eBPF helper functions. It can be 1844 * parsed and used to produce a manual page. The workflow is the following, 1845 * and requires the rst2man utility: 1846 * 1847 * $ ./scripts/bpf_doc.py \ 1848 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1849 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1850 * $ man /tmp/bpf-helpers.7 1851 * 1852 * Note that in order to produce this external documentation, some RST 1853 * formatting is used in the descriptions to get "bold" and "italics" in 1854 * manual pages. Also note that the few trailing white spaces are 1855 * intentional, removing them would break paragraphs for rst2man. 1856 * 1857 * Start of BPF helper function descriptions: 1858 * 1859 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1860 * Description 1861 * Perform a lookup in *map* for an entry associated to *key*. 1862 * Return 1863 * Map value associated to *key*, or **NULL** if no entry was 1864 * found. 1865 * 1866 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1867 * Description 1868 * Add or update the value of the entry associated to *key* in 1869 * *map* with *value*. *flags* is one of: 1870 * 1871 * **BPF_NOEXIST** 1872 * The entry for *key* must not exist in the map. 1873 * **BPF_EXIST** 1874 * The entry for *key* must already exist in the map. 1875 * **BPF_ANY** 1876 * No condition on the existence of the entry for *key*. 1877 * 1878 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1879 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1880 * elements always exist), the helper would return an error. 1881 * Return 1882 * 0 on success, or a negative error in case of failure. 1883 * 1884 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1885 * Description 1886 * Delete entry with *key* from *map*. 1887 * Return 1888 * 0 on success, or a negative error in case of failure. 1889 * 1890 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1891 * Description 1892 * For tracing programs, safely attempt to read *size* bytes from 1893 * kernel space address *unsafe_ptr* and store the data in *dst*. 1894 * 1895 * Generally, use **bpf_probe_read_user**\ () or 1896 * **bpf_probe_read_kernel**\ () instead. 1897 * Return 1898 * 0 on success, or a negative error in case of failure. 1899 * 1900 * u64 bpf_ktime_get_ns(void) 1901 * Description 1902 * Return the time elapsed since system boot, in nanoseconds. 1903 * Does not include time the system was suspended. 1904 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1905 * Return 1906 * Current *ktime*. 1907 * 1908 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1909 * Description 1910 * This helper is a "printk()-like" facility for debugging. It 1911 * prints a message defined by format *fmt* (of size *fmt_size*) 1912 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1913 * available. It can take up to three additional **u64** 1914 * arguments (as an eBPF helpers, the total number of arguments is 1915 * limited to five). 1916 * 1917 * Each time the helper is called, it appends a line to the trace. 1918 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1919 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1920 * The format of the trace is customizable, and the exact output 1921 * one will get depends on the options set in 1922 * *\/sys/kernel/tracing/trace_options* (see also the 1923 * *README* file under the same directory). However, it usually 1924 * defaults to something like: 1925 * 1926 * :: 1927 * 1928 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1929 * 1930 * In the above: 1931 * 1932 * * ``telnet`` is the name of the current task. 1933 * * ``470`` is the PID of the current task. 1934 * * ``001`` is the CPU number on which the task is 1935 * running. 1936 * * In ``.N..``, each character refers to a set of 1937 * options (whether irqs are enabled, scheduling 1938 * options, whether hard/softirqs are running, level of 1939 * preempt_disabled respectively). **N** means that 1940 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1941 * are set. 1942 * * ``419421.045894`` is a timestamp. 1943 * * ``0x00000001`` is a fake value used by BPF for the 1944 * instruction pointer register. 1945 * * ``<formatted msg>`` is the message formatted with 1946 * *fmt*. 1947 * 1948 * The conversion specifiers supported by *fmt* are similar, but 1949 * more limited than for printk(). They are **%d**, **%i**, 1950 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1951 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1952 * of field, padding with zeroes, etc.) is available, and the 1953 * helper will return **-EINVAL** (but print nothing) if it 1954 * encounters an unknown specifier. 1955 * 1956 * Also, note that **bpf_trace_printk**\ () is slow, and should 1957 * only be used for debugging purposes. For this reason, a notice 1958 * block (spanning several lines) is printed to kernel logs and 1959 * states that the helper should not be used "for production use" 1960 * the first time this helper is used (or more precisely, when 1961 * **trace_printk**\ () buffers are allocated). For passing values 1962 * to user space, perf events should be preferred. 1963 * Return 1964 * The number of bytes written to the buffer, or a negative error 1965 * in case of failure. 1966 * 1967 * u32 bpf_get_prandom_u32(void) 1968 * Description 1969 * Get a pseudo-random number. 1970 * 1971 * From a security point of view, this helper uses its own 1972 * pseudo-random internal state, and cannot be used to infer the 1973 * seed of other random functions in the kernel. However, it is 1974 * essential to note that the generator used by the helper is not 1975 * cryptographically secure. 1976 * Return 1977 * A random 32-bit unsigned value. 1978 * 1979 * u32 bpf_get_smp_processor_id(void) 1980 * Description 1981 * Get the SMP (symmetric multiprocessing) processor id. Note that 1982 * all programs run with migration disabled, which means that the 1983 * SMP processor id is stable during all the execution of the 1984 * program. 1985 * Return 1986 * The SMP id of the processor running the program. 1987 * 1988 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1989 * Description 1990 * Store *len* bytes from address *from* into the packet 1991 * associated to *skb*, at *offset*. *flags* are a combination of 1992 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1993 * checksum for the packet after storing the bytes) and 1994 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1995 * **->swhash** and *skb*\ **->l4hash** to 0). 1996 * 1997 * A call to this helper is susceptible to change the underlying 1998 * packet buffer. Therefore, at load time, all checks on pointers 1999 * previously done by the verifier are invalidated and must be 2000 * performed again, if the helper is used in combination with 2001 * direct packet access. 2002 * Return 2003 * 0 on success, or a negative error in case of failure. 2004 * 2005 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 2006 * Description 2007 * Recompute the layer 3 (e.g. IP) checksum for the packet 2008 * associated to *skb*. Computation is incremental, so the helper 2009 * must know the former value of the header field that was 2010 * modified (*from*), the new value of this field (*to*), and the 2011 * number of bytes (2 or 4) for this field, stored in *size*. 2012 * Alternatively, it is possible to store the difference between 2013 * the previous and the new values of the header field in *to*, by 2014 * setting *from* and *size* to 0. For both methods, *offset* 2015 * indicates the location of the IP checksum within the packet. 2016 * 2017 * This helper works in combination with **bpf_csum_diff**\ (), 2018 * which does not update the checksum in-place, but offers more 2019 * flexibility and can handle sizes larger than 2 or 4 for the 2020 * checksum to update. 2021 * 2022 * A call to this helper is susceptible to change the underlying 2023 * packet buffer. Therefore, at load time, all checks on pointers 2024 * previously done by the verifier are invalidated and must be 2025 * performed again, if the helper is used in combination with 2026 * direct packet access. 2027 * Return 2028 * 0 on success, or a negative error in case of failure. 2029 * 2030 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 2031 * Description 2032 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 2033 * packet associated to *skb*. Computation is incremental, so the 2034 * helper must know the former value of the header field that was 2035 * modified (*from*), the new value of this field (*to*), and the 2036 * number of bytes (2 or 4) for this field, stored on the lowest 2037 * four bits of *flags*. Alternatively, it is possible to store 2038 * the difference between the previous and the new values of the 2039 * header field in *to*, by setting *from* and the four lowest 2040 * bits of *flags* to 0. For both methods, *offset* indicates the 2041 * location of the IP checksum within the packet. In addition to 2042 * the size of the field, *flags* can be added (bitwise OR) actual 2043 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 2044 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 2045 * for updates resulting in a null checksum the value is set to 2046 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 2047 * the checksum is to be computed against a pseudo-header. 2048 * Flag **BPF_F_IPV6** should be set for IPv6 packets. 2049 * 2050 * This helper works in combination with **bpf_csum_diff**\ (), 2051 * which does not update the checksum in-place, but offers more 2052 * flexibility and can handle sizes larger than 2 or 4 for the 2053 * checksum to update. 2054 * 2055 * A call to this helper is susceptible to change the underlying 2056 * packet buffer. Therefore, at load time, all checks on pointers 2057 * previously done by the verifier are invalidated and must be 2058 * performed again, if the helper is used in combination with 2059 * direct packet access. 2060 * Return 2061 * 0 on success, or a negative error in case of failure. 2062 * 2063 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2064 * Description 2065 * This special helper is used to trigger a "tail call", or in 2066 * other words, to jump into another eBPF program. The same stack 2067 * frame is used (but values on stack and in registers for the 2068 * caller are not accessible to the callee). This mechanism allows 2069 * for program chaining, either for raising the maximum number of 2070 * available eBPF instructions, or to execute given programs in 2071 * conditional blocks. For security reasons, there is an upper 2072 * limit to the number of successive tail calls that can be 2073 * performed. 2074 * 2075 * Upon call of this helper, the program attempts to jump into a 2076 * program referenced at index *index* in *prog_array_map*, a 2077 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2078 * *ctx*, a pointer to the context. 2079 * 2080 * If the call succeeds, the kernel immediately runs the first 2081 * instruction of the new program. This is not a function call, 2082 * and it never returns to the previous program. If the call 2083 * fails, then the helper has no effect, and the caller continues 2084 * to run its subsequent instructions. A call can fail if the 2085 * destination program for the jump does not exist (i.e. *index* 2086 * is superior to the number of entries in *prog_array_map*), or 2087 * if the maximum number of tail calls has been reached for this 2088 * chain of programs. This limit is defined in the kernel by the 2089 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2090 * which is currently set to 33. 2091 * Return 2092 * 0 on success, or a negative error in case of failure. 2093 * 2094 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2095 * Description 2096 * Clone and redirect the packet associated to *skb* to another 2097 * net device of index *ifindex*. Both ingress and egress 2098 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2099 * value in *flags* is used to make the distinction (ingress path 2100 * is selected if the flag is present, egress path otherwise). 2101 * This is the only flag supported for now. 2102 * 2103 * In comparison with **bpf_redirect**\ () helper, 2104 * **bpf_clone_redirect**\ () has the associated cost of 2105 * duplicating the packet buffer, but this can be executed out of 2106 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2107 * efficient, but it is handled through an action code where the 2108 * redirection happens only after the eBPF program has returned. 2109 * 2110 * A call to this helper is susceptible to change the underlying 2111 * packet buffer. Therefore, at load time, all checks on pointers 2112 * previously done by the verifier are invalidated and must be 2113 * performed again, if the helper is used in combination with 2114 * direct packet access. 2115 * Return 2116 * 0 on success, or a negative error in case of failure. Positive 2117 * error indicates a potential drop or congestion in the target 2118 * device. The particular positive error codes are not defined. 2119 * 2120 * u64 bpf_get_current_pid_tgid(void) 2121 * Description 2122 * Get the current pid and tgid. 2123 * Return 2124 * A 64-bit integer containing the current tgid and pid, and 2125 * created as such: 2126 * *current_task*\ **->tgid << 32 \|** 2127 * *current_task*\ **->pid**. 2128 * 2129 * u64 bpf_get_current_uid_gid(void) 2130 * Description 2131 * Get the current uid and gid. 2132 * Return 2133 * A 64-bit integer containing the current GID and UID, and 2134 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2135 * 2136 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2137 * Description 2138 * Copy the **comm** attribute of the current task into *buf* of 2139 * *size_of_buf*. The **comm** attribute contains the name of 2140 * the executable (excluding the path) for the current task. The 2141 * *size_of_buf* must be strictly positive. On success, the 2142 * helper makes sure that the *buf* is NUL-terminated. On failure, 2143 * it is filled with zeroes. 2144 * Return 2145 * 0 on success, or a negative error in case of failure. 2146 * 2147 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2148 * Description 2149 * Retrieve the classid for the current task, i.e. for the net_cls 2150 * cgroup to which *skb* belongs. 2151 * 2152 * This helper can be used on TC egress path, but not on ingress. 2153 * 2154 * The net_cls cgroup provides an interface to tag network packets 2155 * based on a user-provided identifier for all traffic coming from 2156 * the tasks belonging to the related cgroup. See also the related 2157 * kernel documentation, available from the Linux sources in file 2158 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2159 * 2160 * The Linux kernel has two versions for cgroups: there are 2161 * cgroups v1 and cgroups v2. Both are available to users, who can 2162 * use a mixture of them, but note that the net_cls cgroup is for 2163 * cgroup v1 only. This makes it incompatible with BPF programs 2164 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2165 * only hold data for one version of cgroups at a time). 2166 * 2167 * This helper is only available is the kernel was compiled with 2168 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2169 * "**y**" or to "**m**". 2170 * Return 2171 * The classid, or 0 for the default unconfigured classid. 2172 * 2173 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2174 * Description 2175 * Push a *vlan_tci* (VLAN tag control information) of protocol 2176 * *vlan_proto* to the packet associated to *skb*, then update 2177 * the checksum. Note that if *vlan_proto* is different from 2178 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2179 * be **ETH_P_8021Q**. 2180 * 2181 * A call to this helper is susceptible to change the underlying 2182 * packet buffer. Therefore, at load time, all checks on pointers 2183 * previously done by the verifier are invalidated and must be 2184 * performed again, if the helper is used in combination with 2185 * direct packet access. 2186 * Return 2187 * 0 on success, or a negative error in case of failure. 2188 * 2189 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2190 * Description 2191 * Pop a VLAN header from the packet associated to *skb*. 2192 * 2193 * A call to this helper is susceptible to change the underlying 2194 * packet buffer. Therefore, at load time, all checks on pointers 2195 * previously done by the verifier are invalidated and must be 2196 * performed again, if the helper is used in combination with 2197 * direct packet access. 2198 * Return 2199 * 0 on success, or a negative error in case of failure. 2200 * 2201 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2202 * Description 2203 * Get tunnel metadata. This helper takes a pointer *key* to an 2204 * empty **struct bpf_tunnel_key** of **size**, that will be 2205 * filled with tunnel metadata for the packet associated to *skb*. 2206 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2207 * indicates that the tunnel is based on IPv6 protocol instead of 2208 * IPv4. 2209 * 2210 * The **struct bpf_tunnel_key** is an object that generalizes the 2211 * principal parameters used by various tunneling protocols into a 2212 * single struct. This way, it can be used to easily make a 2213 * decision based on the contents of the encapsulation header, 2214 * "summarized" in this struct. In particular, it holds the IP 2215 * address of the remote end (IPv4 or IPv6, depending on the case) 2216 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2217 * this struct exposes the *key*\ **->tunnel_id**, which is 2218 * generally mapped to a VNI (Virtual Network Identifier), making 2219 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2220 * () helper. 2221 * 2222 * Let's imagine that the following code is part of a program 2223 * attached to the TC ingress interface, on one end of a GRE 2224 * tunnel, and is supposed to filter out all messages coming from 2225 * remote ends with IPv4 address other than 10.0.0.1: 2226 * 2227 * :: 2228 * 2229 * int ret; 2230 * struct bpf_tunnel_key key = {}; 2231 * 2232 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2233 * if (ret < 0) 2234 * return TC_ACT_SHOT; // drop packet 2235 * 2236 * if (key.remote_ipv4 != 0x0a000001) 2237 * return TC_ACT_SHOT; // drop packet 2238 * 2239 * return TC_ACT_OK; // accept packet 2240 * 2241 * This interface can also be used with all encapsulation devices 2242 * that can operate in "collect metadata" mode: instead of having 2243 * one network device per specific configuration, the "collect 2244 * metadata" mode only requires a single device where the 2245 * configuration can be extracted from this helper. 2246 * 2247 * This can be used together with various tunnels such as VXLan, 2248 * Geneve, GRE or IP in IP (IPIP). 2249 * Return 2250 * 0 on success, or a negative error in case of failure. 2251 * 2252 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2253 * Description 2254 * Populate tunnel metadata for packet associated to *skb.* The 2255 * tunnel metadata is set to the contents of *key*, of *size*. The 2256 * *flags* can be set to a combination of the following values: 2257 * 2258 * **BPF_F_TUNINFO_IPV6** 2259 * Indicate that the tunnel is based on IPv6 protocol 2260 * instead of IPv4. 2261 * **BPF_F_ZERO_CSUM_TX** 2262 * For IPv4 packets, add a flag to tunnel metadata 2263 * indicating that checksum computation should be skipped 2264 * and checksum set to zeroes. 2265 * **BPF_F_DONT_FRAGMENT** 2266 * Add a flag to tunnel metadata indicating that the 2267 * packet should not be fragmented. 2268 * **BPF_F_SEQ_NUMBER** 2269 * Add a flag to tunnel metadata indicating that a 2270 * sequence number should be added to tunnel header before 2271 * sending the packet. This flag was added for GRE 2272 * encapsulation, but might be used with other protocols 2273 * as well in the future. 2274 * **BPF_F_NO_TUNNEL_KEY** 2275 * Add a flag to tunnel metadata indicating that no tunnel 2276 * key should be set in the resulting tunnel header. 2277 * 2278 * Here is a typical usage on the transmit path: 2279 * 2280 * :: 2281 * 2282 * struct bpf_tunnel_key key; 2283 * populate key ... 2284 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2285 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2286 * 2287 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2288 * helper for additional information. 2289 * Return 2290 * 0 on success, or a negative error in case of failure. 2291 * 2292 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2293 * Description 2294 * Read the value of a perf event counter. This helper relies on a 2295 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2296 * the perf event counter is selected when *map* is updated with 2297 * perf event file descriptors. The *map* is an array whose size 2298 * is the number of available CPUs, and each cell contains a value 2299 * relative to one CPU. The value to retrieve is indicated by 2300 * *flags*, that contains the index of the CPU to look up, masked 2301 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2302 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2303 * current CPU should be retrieved. 2304 * 2305 * Note that before Linux 4.13, only hardware perf event can be 2306 * retrieved. 2307 * 2308 * Also, be aware that the newer helper 2309 * **bpf_perf_event_read_value**\ () is recommended over 2310 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2311 * quirks where error and counter value are used as a return code 2312 * (which is wrong to do since ranges may overlap). This issue is 2313 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2314 * time provides more features over the **bpf_perf_event_read**\ 2315 * () interface. Please refer to the description of 2316 * **bpf_perf_event_read_value**\ () for details. 2317 * Return 2318 * The value of the perf event counter read from the map, or a 2319 * negative error code in case of failure. 2320 * 2321 * long bpf_redirect(u32 ifindex, u64 flags) 2322 * Description 2323 * Redirect the packet to another net device of index *ifindex*. 2324 * This helper is somewhat similar to **bpf_clone_redirect**\ 2325 * (), except that the packet is not cloned, which provides 2326 * increased performance. 2327 * 2328 * Except for XDP, both ingress and egress interfaces can be used 2329 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2330 * to make the distinction (ingress path is selected if the flag 2331 * is present, egress path otherwise). Currently, XDP only 2332 * supports redirection to the egress interface, and accepts no 2333 * flag at all. 2334 * 2335 * The same effect can also be attained with the more generic 2336 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2337 * redirect target instead of providing it directly to the helper. 2338 * Return 2339 * For XDP, the helper returns **XDP_REDIRECT** on success or 2340 * **XDP_ABORTED** on error. For other program types, the values 2341 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2342 * error. 2343 * 2344 * u32 bpf_get_route_realm(struct sk_buff *skb) 2345 * Description 2346 * Retrieve the realm or the route, that is to say the 2347 * **tclassid** field of the destination for the *skb*. The 2348 * identifier retrieved is a user-provided tag, similar to the 2349 * one used with the net_cls cgroup (see description for 2350 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2351 * held by a route (a destination entry), not by a task. 2352 * 2353 * Retrieving this identifier works with the clsact TC egress hook 2354 * (see also **tc-bpf(8)**), or alternatively on conventional 2355 * classful egress qdiscs, but not on TC ingress path. In case of 2356 * clsact TC egress hook, this has the advantage that, internally, 2357 * the destination entry has not been dropped yet in the transmit 2358 * path. Therefore, the destination entry does not need to be 2359 * artificially held via **netif_keep_dst**\ () for a classful 2360 * qdisc until the *skb* is freed. 2361 * 2362 * This helper is available only if the kernel was compiled with 2363 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2364 * Return 2365 * The realm of the route for the packet associated to *skb*, or 0 2366 * if none was found. 2367 * 2368 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2369 * Description 2370 * Write raw *data* blob into a special BPF perf event held by 2371 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2372 * event must have the following attributes: **PERF_SAMPLE_RAW** 2373 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2374 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2375 * 2376 * The *flags* are used to indicate the index in *map* for which 2377 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2378 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2379 * to indicate that the index of the current CPU core should be 2380 * used. 2381 * 2382 * The value to write, of *size*, is passed through eBPF stack and 2383 * pointed by *data*. 2384 * 2385 * The context of the program *ctx* needs also be passed to the 2386 * helper. 2387 * 2388 * On user space, a program willing to read the values needs to 2389 * call **perf_event_open**\ () on the perf event (either for 2390 * one or for all CPUs) and to store the file descriptor into the 2391 * *map*. This must be done before the eBPF program can send data 2392 * into it. An example is available in file 2393 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2394 * tree (the eBPF program counterpart is in 2395 * *samples/bpf/trace_output_kern.c*). 2396 * 2397 * **bpf_perf_event_output**\ () achieves better performance 2398 * than **bpf_trace_printk**\ () for sharing data with user 2399 * space, and is much better suitable for streaming data from eBPF 2400 * programs. 2401 * 2402 * Note that this helper is not restricted to tracing use cases 2403 * and can be used with programs attached to TC or XDP as well, 2404 * where it allows for passing data to user space listeners. Data 2405 * can be: 2406 * 2407 * * Only custom structs, 2408 * * Only the packet payload, or 2409 * * A combination of both. 2410 * Return 2411 * 0 on success, or a negative error in case of failure. 2412 * 2413 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2414 * Description 2415 * This helper was provided as an easy way to load data from a 2416 * packet. It can be used to load *len* bytes from *offset* from 2417 * the packet associated to *skb*, into the buffer pointed by 2418 * *to*. 2419 * 2420 * Since Linux 4.7, usage of this helper has mostly been replaced 2421 * by "direct packet access", enabling packet data to be 2422 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2423 * pointing respectively to the first byte of packet data and to 2424 * the byte after the last byte of packet data. However, it 2425 * remains useful if one wishes to read large quantities of data 2426 * at once from a packet into the eBPF stack. 2427 * Return 2428 * 0 on success, or a negative error in case of failure. 2429 * 2430 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2431 * Description 2432 * Walk a user or a kernel stack and return its id. To achieve 2433 * this, the helper needs *ctx*, which is a pointer to the context 2434 * on which the tracing program is executed, and a pointer to a 2435 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2436 * 2437 * The last argument, *flags*, holds the number of stack frames to 2438 * skip (from 0 to 255), masked with 2439 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2440 * a combination of the following flags: 2441 * 2442 * **BPF_F_USER_STACK** 2443 * Collect a user space stack instead of a kernel stack. 2444 * **BPF_F_FAST_STACK_CMP** 2445 * Compare stacks by hash only. 2446 * **BPF_F_REUSE_STACKID** 2447 * If two different stacks hash into the same *stackid*, 2448 * discard the old one. 2449 * 2450 * The stack id retrieved is a 32 bit long integer handle which 2451 * can be further combined with other data (including other stack 2452 * ids) and used as a key into maps. This can be useful for 2453 * generating a variety of graphs (such as flame graphs or off-cpu 2454 * graphs). 2455 * 2456 * For walking a stack, this helper is an improvement over 2457 * **bpf_probe_read**\ (), which can be used with unrolled loops 2458 * but is not efficient and consumes a lot of eBPF instructions. 2459 * Instead, **bpf_get_stackid**\ () can collect up to 2460 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2461 * this limit can be controlled with the **sysctl** program, and 2462 * that it should be manually increased in order to profile long 2463 * user stacks (such as stacks for Java programs). To do so, use: 2464 * 2465 * :: 2466 * 2467 * # sysctl kernel.perf_event_max_stack=<new value> 2468 * Return 2469 * The positive or null stack id on success, or a negative error 2470 * in case of failure. 2471 * 2472 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2473 * Description 2474 * Compute a checksum difference, from the raw buffer pointed by 2475 * *from*, of length *from_size* (that must be a multiple of 4), 2476 * towards the raw buffer pointed by *to*, of size *to_size* 2477 * (same remark). An optional *seed* can be added to the value 2478 * (this can be cascaded, the seed may come from a previous call 2479 * to the helper). 2480 * 2481 * This is flexible enough to be used in several ways: 2482 * 2483 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2484 * checksum, it can be used when pushing new data. 2485 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2486 * checksum, it can be used when removing data from a packet. 2487 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2488 * can be used to compute a diff. Note that *from_size* and 2489 * *to_size* do not need to be equal. 2490 * 2491 * This helper can be used in combination with 2492 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2493 * which one can feed in the difference computed with 2494 * **bpf_csum_diff**\ (). 2495 * Return 2496 * The checksum result, or a negative error code in case of 2497 * failure. 2498 * 2499 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2500 * Description 2501 * Retrieve tunnel options metadata for the packet associated to 2502 * *skb*, and store the raw tunnel option data to the buffer *opt* 2503 * of *size*. 2504 * 2505 * This helper can be used with encapsulation devices that can 2506 * operate in "collect metadata" mode (please refer to the related 2507 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2508 * more details). A particular example where this can be used is 2509 * in combination with the Geneve encapsulation protocol, where it 2510 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2511 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2512 * the eBPF program. This allows for full customization of these 2513 * headers. 2514 * Return 2515 * The size of the option data retrieved. 2516 * 2517 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2518 * Description 2519 * Set tunnel options metadata for the packet associated to *skb* 2520 * to the option data contained in the raw buffer *opt* of *size*. 2521 * 2522 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2523 * helper for additional information. 2524 * Return 2525 * 0 on success, or a negative error in case of failure. 2526 * 2527 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2528 * Description 2529 * Change the protocol of the *skb* to *proto*. Currently 2530 * supported are transition from IPv4 to IPv6, and from IPv6 to 2531 * IPv4. The helper takes care of the groundwork for the 2532 * transition, including resizing the socket buffer. The eBPF 2533 * program is expected to fill the new headers, if any, via 2534 * **skb_store_bytes**\ () and to recompute the checksums with 2535 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2536 * (). The main case for this helper is to perform NAT64 2537 * operations out of an eBPF program. 2538 * 2539 * Internally, the GSO type is marked as dodgy so that headers are 2540 * checked and segments are recalculated by the GSO/GRO engine. 2541 * The size for GSO target is adapted as well. 2542 * 2543 * All values for *flags* are reserved for future usage, and must 2544 * be left at zero. 2545 * 2546 * A call to this helper is susceptible to change the underlying 2547 * packet buffer. Therefore, at load time, all checks on pointers 2548 * previously done by the verifier are invalidated and must be 2549 * performed again, if the helper is used in combination with 2550 * direct packet access. 2551 * Return 2552 * 0 on success, or a negative error in case of failure. 2553 * 2554 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2555 * Description 2556 * Change the packet type for the packet associated to *skb*. This 2557 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2558 * the eBPF program does not have a write access to *skb*\ 2559 * **->pkt_type** beside this helper. Using a helper here allows 2560 * for graceful handling of errors. 2561 * 2562 * The major use case is to change incoming *skb*s to 2563 * **PACKET_HOST** in a programmatic way instead of having to 2564 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2565 * example. 2566 * 2567 * Note that *type* only allows certain values. At this time, they 2568 * are: 2569 * 2570 * **PACKET_HOST** 2571 * Packet is for us. 2572 * **PACKET_BROADCAST** 2573 * Send packet to all. 2574 * **PACKET_MULTICAST** 2575 * Send packet to group. 2576 * **PACKET_OTHERHOST** 2577 * Send packet to someone else. 2578 * Return 2579 * 0 on success, or a negative error in case of failure. 2580 * 2581 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2582 * Description 2583 * Check whether *skb* is a descendant of the cgroup2 held by 2584 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2585 * Return 2586 * The return value depends on the result of the test, and can be: 2587 * 2588 * * 0, if the *skb* failed the cgroup2 descendant test. 2589 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2590 * * A negative error code, if an error occurred. 2591 * 2592 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2593 * Description 2594 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2595 * not set, in particular if the hash was cleared due to mangling, 2596 * recompute this hash. Later accesses to the hash can be done 2597 * directly with *skb*\ **->hash**. 2598 * 2599 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2600 * prototype with **bpf_skb_change_proto**\ (), or calling 2601 * **bpf_skb_store_bytes**\ () with the 2602 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2603 * the hash and to trigger a new computation for the next call to 2604 * **bpf_get_hash_recalc**\ (). 2605 * Return 2606 * The 32-bit hash. 2607 * 2608 * u64 bpf_get_current_task(void) 2609 * Description 2610 * Get the current task. 2611 * Return 2612 * A pointer to the current task struct. 2613 * 2614 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2615 * Description 2616 * Attempt in a safe way to write *len* bytes from the buffer 2617 * *src* to *dst* in memory. It only works for threads that are in 2618 * user context, and *dst* must be a valid user space address. 2619 * 2620 * This helper should not be used to implement any kind of 2621 * security mechanism because of TOC-TOU attacks, but rather to 2622 * debug, divert, and manipulate execution of semi-cooperative 2623 * processes. 2624 * 2625 * Keep in mind that this feature is meant for experiments, and it 2626 * has a risk of crashing the system and running programs. 2627 * Therefore, when an eBPF program using this helper is attached, 2628 * a warning including PID and process name is printed to kernel 2629 * logs. 2630 * Return 2631 * 0 on success, or a negative error in case of failure. 2632 * 2633 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2634 * Description 2635 * Check whether the probe is being run is the context of a given 2636 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2637 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2638 * Return 2639 * The return value depends on the result of the test, and can be: 2640 * 2641 * * 1, if current task belongs to the cgroup2. 2642 * * 0, if current task does not belong to the cgroup2. 2643 * * A negative error code, if an error occurred. 2644 * 2645 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2646 * Description 2647 * Resize (trim or grow) the packet associated to *skb* to the 2648 * new *len*. The *flags* are reserved for future usage, and must 2649 * be left at zero. 2650 * 2651 * The basic idea is that the helper performs the needed work to 2652 * change the size of the packet, then the eBPF program rewrites 2653 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2654 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2655 * and others. This helper is a slow path utility intended for 2656 * replies with control messages. And because it is targeted for 2657 * slow path, the helper itself can afford to be slow: it 2658 * implicitly linearizes, unclones and drops offloads from the 2659 * *skb*. 2660 * 2661 * A call to this helper is susceptible to change the underlying 2662 * packet buffer. Therefore, at load time, all checks on pointers 2663 * previously done by the verifier are invalidated and must be 2664 * performed again, if the helper is used in combination with 2665 * direct packet access. 2666 * Return 2667 * 0 on success, or a negative error in case of failure. 2668 * 2669 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2670 * Description 2671 * Pull in non-linear data in case the *skb* is non-linear and not 2672 * all of *len* are part of the linear section. Make *len* bytes 2673 * from *skb* readable and writable. If a zero value is passed for 2674 * *len*, then all bytes in the linear part of *skb* will be made 2675 * readable and writable. 2676 * 2677 * This helper is only needed for reading and writing with direct 2678 * packet access. 2679 * 2680 * For direct packet access, testing that offsets to access 2681 * are within packet boundaries (test on *skb*\ **->data_end**) is 2682 * susceptible to fail if offsets are invalid, or if the requested 2683 * data is in non-linear parts of the *skb*. On failure the 2684 * program can just bail out, or in the case of a non-linear 2685 * buffer, use a helper to make the data available. The 2686 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2687 * the data. Another one consists in using **bpf_skb_pull_data** 2688 * to pull in once the non-linear parts, then retesting and 2689 * eventually access the data. 2690 * 2691 * At the same time, this also makes sure the *skb* is uncloned, 2692 * which is a necessary condition for direct write. As this needs 2693 * to be an invariant for the write part only, the verifier 2694 * detects writes and adds a prologue that is calling 2695 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2696 * the very beginning in case it is indeed cloned. 2697 * 2698 * A call to this helper is susceptible to change the underlying 2699 * packet buffer. Therefore, at load time, all checks on pointers 2700 * previously done by the verifier are invalidated and must be 2701 * performed again, if the helper is used in combination with 2702 * direct packet access. 2703 * Return 2704 * 0 on success, or a negative error in case of failure. 2705 * 2706 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2707 * Description 2708 * Add the checksum *csum* into *skb*\ **->csum** in case the 2709 * driver has supplied a checksum for the entire packet into that 2710 * field. Return an error otherwise. This helper is intended to be 2711 * used in combination with **bpf_csum_diff**\ (), in particular 2712 * when the checksum needs to be updated after data has been 2713 * written into the packet through direct packet access. 2714 * Return 2715 * The checksum on success, or a negative error code in case of 2716 * failure. 2717 * 2718 * void bpf_set_hash_invalid(struct sk_buff *skb) 2719 * Description 2720 * Invalidate the current *skb*\ **->hash**. It can be used after 2721 * mangling on headers through direct packet access, in order to 2722 * indicate that the hash is outdated and to trigger a 2723 * recalculation the next time the kernel tries to access this 2724 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2725 * Return 2726 * void. 2727 * 2728 * long bpf_get_numa_node_id(void) 2729 * Description 2730 * Return the id of the current NUMA node. The primary use case 2731 * for this helper is the selection of sockets for the local NUMA 2732 * node, when the program is attached to sockets using the 2733 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2734 * but the helper is also available to other eBPF program types, 2735 * similarly to **bpf_get_smp_processor_id**\ (). 2736 * Return 2737 * The id of current NUMA node. 2738 * 2739 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2740 * Description 2741 * Grows headroom of packet associated to *skb* and adjusts the 2742 * offset of the MAC header accordingly, adding *len* bytes of 2743 * space. It automatically extends and reallocates memory as 2744 * required. 2745 * 2746 * This helper can be used on a layer 3 *skb* to push a MAC header 2747 * for redirection into a layer 2 device. 2748 * 2749 * All values for *flags* are reserved for future usage, and must 2750 * be left at zero. 2751 * 2752 * A call to this helper is susceptible to change the underlying 2753 * packet buffer. Therefore, at load time, all checks on pointers 2754 * previously done by the verifier are invalidated and must be 2755 * performed again, if the helper is used in combination with 2756 * direct packet access. 2757 * Return 2758 * 0 on success, or a negative error in case of failure. 2759 * 2760 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2761 * Description 2762 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2763 * it is possible to use a negative value for *delta*. This helper 2764 * can be used to prepare the packet for pushing or popping 2765 * headers. 2766 * 2767 * A call to this helper is susceptible to change the underlying 2768 * packet buffer. Therefore, at load time, all checks on pointers 2769 * previously done by the verifier are invalidated and must be 2770 * performed again, if the helper is used in combination with 2771 * direct packet access. 2772 * Return 2773 * 0 on success, or a negative error in case of failure. 2774 * 2775 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2776 * Description 2777 * Copy a NUL terminated string from an unsafe kernel address 2778 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2779 * more details. 2780 * 2781 * Generally, use **bpf_probe_read_user_str**\ () or 2782 * **bpf_probe_read_kernel_str**\ () instead. 2783 * Return 2784 * On success, the strictly positive length of the string, 2785 * including the trailing NUL character. On error, a negative 2786 * value. 2787 * 2788 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2789 * Description 2790 * If the **struct sk_buff** pointed by *skb* has a known socket, 2791 * retrieve the cookie (generated by the kernel) of this socket. 2792 * If no cookie has been set yet, generate a new cookie. Once 2793 * generated, the socket cookie remains stable for the life of the 2794 * socket. This helper can be useful for monitoring per socket 2795 * networking traffic statistics as it provides a global socket 2796 * identifier that can be assumed unique. 2797 * Return 2798 * A 8-byte long unique number on success, or 0 if the socket 2799 * field is missing inside *skb*. 2800 * 2801 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2802 * Description 2803 * Equivalent to bpf_get_socket_cookie() helper that accepts 2804 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2805 * Return 2806 * A 8-byte long unique number. 2807 * 2808 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2809 * Description 2810 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2811 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2812 * Return 2813 * A 8-byte long unique number. 2814 * 2815 * u64 bpf_get_socket_cookie(struct sock *sk) 2816 * Description 2817 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2818 * *sk*, but gets socket from a BTF **struct sock**. This helper 2819 * also works for sleepable programs. 2820 * Return 2821 * A 8-byte long unique number or 0 if *sk* is NULL. 2822 * 2823 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2824 * Description 2825 * Get the owner UID of the socked associated to *skb*. 2826 * Return 2827 * The owner UID of the socket associated to *skb*. If the socket 2828 * is **NULL**, or if it is not a full socket (i.e. if it is a 2829 * time-wait or a request socket instead), **overflowuid** value 2830 * is returned (note that **overflowuid** might also be the actual 2831 * UID value for the socket). 2832 * 2833 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2834 * Description 2835 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2836 * to value *hash*. 2837 * Return 2838 * 0 2839 * 2840 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2841 * Description 2842 * Emulate a call to **setsockopt()** on the socket associated to 2843 * *bpf_socket*, which must be a full socket. The *level* at 2844 * which the option resides and the name *optname* of the option 2845 * must be specified, see **setsockopt(2)** for more information. 2846 * The option value of length *optlen* is pointed by *optval*. 2847 * 2848 * *bpf_socket* should be one of the following: 2849 * 2850 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2851 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2852 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2853 * 2854 * This helper actually implements a subset of **setsockopt()**. 2855 * It supports the following *level*\ s: 2856 * 2857 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2858 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2859 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2860 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2861 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2862 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2863 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2864 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2865 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2866 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2867 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2868 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2869 * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**. 2870 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2871 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2872 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2873 * Return 2874 * 0 on success, or a negative error in case of failure. 2875 * 2876 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2877 * Description 2878 * Grow or shrink the room for data in the packet associated to 2879 * *skb* by *len_diff*, and according to the selected *mode*. 2880 * 2881 * By default, the helper will reset any offloaded checksum 2882 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2883 * by the following flag: 2884 * 2885 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2886 * checksum data of the skb to CHECKSUM_NONE. 2887 * 2888 * There are two supported modes at this time: 2889 * 2890 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2891 * (room space is added or removed between the layer 2 and 2892 * layer 3 headers). 2893 * 2894 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2895 * (room space is added or removed between the layer 3 and 2896 * layer 4 headers). 2897 * 2898 * The following flags are supported at this time: 2899 * 2900 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2901 * Adjusting mss in this way is not allowed for datagrams. 2902 * 2903 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2904 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2905 * Any new space is reserved to hold a tunnel header. 2906 * Configure skb offsets and other fields accordingly. 2907 * 2908 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2909 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2910 * Use with ENCAP_L3 flags to further specify the tunnel type. 2911 * 2912 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2913 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2914 * type; *len* is the length of the inner MAC header. 2915 * 2916 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2917 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2918 * L2 type as Ethernet. 2919 * 2920 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2921 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2922 * Indicate the new IP header version after decapsulating the outer 2923 * IP header. Used when the inner and outer IP versions are different. 2924 * 2925 * A call to this helper is susceptible to change the underlying 2926 * packet buffer. Therefore, at load time, all checks on pointers 2927 * previously done by the verifier are invalidated and must be 2928 * performed again, if the helper is used in combination with 2929 * direct packet access. 2930 * Return 2931 * 0 on success, or a negative error in case of failure. 2932 * 2933 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2934 * Description 2935 * Redirect the packet to the endpoint referenced by *map* at 2936 * index *key*. Depending on its type, this *map* can contain 2937 * references to net devices (for forwarding packets through other 2938 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2939 * but this is only implemented for native XDP (with driver 2940 * support) as of this writing). 2941 * 2942 * The lower two bits of *flags* are used as the return code if 2943 * the map lookup fails. This is so that the return value can be 2944 * one of the XDP program return codes up to **XDP_TX**, as chosen 2945 * by the caller. The higher bits of *flags* can be set to 2946 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2947 * 2948 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2949 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2950 * interface will be excluded when do broadcasting. 2951 * 2952 * See also **bpf_redirect**\ (), which only supports redirecting 2953 * to an ifindex, but doesn't require a map to do so. 2954 * Return 2955 * **XDP_REDIRECT** on success, or the value of the two lower bits 2956 * of the *flags* argument on error. 2957 * 2958 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2959 * Description 2960 * Redirect the packet to the socket referenced by *map* (of type 2961 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2962 * egress interfaces can be used for redirection. The 2963 * **BPF_F_INGRESS** value in *flags* is used to make the 2964 * distinction (ingress path is selected if the flag is present, 2965 * egress path otherwise). This is the only flag supported for now. 2966 * Return 2967 * **SK_PASS** on success, or **SK_DROP** on error. 2968 * 2969 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2970 * Description 2971 * Add an entry to, or update a *map* referencing sockets. The 2972 * *skops* is used as a new value for the entry associated to 2973 * *key*. *flags* is one of: 2974 * 2975 * **BPF_NOEXIST** 2976 * The entry for *key* must not exist in the map. 2977 * **BPF_EXIST** 2978 * The entry for *key* must already exist in the map. 2979 * **BPF_ANY** 2980 * No condition on the existence of the entry for *key*. 2981 * 2982 * If the *map* has eBPF programs (parser and verdict), those will 2983 * be inherited by the socket being added. If the socket is 2984 * already attached to eBPF programs, this results in an error. 2985 * Return 2986 * 0 on success, or a negative error in case of failure. 2987 * 2988 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2989 * Description 2990 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2991 * *delta* (which can be positive or negative). Note that this 2992 * operation modifies the address stored in *xdp_md*\ **->data**, 2993 * so the latter must be loaded only after the helper has been 2994 * called. 2995 * 2996 * The use of *xdp_md*\ **->data_meta** is optional and programs 2997 * are not required to use it. The rationale is that when the 2998 * packet is processed with XDP (e.g. as DoS filter), it is 2999 * possible to push further meta data along with it before passing 3000 * to the stack, and to give the guarantee that an ingress eBPF 3001 * program attached as a TC classifier on the same device can pick 3002 * this up for further post-processing. Since TC works with socket 3003 * buffers, it remains possible to set from XDP the **mark** or 3004 * **priority** pointers, or other pointers for the socket buffer. 3005 * Having this scratch space generic and programmable allows for 3006 * more flexibility as the user is free to store whatever meta 3007 * data they need. 3008 * 3009 * A call to this helper is susceptible to change the underlying 3010 * packet buffer. Therefore, at load time, all checks on pointers 3011 * previously done by the verifier are invalidated and must be 3012 * performed again, if the helper is used in combination with 3013 * direct packet access. 3014 * Return 3015 * 0 on success, or a negative error in case of failure. 3016 * 3017 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 3018 * Description 3019 * Read the value of a perf event counter, and store it into *buf* 3020 * of size *buf_size*. This helper relies on a *map* of type 3021 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 3022 * counter is selected when *map* is updated with perf event file 3023 * descriptors. The *map* is an array whose size is the number of 3024 * available CPUs, and each cell contains a value relative to one 3025 * CPU. The value to retrieve is indicated by *flags*, that 3026 * contains the index of the CPU to look up, masked with 3027 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 3028 * **BPF_F_CURRENT_CPU** to indicate that the value for the 3029 * current CPU should be retrieved. 3030 * 3031 * This helper behaves in a way close to 3032 * **bpf_perf_event_read**\ () helper, save that instead of 3033 * just returning the value observed, it fills the *buf* 3034 * structure. This allows for additional data to be retrieved: in 3035 * particular, the enabled and running times (in *buf*\ 3036 * **->enabled** and *buf*\ **->running**, respectively) are 3037 * copied. In general, **bpf_perf_event_read_value**\ () is 3038 * recommended over **bpf_perf_event_read**\ (), which has some 3039 * ABI issues and provides fewer functionalities. 3040 * 3041 * These values are interesting, because hardware PMU (Performance 3042 * Monitoring Unit) counters are limited resources. When there are 3043 * more PMU based perf events opened than available counters, 3044 * kernel will multiplex these events so each event gets certain 3045 * percentage (but not all) of the PMU time. In case that 3046 * multiplexing happens, the number of samples or counter value 3047 * will not reflect the case compared to when no multiplexing 3048 * occurs. This makes comparison between different runs difficult. 3049 * Typically, the counter value should be normalized before 3050 * comparing to other experiments. The usual normalization is done 3051 * as follows. 3052 * 3053 * :: 3054 * 3055 * normalized_counter = counter * t_enabled / t_running 3056 * 3057 * Where t_enabled is the time enabled for event and t_running is 3058 * the time running for event since last normalization. The 3059 * enabled and running times are accumulated since the perf event 3060 * open. To achieve scaling factor between two invocations of an 3061 * eBPF program, users can use CPU id as the key (which is 3062 * typical for perf array usage model) to remember the previous 3063 * value and do the calculation inside the eBPF program. 3064 * Return 3065 * 0 on success, or a negative error in case of failure. 3066 * 3067 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3068 * Description 3069 * For an eBPF program attached to a perf event, retrieve the 3070 * value of the event counter associated to *ctx* and store it in 3071 * the structure pointed by *buf* and of size *buf_size*. Enabled 3072 * and running times are also stored in the structure (see 3073 * description of helper **bpf_perf_event_read_value**\ () for 3074 * more details). 3075 * Return 3076 * 0 on success, or a negative error in case of failure. 3077 * 3078 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3079 * Description 3080 * Emulate a call to **getsockopt()** on the socket associated to 3081 * *bpf_socket*, which must be a full socket. The *level* at 3082 * which the option resides and the name *optname* of the option 3083 * must be specified, see **getsockopt(2)** for more information. 3084 * The retrieved value is stored in the structure pointed by 3085 * *opval* and of length *optlen*. 3086 * 3087 * *bpf_socket* should be one of the following: 3088 * 3089 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3090 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3091 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3092 * 3093 * This helper actually implements a subset of **getsockopt()**. 3094 * It supports the same set of *optname*\ s that is supported by 3095 * the **bpf_setsockopt**\ () helper. The exceptions are 3096 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3097 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3098 * Return 3099 * 0 on success, or a negative error in case of failure. 3100 * 3101 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3102 * Description 3103 * Used for error injection, this helper uses kprobes to override 3104 * the return value of the probed function, and to set it to *rc*. 3105 * The first argument is the context *regs* on which the kprobe 3106 * works. 3107 * 3108 * This helper works by setting the PC (program counter) 3109 * to an override function which is run in place of the original 3110 * probed function. This means the probed function is not run at 3111 * all. The replacement function just returns with the required 3112 * value. 3113 * 3114 * This helper has security implications, and thus is subject to 3115 * restrictions. It is only available if the kernel was compiled 3116 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3117 * option, and in this case it only works on functions tagged with 3118 * **ALLOW_ERROR_INJECTION** in the kernel code. 3119 * 3120 * Also, the helper is only available for the architectures having 3121 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 3122 * x86 architecture is the only one to support this feature. 3123 * Return 3124 * 0 3125 * 3126 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3127 * Description 3128 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3129 * for the full TCP socket associated to *bpf_sock_ops* to 3130 * *argval*. 3131 * 3132 * The primary use of this field is to determine if there should 3133 * be calls to eBPF programs of type 3134 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3135 * code. A program of the same type can change its value, per 3136 * connection and as necessary, when the connection is 3137 * established. This field is directly accessible for reading, but 3138 * this helper must be used for updates in order to return an 3139 * error if an eBPF program tries to set a callback that is not 3140 * supported in the current kernel. 3141 * 3142 * *argval* is a flag array which can combine these flags: 3143 * 3144 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3145 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3146 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3147 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3148 * 3149 * Therefore, this function can be used to clear a callback flag by 3150 * setting the appropriate bit to zero. e.g. to disable the RTO 3151 * callback: 3152 * 3153 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3154 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3155 * 3156 * Here are some examples of where one could call such eBPF 3157 * program: 3158 * 3159 * * When RTO fires. 3160 * * When a packet is retransmitted. 3161 * * When the connection terminates. 3162 * * When a packet is sent. 3163 * * When a packet is received. 3164 * Return 3165 * Code **-EINVAL** if the socket is not a full TCP socket; 3166 * otherwise, a positive number containing the bits that could not 3167 * be set is returned (which comes down to 0 if all bits were set 3168 * as required). 3169 * 3170 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3171 * Description 3172 * This helper is used in programs implementing policies at the 3173 * socket level. If the message *msg* is allowed to pass (i.e. if 3174 * the verdict eBPF program returns **SK_PASS**), redirect it to 3175 * the socket referenced by *map* (of type 3176 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3177 * egress interfaces can be used for redirection. The 3178 * **BPF_F_INGRESS** value in *flags* is used to make the 3179 * distinction (ingress path is selected if the flag is present, 3180 * egress path otherwise). This is the only flag supported for now. 3181 * Return 3182 * **SK_PASS** on success, or **SK_DROP** on error. 3183 * 3184 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3185 * Description 3186 * For socket policies, apply the verdict of the eBPF program to 3187 * the next *bytes* (number of bytes) of message *msg*. 3188 * 3189 * For example, this helper can be used in the following cases: 3190 * 3191 * * A single **sendmsg**\ () or **sendfile**\ () system call 3192 * contains multiple logical messages that the eBPF program is 3193 * supposed to read and for which it should apply a verdict. 3194 * * An eBPF program only cares to read the first *bytes* of a 3195 * *msg*. If the message has a large payload, then setting up 3196 * and calling the eBPF program repeatedly for all bytes, even 3197 * though the verdict is already known, would create unnecessary 3198 * overhead. 3199 * 3200 * When called from within an eBPF program, the helper sets a 3201 * counter internal to the BPF infrastructure, that is used to 3202 * apply the last verdict to the next *bytes*. If *bytes* is 3203 * smaller than the current data being processed from a 3204 * **sendmsg**\ () or **sendfile**\ () system call, the first 3205 * *bytes* will be sent and the eBPF program will be re-run with 3206 * the pointer for start of data pointing to byte number *bytes* 3207 * **+ 1**. If *bytes* is larger than the current data being 3208 * processed, then the eBPF verdict will be applied to multiple 3209 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3210 * consumed. 3211 * 3212 * Note that if a socket closes with the internal counter holding 3213 * a non-zero value, this is not a problem because data is not 3214 * being buffered for *bytes* and is sent as it is received. 3215 * Return 3216 * 0 3217 * 3218 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3219 * Description 3220 * For socket policies, prevent the execution of the verdict eBPF 3221 * program for message *msg* until *bytes* (byte number) have been 3222 * accumulated. 3223 * 3224 * This can be used when one needs a specific number of bytes 3225 * before a verdict can be assigned, even if the data spans 3226 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3227 * case would be a user calling **sendmsg**\ () repeatedly with 3228 * 1-byte long message segments. Obviously, this is bad for 3229 * performance, but it is still valid. If the eBPF program needs 3230 * *bytes* bytes to validate a header, this helper can be used to 3231 * prevent the eBPF program to be called again until *bytes* have 3232 * been accumulated. 3233 * Return 3234 * 0 3235 * 3236 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3237 * Description 3238 * For socket policies, pull in non-linear data from user space 3239 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3240 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3241 * respectively. 3242 * 3243 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3244 * *msg* it can only parse data that the (**data**, **data_end**) 3245 * pointers have already consumed. For **sendmsg**\ () hooks this 3246 * is likely the first scatterlist element. But for calls relying 3247 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3248 * be the range (**0**, **0**) because the data is shared with 3249 * user space and by default the objective is to avoid allowing 3250 * user space to modify data while (or after) eBPF verdict is 3251 * being decided. This helper can be used to pull in data and to 3252 * set the start and end pointer to given values. Data will be 3253 * copied if necessary (i.e. if data was not linear and if start 3254 * and end pointers do not point to the same chunk). 3255 * 3256 * A call to this helper is susceptible to change the underlying 3257 * packet buffer. Therefore, at load time, all checks on pointers 3258 * previously done by the verifier are invalidated and must be 3259 * performed again, if the helper is used in combination with 3260 * direct packet access. 3261 * 3262 * All values for *flags* are reserved for future usage, and must 3263 * be left at zero. 3264 * Return 3265 * 0 on success, or a negative error in case of failure. 3266 * 3267 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3268 * Description 3269 * Bind the socket associated to *ctx* to the address pointed by 3270 * *addr*, of length *addr_len*. This allows for making outgoing 3271 * connection from the desired IP address, which can be useful for 3272 * example when all processes inside a cgroup should use one 3273 * single IP address on a host that has multiple IP configured. 3274 * 3275 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3276 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3277 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3278 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3279 * behavior and lets the kernel efficiently pick up an unused 3280 * port as long as 4-tuple is unique. Passing non-zero port might 3281 * lead to degraded performance. 3282 * Return 3283 * 0 on success, or a negative error in case of failure. 3284 * 3285 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3286 * Description 3287 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3288 * possible to both shrink and grow the packet tail. 3289 * Shrink done via *delta* being a negative integer. 3290 * 3291 * A call to this helper is susceptible to change the underlying 3292 * packet buffer. Therefore, at load time, all checks on pointers 3293 * previously done by the verifier are invalidated and must be 3294 * performed again, if the helper is used in combination with 3295 * direct packet access. 3296 * Return 3297 * 0 on success, or a negative error in case of failure. 3298 * 3299 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3300 * Description 3301 * Retrieve the XFRM state (IP transform framework, see also 3302 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3303 * 3304 * The retrieved value is stored in the **struct bpf_xfrm_state** 3305 * pointed by *xfrm_state* and of length *size*. 3306 * 3307 * All values for *flags* are reserved for future usage, and must 3308 * be left at zero. 3309 * 3310 * This helper is available only if the kernel was compiled with 3311 * **CONFIG_XFRM** configuration option. 3312 * Return 3313 * 0 on success, or a negative error in case of failure. 3314 * 3315 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3316 * Description 3317 * Return a user or a kernel stack in bpf program provided buffer. 3318 * To achieve this, the helper needs *ctx*, which is a pointer 3319 * to the context on which the tracing program is executed. 3320 * To store the stacktrace, the bpf program provides *buf* with 3321 * a nonnegative *size*. 3322 * 3323 * The last argument, *flags*, holds the number of stack frames to 3324 * skip (from 0 to 255), masked with 3325 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3326 * the following flags: 3327 * 3328 * **BPF_F_USER_STACK** 3329 * Collect a user space stack instead of a kernel stack. 3330 * **BPF_F_USER_BUILD_ID** 3331 * Collect (build_id, file_offset) instead of ips for user 3332 * stack, only valid if **BPF_F_USER_STACK** is also 3333 * specified. 3334 * 3335 * *file_offset* is an offset relative to the beginning 3336 * of the executable or shared object file backing the vma 3337 * which the *ip* falls in. It is *not* an offset relative 3338 * to that object's base address. Accordingly, it must be 3339 * adjusted by adding (sh_addr - sh_offset), where 3340 * sh_{addr,offset} correspond to the executable section 3341 * containing *file_offset* in the object, for comparisons 3342 * to symbols' st_value to be valid. 3343 * 3344 * **bpf_get_stack**\ () can collect up to 3345 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3346 * to sufficient large buffer size. Note that 3347 * this limit can be controlled with the **sysctl** program, and 3348 * that it should be manually increased in order to profile long 3349 * user stacks (such as stacks for Java programs). To do so, use: 3350 * 3351 * :: 3352 * 3353 * # sysctl kernel.perf_event_max_stack=<new value> 3354 * Return 3355 * The non-negative copied *buf* length equal to or less than 3356 * *size* on success, or a negative error in case of failure. 3357 * 3358 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3359 * Description 3360 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3361 * it provides an easy way to load *len* bytes from *offset* 3362 * from the packet associated to *skb*, into the buffer pointed 3363 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3364 * a fifth argument *start_header* exists in order to select a 3365 * base offset to start from. *start_header* can be one of: 3366 * 3367 * **BPF_HDR_START_MAC** 3368 * Base offset to load data from is *skb*'s mac header. 3369 * **BPF_HDR_START_NET** 3370 * Base offset to load data from is *skb*'s network header. 3371 * 3372 * In general, "direct packet access" is the preferred method to 3373 * access packet data, however, this helper is in particular useful 3374 * in socket filters where *skb*\ **->data** does not always point 3375 * to the start of the mac header and where "direct packet access" 3376 * is not available. 3377 * Return 3378 * 0 on success, or a negative error in case of failure. 3379 * 3380 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3381 * Description 3382 * Do FIB lookup in kernel tables using parameters in *params*. 3383 * If lookup is successful and result shows packet is to be 3384 * forwarded, the neighbor tables are searched for the nexthop. 3385 * If successful (ie., FIB lookup shows forwarding and nexthop 3386 * is resolved), the nexthop address is returned in ipv4_dst 3387 * or ipv6_dst based on family, smac is set to mac address of 3388 * egress device, dmac is set to nexthop mac address, rt_metric 3389 * is set to metric from route (IPv4/IPv6 only), and ifindex 3390 * is set to the device index of the nexthop from the FIB lookup. 3391 * 3392 * *plen* argument is the size of the passed in struct. 3393 * *flags* argument can be a combination of one or more of the 3394 * following values: 3395 * 3396 * **BPF_FIB_LOOKUP_DIRECT** 3397 * Do a direct table lookup vs full lookup using FIB 3398 * rules. 3399 * **BPF_FIB_LOOKUP_TBID** 3400 * Used with BPF_FIB_LOOKUP_DIRECT. 3401 * Use the routing table ID present in *params*->tbid 3402 * for the fib lookup. 3403 * **BPF_FIB_LOOKUP_OUTPUT** 3404 * Perform lookup from an egress perspective (default is 3405 * ingress). 3406 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3407 * Skip the neighbour table lookup. *params*->dmac 3408 * and *params*->smac will not be set as output. A common 3409 * use case is to call **bpf_redirect_neigh**\ () after 3410 * doing **bpf_fib_lookup**\ (). 3411 * **BPF_FIB_LOOKUP_SRC** 3412 * Derive and set source IP addr in *params*->ipv{4,6}_src 3413 * for the nexthop. If the src addr cannot be derived, 3414 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3415 * case, *params*->dmac and *params*->smac are not set either. 3416 * **BPF_FIB_LOOKUP_MARK** 3417 * Use the mark present in *params*->mark for the fib lookup. 3418 * This option should not be used with BPF_FIB_LOOKUP_DIRECT, 3419 * as it only has meaning for full lookups. 3420 * 3421 * *ctx* is either **struct xdp_md** for XDP programs or 3422 * **struct sk_buff** tc cls_act programs. 3423 * Return 3424 * * < 0 if any input argument is invalid 3425 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3426 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3427 * packet is not forwarded or needs assist from full stack 3428 * 3429 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3430 * was exceeded and output params->mtu_result contains the MTU. 3431 * 3432 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3433 * Description 3434 * Add an entry to, or update a sockhash *map* referencing sockets. 3435 * The *skops* is used as a new value for the entry associated to 3436 * *key*. *flags* is one of: 3437 * 3438 * **BPF_NOEXIST** 3439 * The entry for *key* must not exist in the map. 3440 * **BPF_EXIST** 3441 * The entry for *key* must already exist in the map. 3442 * **BPF_ANY** 3443 * No condition on the existence of the entry for *key*. 3444 * 3445 * If the *map* has eBPF programs (parser and verdict), those will 3446 * be inherited by the socket being added. If the socket is 3447 * already attached to eBPF programs, this results in an error. 3448 * Return 3449 * 0 on success, or a negative error in case of failure. 3450 * 3451 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3452 * Description 3453 * This helper is used in programs implementing policies at the 3454 * socket level. If the message *msg* is allowed to pass (i.e. if 3455 * the verdict eBPF program returns **SK_PASS**), redirect it to 3456 * the socket referenced by *map* (of type 3457 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3458 * egress interfaces can be used for redirection. The 3459 * **BPF_F_INGRESS** value in *flags* is used to make the 3460 * distinction (ingress path is selected if the flag is present, 3461 * egress path otherwise). This is the only flag supported for now. 3462 * Return 3463 * **SK_PASS** on success, or **SK_DROP** on error. 3464 * 3465 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3466 * Description 3467 * This helper is used in programs implementing policies at the 3468 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3469 * if the verdict eBPF program returns **SK_PASS**), redirect it 3470 * to the socket referenced by *map* (of type 3471 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3472 * egress interfaces can be used for redirection. The 3473 * **BPF_F_INGRESS** value in *flags* is used to make the 3474 * distinction (ingress path is selected if the flag is present, 3475 * egress otherwise). This is the only flag supported for now. 3476 * Return 3477 * **SK_PASS** on success, or **SK_DROP** on error. 3478 * 3479 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3480 * Description 3481 * Encapsulate the packet associated to *skb* within a Layer 3 3482 * protocol header. This header is provided in the buffer at 3483 * address *hdr*, with *len* its size in bytes. *type* indicates 3484 * the protocol of the header and can be one of: 3485 * 3486 * **BPF_LWT_ENCAP_SEG6** 3487 * IPv6 encapsulation with Segment Routing Header 3488 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3489 * the IPv6 header is computed by the kernel. 3490 * **BPF_LWT_ENCAP_SEG6_INLINE** 3491 * Only works if *skb* contains an IPv6 packet. Insert a 3492 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3493 * the IPv6 header. 3494 * **BPF_LWT_ENCAP_IP** 3495 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3496 * must be IPv4 or IPv6, followed by zero or more 3497 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3498 * total bytes in all prepended headers. Please note that 3499 * if **skb_is_gso**\ (*skb*) is true, no more than two 3500 * headers can be prepended, and the inner header, if 3501 * present, should be either GRE or UDP/GUE. 3502 * 3503 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3504 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3505 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3506 * **BPF_PROG_TYPE_LWT_XMIT**. 3507 * 3508 * A call to this helper is susceptible to change the underlying 3509 * packet buffer. Therefore, at load time, all checks on pointers 3510 * previously done by the verifier are invalidated and must be 3511 * performed again, if the helper is used in combination with 3512 * direct packet access. 3513 * Return 3514 * 0 on success, or a negative error in case of failure. 3515 * 3516 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3517 * Description 3518 * Store *len* bytes from address *from* into the packet 3519 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3520 * inside the outermost IPv6 Segment Routing Header can be 3521 * modified through this helper. 3522 * 3523 * A call to this helper is susceptible to change the underlying 3524 * packet buffer. Therefore, at load time, all checks on pointers 3525 * previously done by the verifier are invalidated and must be 3526 * performed again, if the helper is used in combination with 3527 * direct packet access. 3528 * Return 3529 * 0 on success, or a negative error in case of failure. 3530 * 3531 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3532 * Description 3533 * Adjust the size allocated to TLVs in the outermost IPv6 3534 * Segment Routing Header contained in the packet associated to 3535 * *skb*, at position *offset* by *delta* bytes. Only offsets 3536 * after the segments are accepted. *delta* can be as well 3537 * positive (growing) as negative (shrinking). 3538 * 3539 * A call to this helper is susceptible to change the underlying 3540 * packet buffer. Therefore, at load time, all checks on pointers 3541 * previously done by the verifier are invalidated and must be 3542 * performed again, if the helper is used in combination with 3543 * direct packet access. 3544 * Return 3545 * 0 on success, or a negative error in case of failure. 3546 * 3547 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3548 * Description 3549 * Apply an IPv6 Segment Routing action of type *action* to the 3550 * packet associated to *skb*. Each action takes a parameter 3551 * contained at address *param*, and of length *param_len* bytes. 3552 * *action* can be one of: 3553 * 3554 * **SEG6_LOCAL_ACTION_END_X** 3555 * End.X action: Endpoint with Layer-3 cross-connect. 3556 * Type of *param*: **struct in6_addr**. 3557 * **SEG6_LOCAL_ACTION_END_T** 3558 * End.T action: Endpoint with specific IPv6 table lookup. 3559 * Type of *param*: **int**. 3560 * **SEG6_LOCAL_ACTION_END_B6** 3561 * End.B6 action: Endpoint bound to an SRv6 policy. 3562 * Type of *param*: **struct ipv6_sr_hdr**. 3563 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3564 * End.B6.Encap action: Endpoint bound to an SRv6 3565 * encapsulation policy. 3566 * Type of *param*: **struct ipv6_sr_hdr**. 3567 * 3568 * A call to this helper is susceptible to change the underlying 3569 * packet buffer. Therefore, at load time, all checks on pointers 3570 * previously done by the verifier are invalidated and must be 3571 * performed again, if the helper is used in combination with 3572 * direct packet access. 3573 * Return 3574 * 0 on success, or a negative error in case of failure. 3575 * 3576 * long bpf_rc_repeat(void *ctx) 3577 * Description 3578 * This helper is used in programs implementing IR decoding, to 3579 * report a successfully decoded repeat key message. This delays 3580 * the generation of a key up event for previously generated 3581 * key down event. 3582 * 3583 * Some IR protocols like NEC have a special IR message for 3584 * repeating last button, for when a button is held down. 3585 * 3586 * The *ctx* should point to the lirc sample as passed into 3587 * the program. 3588 * 3589 * This helper is only available is the kernel was compiled with 3590 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3591 * "**y**". 3592 * Return 3593 * 0 3594 * 3595 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3596 * Description 3597 * This helper is used in programs implementing IR decoding, to 3598 * report a successfully decoded key press with *scancode*, 3599 * *toggle* value in the given *protocol*. The scancode will be 3600 * translated to a keycode using the rc keymap, and reported as 3601 * an input key down event. After a period a key up event is 3602 * generated. This period can be extended by calling either 3603 * **bpf_rc_keydown**\ () again with the same values, or calling 3604 * **bpf_rc_repeat**\ (). 3605 * 3606 * Some protocols include a toggle bit, in case the button was 3607 * released and pressed again between consecutive scancodes. 3608 * 3609 * The *ctx* should point to the lirc sample as passed into 3610 * the program. 3611 * 3612 * The *protocol* is the decoded protocol number (see 3613 * **enum rc_proto** for some predefined values). 3614 * 3615 * This helper is only available is the kernel was compiled with 3616 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3617 * "**y**". 3618 * Return 3619 * 0 3620 * 3621 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3622 * Description 3623 * Return the cgroup v2 id of the socket associated with the *skb*. 3624 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3625 * helper for cgroup v1 by providing a tag resp. identifier that 3626 * can be matched on or used for map lookups e.g. to implement 3627 * policy. The cgroup v2 id of a given path in the hierarchy is 3628 * exposed in user space through the f_handle API in order to get 3629 * to the same 64-bit id. 3630 * 3631 * This helper can be used on TC egress path, but not on ingress, 3632 * and is available only if the kernel was compiled with the 3633 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3634 * Return 3635 * The id is returned or 0 in case the id could not be retrieved. 3636 * 3637 * u64 bpf_get_current_cgroup_id(void) 3638 * Description 3639 * Get the current cgroup id based on the cgroup within which 3640 * the current task is running. 3641 * Return 3642 * A 64-bit integer containing the current cgroup id based 3643 * on the cgroup within which the current task is running. 3644 * 3645 * void *bpf_get_local_storage(void *map, u64 flags) 3646 * Description 3647 * Get the pointer to the local storage area. 3648 * The type and the size of the local storage is defined 3649 * by the *map* argument. 3650 * The *flags* meaning is specific for each map type, 3651 * and has to be 0 for cgroup local storage. 3652 * 3653 * Depending on the BPF program type, a local storage area 3654 * can be shared between multiple instances of the BPF program, 3655 * running simultaneously. 3656 * 3657 * A user should care about the synchronization by himself. 3658 * For example, by using the **BPF_ATOMIC** instructions to alter 3659 * the shared data. 3660 * Return 3661 * A pointer to the local storage area. 3662 * 3663 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3664 * Description 3665 * Select a **SO_REUSEPORT** socket from a 3666 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3667 * It checks the selected socket is matching the incoming 3668 * request in the socket buffer. 3669 * Return 3670 * 0 on success, or a negative error in case of failure. 3671 * 3672 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3673 * Description 3674 * Return id of cgroup v2 that is ancestor of cgroup associated 3675 * with the *skb* at the *ancestor_level*. The root cgroup is at 3676 * *ancestor_level* zero and each step down the hierarchy 3677 * increments the level. If *ancestor_level* == level of cgroup 3678 * associated with *skb*, then return value will be same as that 3679 * of **bpf_skb_cgroup_id**\ (). 3680 * 3681 * The helper is useful to implement policies based on cgroups 3682 * that are upper in hierarchy than immediate cgroup associated 3683 * with *skb*. 3684 * 3685 * The format of returned id and helper limitations are same as in 3686 * **bpf_skb_cgroup_id**\ (). 3687 * Return 3688 * The id is returned or 0 in case the id could not be retrieved. 3689 * 3690 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3691 * Description 3692 * Look for TCP socket matching *tuple*, optionally in a child 3693 * network namespace *netns*. The return value must be checked, 3694 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3695 * 3696 * The *ctx* should point to the context of the program, such as 3697 * the skb or socket (depending on the hook in use). This is used 3698 * to determine the base network namespace for the lookup. 3699 * 3700 * *tuple_size* must be one of: 3701 * 3702 * **sizeof**\ (*tuple*\ **->ipv4**) 3703 * Look for an IPv4 socket. 3704 * **sizeof**\ (*tuple*\ **->ipv6**) 3705 * Look for an IPv6 socket. 3706 * 3707 * If the *netns* is a negative signed 32-bit integer, then the 3708 * socket lookup table in the netns associated with the *ctx* 3709 * will be used. For the TC hooks, this is the netns of the device 3710 * in the skb. For socket hooks, this is the netns of the socket. 3711 * If *netns* is any other signed 32-bit value greater than or 3712 * equal to zero then it specifies the ID of the netns relative to 3713 * the netns associated with the *ctx*. *netns* values beyond the 3714 * range of 32-bit integers are reserved for future use. 3715 * 3716 * All values for *flags* are reserved for future usage, and must 3717 * be left at zero. 3718 * 3719 * This helper is available only if the kernel was compiled with 3720 * **CONFIG_NET** configuration option. 3721 * Return 3722 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3723 * For sockets with reuseport option, the **struct bpf_sock** 3724 * result is from *reuse*\ **->socks**\ [] using the hash of the 3725 * tuple. 3726 * 3727 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3728 * Description 3729 * Look for UDP socket matching *tuple*, optionally in a child 3730 * network namespace *netns*. The return value must be checked, 3731 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3732 * 3733 * The *ctx* should point to the context of the program, such as 3734 * the skb or socket (depending on the hook in use). This is used 3735 * to determine the base network namespace for the lookup. 3736 * 3737 * *tuple_size* must be one of: 3738 * 3739 * **sizeof**\ (*tuple*\ **->ipv4**) 3740 * Look for an IPv4 socket. 3741 * **sizeof**\ (*tuple*\ **->ipv6**) 3742 * Look for an IPv6 socket. 3743 * 3744 * If the *netns* is a negative signed 32-bit integer, then the 3745 * socket lookup table in the netns associated with the *ctx* 3746 * will be used. For the TC hooks, this is the netns of the device 3747 * in the skb. For socket hooks, this is the netns of the socket. 3748 * If *netns* is any other signed 32-bit value greater than or 3749 * equal to zero then it specifies the ID of the netns relative to 3750 * the netns associated with the *ctx*. *netns* values beyond the 3751 * range of 32-bit integers are reserved for future use. 3752 * 3753 * All values for *flags* are reserved for future usage, and must 3754 * be left at zero. 3755 * 3756 * This helper is available only if the kernel was compiled with 3757 * **CONFIG_NET** configuration option. 3758 * Return 3759 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3760 * For sockets with reuseport option, the **struct bpf_sock** 3761 * result is from *reuse*\ **->socks**\ [] using the hash of the 3762 * tuple. 3763 * 3764 * long bpf_sk_release(void *sock) 3765 * Description 3766 * Release the reference held by *sock*. *sock* must be a 3767 * non-**NULL** pointer that was returned from 3768 * **bpf_sk_lookup_xxx**\ (). 3769 * Return 3770 * 0 on success, or a negative error in case of failure. 3771 * 3772 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3773 * Description 3774 * Push an element *value* in *map*. *flags* is one of: 3775 * 3776 * **BPF_EXIST** 3777 * If the queue/stack is full, the oldest element is 3778 * removed to make room for this. 3779 * Return 3780 * 0 on success, or a negative error in case of failure. 3781 * 3782 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3783 * Description 3784 * Pop an element from *map*. 3785 * Return 3786 * 0 on success, or a negative error in case of failure. 3787 * 3788 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3789 * Description 3790 * Get an element from *map* without removing it. 3791 * Return 3792 * 0 on success, or a negative error in case of failure. 3793 * 3794 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3795 * Description 3796 * For socket policies, insert *len* bytes into *msg* at offset 3797 * *start*. 3798 * 3799 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3800 * *msg* it may want to insert metadata or options into the *msg*. 3801 * This can later be read and used by any of the lower layer BPF 3802 * hooks. 3803 * 3804 * This helper may fail if under memory pressure (a malloc 3805 * fails) in these cases BPF programs will get an appropriate 3806 * error and BPF programs will need to handle them. 3807 * Return 3808 * 0 on success, or a negative error in case of failure. 3809 * 3810 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3811 * Description 3812 * Will remove *len* bytes from a *msg* starting at byte *start*. 3813 * This may result in **ENOMEM** errors under certain situations if 3814 * an allocation and copy are required due to a full ring buffer. 3815 * However, the helper will try to avoid doing the allocation 3816 * if possible. Other errors can occur if input parameters are 3817 * invalid either due to *start* byte not being valid part of *msg* 3818 * payload and/or *pop* value being to large. 3819 * Return 3820 * 0 on success, or a negative error in case of failure. 3821 * 3822 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3823 * Description 3824 * This helper is used in programs implementing IR decoding, to 3825 * report a successfully decoded pointer movement. 3826 * 3827 * The *ctx* should point to the lirc sample as passed into 3828 * the program. 3829 * 3830 * This helper is only available is the kernel was compiled with 3831 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3832 * "**y**". 3833 * Return 3834 * 0 3835 * 3836 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3837 * Description 3838 * Acquire a spinlock represented by the pointer *lock*, which is 3839 * stored as part of a value of a map. Taking the lock allows to 3840 * safely update the rest of the fields in that value. The 3841 * spinlock can (and must) later be released with a call to 3842 * **bpf_spin_unlock**\ (\ *lock*\ ). 3843 * 3844 * Spinlocks in BPF programs come with a number of restrictions 3845 * and constraints: 3846 * 3847 * * **bpf_spin_lock** objects are only allowed inside maps of 3848 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3849 * list could be extended in the future). 3850 * * BTF description of the map is mandatory. 3851 * * The BPF program can take ONE lock at a time, since taking two 3852 * or more could cause dead locks. 3853 * * Only one **struct bpf_spin_lock** is allowed per map element. 3854 * * When the lock is taken, calls (either BPF to BPF or helpers) 3855 * are not allowed. 3856 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3857 * allowed inside a spinlock-ed region. 3858 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3859 * the lock, on all execution paths, before it returns. 3860 * * The BPF program can access **struct bpf_spin_lock** only via 3861 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3862 * helpers. Loading or storing data into the **struct 3863 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3864 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3865 * of the map value must be a struct and have **struct 3866 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3867 * Nested lock inside another struct is not allowed. 3868 * * The **struct bpf_spin_lock** *lock* field in a map value must 3869 * be aligned on a multiple of 4 bytes in that value. 3870 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3871 * the **bpf_spin_lock** field to user space. 3872 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3873 * a BPF program, do not update the **bpf_spin_lock** field. 3874 * * **bpf_spin_lock** cannot be on the stack or inside a 3875 * networking packet (it can only be inside of a map values). 3876 * * **bpf_spin_lock** is available to root only. 3877 * * Tracing programs and socket filter programs cannot use 3878 * **bpf_spin_lock**\ () due to insufficient preemption checks 3879 * (but this may change in the future). 3880 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3881 * Return 3882 * 0 3883 * 3884 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3885 * Description 3886 * Release the *lock* previously locked by a call to 3887 * **bpf_spin_lock**\ (\ *lock*\ ). 3888 * Return 3889 * 0 3890 * 3891 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3892 * Description 3893 * This helper gets a **struct bpf_sock** pointer such 3894 * that all the fields in this **bpf_sock** can be accessed. 3895 * Return 3896 * A **struct bpf_sock** pointer on success, or **NULL** in 3897 * case of failure. 3898 * 3899 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3900 * Description 3901 * This helper gets a **struct bpf_tcp_sock** pointer from a 3902 * **struct bpf_sock** pointer. 3903 * Return 3904 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3905 * case of failure. 3906 * 3907 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3908 * Description 3909 * Set ECN (Explicit Congestion Notification) field of IP header 3910 * to **CE** (Congestion Encountered) if current value is **ECT** 3911 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3912 * and IPv4. 3913 * Return 3914 * 1 if the **CE** flag is set (either by the current helper call 3915 * or because it was already present), 0 if it is not set. 3916 * 3917 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3918 * Description 3919 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3920 * **bpf_sk_release**\ () is unnecessary and not allowed. 3921 * Return 3922 * A **struct bpf_sock** pointer on success, or **NULL** in 3923 * case of failure. 3924 * 3925 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3926 * Description 3927 * Look for TCP socket matching *tuple*, optionally in a child 3928 * network namespace *netns*. The return value must be checked, 3929 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3930 * 3931 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3932 * that it also returns timewait or request sockets. Use 3933 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3934 * full structure. 3935 * 3936 * This helper is available only if the kernel was compiled with 3937 * **CONFIG_NET** configuration option. 3938 * Return 3939 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3940 * For sockets with reuseport option, the **struct bpf_sock** 3941 * result is from *reuse*\ **->socks**\ [] using the hash of the 3942 * tuple. 3943 * 3944 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3945 * Description 3946 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3947 * the listening socket in *sk*. 3948 * 3949 * *iph* points to the start of the IPv4 or IPv6 header, while 3950 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3951 * **sizeof**\ (**struct ipv6hdr**). 3952 * 3953 * *th* points to the start of the TCP header, while *th_len* 3954 * contains the length of the TCP header (at least 3955 * **sizeof**\ (**struct tcphdr**)). 3956 * Return 3957 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3958 * error otherwise. 3959 * 3960 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3961 * Description 3962 * Get name of sysctl in /proc/sys/ and copy it into provided by 3963 * program buffer *buf* of size *buf_len*. 3964 * 3965 * The buffer is always NUL terminated, unless it's zero-sized. 3966 * 3967 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3968 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3969 * only (e.g. "tcp_mem"). 3970 * Return 3971 * Number of character copied (not including the trailing NUL). 3972 * 3973 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3974 * truncated name in this case). 3975 * 3976 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3977 * Description 3978 * Get current value of sysctl as it is presented in /proc/sys 3979 * (incl. newline, etc), and copy it as a string into provided 3980 * by program buffer *buf* of size *buf_len*. 3981 * 3982 * The whole value is copied, no matter what file position user 3983 * space issued e.g. sys_read at. 3984 * 3985 * The buffer is always NUL terminated, unless it's zero-sized. 3986 * Return 3987 * Number of character copied (not including the trailing NUL). 3988 * 3989 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3990 * truncated name in this case). 3991 * 3992 * **-EINVAL** if current value was unavailable, e.g. because 3993 * sysctl is uninitialized and read returns -EIO for it. 3994 * 3995 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3996 * Description 3997 * Get new value being written by user space to sysctl (before 3998 * the actual write happens) and copy it as a string into 3999 * provided by program buffer *buf* of size *buf_len*. 4000 * 4001 * User space may write new value at file position > 0. 4002 * 4003 * The buffer is always NUL terminated, unless it's zero-sized. 4004 * Return 4005 * Number of character copied (not including the trailing NUL). 4006 * 4007 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 4008 * truncated name in this case). 4009 * 4010 * **-EINVAL** if sysctl is being read. 4011 * 4012 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 4013 * Description 4014 * Override new value being written by user space to sysctl with 4015 * value provided by program in buffer *buf* of size *buf_len*. 4016 * 4017 * *buf* should contain a string in same form as provided by user 4018 * space on sysctl write. 4019 * 4020 * User space may write new value at file position > 0. To override 4021 * the whole sysctl value file position should be set to zero. 4022 * Return 4023 * 0 on success. 4024 * 4025 * **-E2BIG** if the *buf_len* is too big. 4026 * 4027 * **-EINVAL** if sysctl is being read. 4028 * 4029 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 4030 * Description 4031 * Convert the initial part of the string from buffer *buf* of 4032 * size *buf_len* to a long integer according to the given base 4033 * and save the result in *res*. 4034 * 4035 * The string may begin with an arbitrary amount of white space 4036 * (as determined by **isspace**\ (3)) followed by a single 4037 * optional '**-**' sign. 4038 * 4039 * Five least significant bits of *flags* encode base, other bits 4040 * are currently unused. 4041 * 4042 * Base must be either 8, 10, 16 or 0 to detect it automatically 4043 * similar to user space **strtol**\ (3). 4044 * Return 4045 * Number of characters consumed on success. Must be positive but 4046 * no more than *buf_len*. 4047 * 4048 * **-EINVAL** if no valid digits were found or unsupported base 4049 * was provided. 4050 * 4051 * **-ERANGE** if resulting value was out of range. 4052 * 4053 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 4054 * Description 4055 * Convert the initial part of the string from buffer *buf* of 4056 * size *buf_len* to an unsigned long integer according to the 4057 * given base and save the result in *res*. 4058 * 4059 * The string may begin with an arbitrary amount of white space 4060 * (as determined by **isspace**\ (3)). 4061 * 4062 * Five least significant bits of *flags* encode base, other bits 4063 * are currently unused. 4064 * 4065 * Base must be either 8, 10, 16 or 0 to detect it automatically 4066 * similar to user space **strtoul**\ (3). 4067 * Return 4068 * Number of characters consumed on success. Must be positive but 4069 * no more than *buf_len*. 4070 * 4071 * **-EINVAL** if no valid digits were found or unsupported base 4072 * was provided. 4073 * 4074 * **-ERANGE** if resulting value was out of range. 4075 * 4076 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4077 * Description 4078 * Get a bpf-local-storage from a *sk*. 4079 * 4080 * Logically, it could be thought of getting the value from 4081 * a *map* with *sk* as the **key**. From this 4082 * perspective, the usage is not much different from 4083 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4084 * helper enforces the key must be a full socket and the map must 4085 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4086 * 4087 * Underneath, the value is stored locally at *sk* instead of 4088 * the *map*. The *map* is used as the bpf-local-storage 4089 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4090 * searched against all bpf-local-storages residing at *sk*. 4091 * 4092 * *sk* is a kernel **struct sock** pointer for LSM program. 4093 * *sk* is a **struct bpf_sock** pointer for other program types. 4094 * 4095 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4096 * used such that a new bpf-local-storage will be 4097 * created if one does not exist. *value* can be used 4098 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4099 * the initial value of a bpf-local-storage. If *value* is 4100 * **NULL**, the new bpf-local-storage will be zero initialized. 4101 * Return 4102 * A bpf-local-storage pointer is returned on success. 4103 * 4104 * **NULL** if not found or there was an error in adding 4105 * a new bpf-local-storage. 4106 * 4107 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4108 * Description 4109 * Delete a bpf-local-storage from a *sk*. 4110 * Return 4111 * 0 on success. 4112 * 4113 * **-ENOENT** if the bpf-local-storage cannot be found. 4114 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4115 * 4116 * long bpf_send_signal(u32 sig) 4117 * Description 4118 * Send signal *sig* to the process of the current task. 4119 * The signal may be delivered to any of this process's threads. 4120 * Return 4121 * 0 on success or successfully queued. 4122 * 4123 * **-EBUSY** if work queue under nmi is full. 4124 * 4125 * **-EINVAL** if *sig* is invalid. 4126 * 4127 * **-EPERM** if no permission to send the *sig*. 4128 * 4129 * **-EAGAIN** if bpf program can try again. 4130 * 4131 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4132 * Description 4133 * Try to issue a SYN cookie for the packet with corresponding 4134 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4135 * 4136 * *iph* points to the start of the IPv4 or IPv6 header, while 4137 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4138 * **sizeof**\ (**struct ipv6hdr**). 4139 * 4140 * *th* points to the start of the TCP header, while *th_len* 4141 * contains the length of the TCP header with options (at least 4142 * **sizeof**\ (**struct tcphdr**)). 4143 * Return 4144 * On success, lower 32 bits hold the generated SYN cookie in 4145 * followed by 16 bits which hold the MSS value for that cookie, 4146 * and the top 16 bits are unused. 4147 * 4148 * On failure, the returned value is one of the following: 4149 * 4150 * **-EINVAL** SYN cookie cannot be issued due to error 4151 * 4152 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4153 * 4154 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4155 * 4156 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4157 * 4158 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4159 * Description 4160 * Write raw *data* blob into a special BPF perf event held by 4161 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4162 * event must have the following attributes: **PERF_SAMPLE_RAW** 4163 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4164 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4165 * 4166 * The *flags* are used to indicate the index in *map* for which 4167 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4168 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4169 * to indicate that the index of the current CPU core should be 4170 * used. 4171 * 4172 * The value to write, of *size*, is passed through eBPF stack and 4173 * pointed by *data*. 4174 * 4175 * *ctx* is a pointer to in-kernel struct sk_buff. 4176 * 4177 * This helper is similar to **bpf_perf_event_output**\ () but 4178 * restricted to raw_tracepoint bpf programs. 4179 * Return 4180 * 0 on success, or a negative error in case of failure. 4181 * 4182 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4183 * Description 4184 * Safely attempt to read *size* bytes from user space address 4185 * *unsafe_ptr* and store the data in *dst*. 4186 * Return 4187 * 0 on success, or a negative error in case of failure. 4188 * 4189 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4190 * Description 4191 * Safely attempt to read *size* bytes from kernel space address 4192 * *unsafe_ptr* and store the data in *dst*. 4193 * Return 4194 * 0 on success, or a negative error in case of failure. 4195 * 4196 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4197 * Description 4198 * Copy a NUL terminated string from an unsafe user address 4199 * *unsafe_ptr* to *dst*. The *size* should include the 4200 * terminating NUL byte. In case the string length is smaller than 4201 * *size*, the target is not padded with further NUL bytes. If the 4202 * string length is larger than *size*, just *size*-1 bytes are 4203 * copied and the last byte is set to NUL. 4204 * 4205 * On success, returns the number of bytes that were written, 4206 * including the terminal NUL. This makes this helper useful in 4207 * tracing programs for reading strings, and more importantly to 4208 * get its length at runtime. See the following snippet: 4209 * 4210 * :: 4211 * 4212 * SEC("kprobe/sys_open") 4213 * void bpf_sys_open(struct pt_regs *ctx) 4214 * { 4215 * char buf[PATHLEN]; // PATHLEN is defined to 256 4216 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4217 * ctx->di); 4218 * 4219 * // Consume buf, for example push it to 4220 * // userspace via bpf_perf_event_output(); we 4221 * // can use res (the string length) as event 4222 * // size, after checking its boundaries. 4223 * } 4224 * 4225 * In comparison, using **bpf_probe_read_user**\ () helper here 4226 * instead to read the string would require to estimate the length 4227 * at compile time, and would often result in copying more memory 4228 * than necessary. 4229 * 4230 * Another useful use case is when parsing individual process 4231 * arguments or individual environment variables navigating 4232 * *current*\ **->mm->arg_start** and *current*\ 4233 * **->mm->env_start**: using this helper and the return value, 4234 * one can quickly iterate at the right offset of the memory area. 4235 * Return 4236 * On success, the strictly positive length of the output string, 4237 * including the trailing NUL character. On error, a negative 4238 * value. 4239 * 4240 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4241 * Description 4242 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4243 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4244 * Return 4245 * On success, the strictly positive length of the string, including 4246 * the trailing NUL character. On error, a negative value. 4247 * 4248 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4249 * Description 4250 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4251 * *rcv_nxt* is the ack_seq to be sent out. 4252 * Return 4253 * 0 on success, or a negative error in case of failure. 4254 * 4255 * long bpf_send_signal_thread(u32 sig) 4256 * Description 4257 * Send signal *sig* to the thread corresponding to the current task. 4258 * Return 4259 * 0 on success or successfully queued. 4260 * 4261 * **-EBUSY** if work queue under nmi is full. 4262 * 4263 * **-EINVAL** if *sig* is invalid. 4264 * 4265 * **-EPERM** if no permission to send the *sig*. 4266 * 4267 * **-EAGAIN** if bpf program can try again. 4268 * 4269 * u64 bpf_jiffies64(void) 4270 * Description 4271 * Obtain the 64bit jiffies 4272 * Return 4273 * The 64 bit jiffies 4274 * 4275 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4276 * Description 4277 * For an eBPF program attached to a perf event, retrieve the 4278 * branch records (**struct perf_branch_entry**) associated to *ctx* 4279 * and store it in the buffer pointed by *buf* up to size 4280 * *size* bytes. 4281 * Return 4282 * On success, number of bytes written to *buf*. On error, a 4283 * negative value. 4284 * 4285 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4286 * instead return the number of bytes required to store all the 4287 * branch entries. If this flag is set, *buf* may be NULL. 4288 * 4289 * **-EINVAL** if arguments invalid or **size** not a multiple 4290 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4291 * 4292 * **-ENOENT** if architecture does not support branch records. 4293 * 4294 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4295 * Description 4296 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4297 * *namespace* will be returned in *nsdata*. 4298 * Return 4299 * 0 on success, or one of the following in case of failure: 4300 * 4301 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4302 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4303 * 4304 * **-ENOENT** if pidns does not exists for the current task. 4305 * 4306 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4307 * Description 4308 * Write raw *data* blob into a special BPF perf event held by 4309 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4310 * event must have the following attributes: **PERF_SAMPLE_RAW** 4311 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4312 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4313 * 4314 * The *flags* are used to indicate the index in *map* for which 4315 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4316 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4317 * to indicate that the index of the current CPU core should be 4318 * used. 4319 * 4320 * The value to write, of *size*, is passed through eBPF stack and 4321 * pointed by *data*. 4322 * 4323 * *ctx* is a pointer to in-kernel struct xdp_buff. 4324 * 4325 * This helper is similar to **bpf_perf_eventoutput**\ () but 4326 * restricted to raw_tracepoint bpf programs. 4327 * Return 4328 * 0 on success, or a negative error in case of failure. 4329 * 4330 * u64 bpf_get_netns_cookie(void *ctx) 4331 * Description 4332 * Retrieve the cookie (generated by the kernel) of the network 4333 * namespace the input *ctx* is associated with. The network 4334 * namespace cookie remains stable for its lifetime and provides 4335 * a global identifier that can be assumed unique. If *ctx* is 4336 * NULL, then the helper returns the cookie for the initial 4337 * network namespace. The cookie itself is very similar to that 4338 * of **bpf_get_socket_cookie**\ () helper, but for network 4339 * namespaces instead of sockets. 4340 * Return 4341 * A 8-byte long opaque number. 4342 * 4343 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4344 * Description 4345 * Return id of cgroup v2 that is ancestor of the cgroup associated 4346 * with the current task at the *ancestor_level*. The root cgroup 4347 * is at *ancestor_level* zero and each step down the hierarchy 4348 * increments the level. If *ancestor_level* == level of cgroup 4349 * associated with the current task, then return value will be the 4350 * same as that of **bpf_get_current_cgroup_id**\ (). 4351 * 4352 * The helper is useful to implement policies based on cgroups 4353 * that are upper in hierarchy than immediate cgroup associated 4354 * with the current task. 4355 * 4356 * The format of returned id and helper limitations are same as in 4357 * **bpf_get_current_cgroup_id**\ (). 4358 * Return 4359 * The id is returned or 0 in case the id could not be retrieved. 4360 * 4361 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4362 * Description 4363 * Helper is overloaded depending on BPF program type. This 4364 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4365 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4366 * 4367 * Assign the *sk* to the *skb*. When combined with appropriate 4368 * routing configuration to receive the packet towards the socket, 4369 * will cause *skb* to be delivered to the specified socket. 4370 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4371 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4372 * interfere with successful delivery to the socket. 4373 * 4374 * This operation is only valid from TC ingress path. 4375 * 4376 * The *flags* argument must be zero. 4377 * Return 4378 * 0 on success, or a negative error in case of failure: 4379 * 4380 * **-EINVAL** if specified *flags* are not supported. 4381 * 4382 * **-ENOENT** if the socket is unavailable for assignment. 4383 * 4384 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4385 * 4386 * **-EOPNOTSUPP** if the operation is not supported, for example 4387 * a call from outside of TC ingress. 4388 * 4389 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4390 * Description 4391 * Helper is overloaded depending on BPF program type. This 4392 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4393 * 4394 * Select the *sk* as a result of a socket lookup. 4395 * 4396 * For the operation to succeed passed socket must be compatible 4397 * with the packet description provided by the *ctx* object. 4398 * 4399 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4400 * be an exact match. While IP family (**AF_INET** or 4401 * **AF_INET6**) must be compatible, that is IPv6 sockets 4402 * that are not v6-only can be selected for IPv4 packets. 4403 * 4404 * Only TCP listeners and UDP unconnected sockets can be 4405 * selected. *sk* can also be NULL to reset any previous 4406 * selection. 4407 * 4408 * *flags* argument can combination of following values: 4409 * 4410 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4411 * socket selection, potentially done by a BPF program 4412 * that ran before us. 4413 * 4414 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4415 * load-balancing within reuseport group for the socket 4416 * being selected. 4417 * 4418 * On success *ctx->sk* will point to the selected socket. 4419 * 4420 * Return 4421 * 0 on success, or a negative errno in case of failure. 4422 * 4423 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4424 * not compatible with packet family (*ctx->family*). 4425 * 4426 * * **-EEXIST** if socket has been already selected, 4427 * potentially by another program, and 4428 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4429 * 4430 * * **-EINVAL** if unsupported flags were specified. 4431 * 4432 * * **-EPROTOTYPE** if socket L4 protocol 4433 * (*sk->protocol*) doesn't match packet protocol 4434 * (*ctx->protocol*). 4435 * 4436 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4437 * state (TCP listening or UDP unconnected). 4438 * 4439 * u64 bpf_ktime_get_boot_ns(void) 4440 * Description 4441 * Return the time elapsed since system boot, in nanoseconds. 4442 * Does include the time the system was suspended. 4443 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4444 * Return 4445 * Current *ktime*. 4446 * 4447 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4448 * Description 4449 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4450 * out the format string. 4451 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4452 * the format string itself. The *data* and *data_len* are format string 4453 * arguments. The *data* are a **u64** array and corresponding format string 4454 * values are stored in the array. For strings and pointers where pointees 4455 * are accessed, only the pointer values are stored in the *data* array. 4456 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4457 * 4458 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4459 * Reading kernel memory may fail due to either invalid address or 4460 * valid address but requiring a major memory fault. If reading kernel memory 4461 * fails, the string for **%s** will be an empty string, and the ip 4462 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4463 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4464 * Return 4465 * 0 on success, or a negative error in case of failure: 4466 * 4467 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4468 * by returning 1 from bpf program. 4469 * 4470 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4471 * 4472 * **-E2BIG** if *fmt* contains too many format specifiers. 4473 * 4474 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4475 * 4476 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4477 * Description 4478 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4479 * The *m* represents the seq_file. The *data* and *len* represent the 4480 * data to write in bytes. 4481 * Return 4482 * 0 on success, or a negative error in case of failure: 4483 * 4484 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4485 * 4486 * u64 bpf_sk_cgroup_id(void *sk) 4487 * Description 4488 * Return the cgroup v2 id of the socket *sk*. 4489 * 4490 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4491 * returned from **bpf_sk_lookup_xxx**\ (), 4492 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4493 * same as in **bpf_skb_cgroup_id**\ (). 4494 * 4495 * This helper is available only if the kernel was compiled with 4496 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4497 * Return 4498 * The id is returned or 0 in case the id could not be retrieved. 4499 * 4500 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4501 * Description 4502 * Return id of cgroup v2 that is ancestor of cgroup associated 4503 * with the *sk* at the *ancestor_level*. The root cgroup is at 4504 * *ancestor_level* zero and each step down the hierarchy 4505 * increments the level. If *ancestor_level* == level of cgroup 4506 * associated with *sk*, then return value will be same as that 4507 * of **bpf_sk_cgroup_id**\ (). 4508 * 4509 * The helper is useful to implement policies based on cgroups 4510 * that are upper in hierarchy than immediate cgroup associated 4511 * with *sk*. 4512 * 4513 * The format of returned id and helper limitations are same as in 4514 * **bpf_sk_cgroup_id**\ (). 4515 * Return 4516 * The id is returned or 0 in case the id could not be retrieved. 4517 * 4518 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4519 * Description 4520 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4521 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4522 * of new data availability is sent. 4523 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4524 * of new data availability is sent unconditionally. 4525 * If **0** is specified in *flags*, an adaptive notification 4526 * of new data availability is sent. 4527 * 4528 * An adaptive notification is a notification sent whenever the user-space 4529 * process has caught up and consumed all available payloads. In case the user-space 4530 * process is still processing a previous payload, then no notification is needed 4531 * as it will process the newly added payload automatically. 4532 * Return 4533 * 0 on success, or a negative error in case of failure. 4534 * 4535 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4536 * Description 4537 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4538 * *flags* must be 0. 4539 * Return 4540 * Valid pointer with *size* bytes of memory available; NULL, 4541 * otherwise. 4542 * 4543 * void bpf_ringbuf_submit(void *data, u64 flags) 4544 * Description 4545 * Submit reserved ring buffer sample, pointed to by *data*. 4546 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4547 * of new data availability is sent. 4548 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4549 * of new data availability is sent unconditionally. 4550 * If **0** is specified in *flags*, an adaptive notification 4551 * of new data availability is sent. 4552 * 4553 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4554 * Return 4555 * Nothing. Always succeeds. 4556 * 4557 * void bpf_ringbuf_discard(void *data, u64 flags) 4558 * Description 4559 * Discard reserved ring buffer sample, pointed to by *data*. 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 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4568 * Return 4569 * Nothing. Always succeeds. 4570 * 4571 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4572 * Description 4573 * Query various characteristics of provided ring buffer. What 4574 * exactly is queries is determined by *flags*: 4575 * 4576 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4577 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4578 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4579 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4580 * 4581 * Data returned is just a momentary snapshot of actual values 4582 * and could be inaccurate, so this facility should be used to 4583 * power heuristics and for reporting, not to make 100% correct 4584 * calculation. 4585 * Return 4586 * Requested value, or 0, if *flags* are not recognized. 4587 * 4588 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4589 * Description 4590 * Change the skbs checksum level by one layer up or down, or 4591 * reset it entirely to none in order to have the stack perform 4592 * checksum validation. The level is applicable to the following 4593 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4594 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4595 * through **bpf_skb_adjust_room**\ () helper with passing in 4596 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4597 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4598 * the UDP header is removed. Similarly, an encap of the latter 4599 * into the former could be accompanied by a helper call to 4600 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4601 * skb is still intended to be processed in higher layers of the 4602 * stack instead of just egressing at tc. 4603 * 4604 * There are three supported level settings at this time: 4605 * 4606 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4607 * with CHECKSUM_UNNECESSARY. 4608 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4609 * with CHECKSUM_UNNECESSARY. 4610 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4611 * sets CHECKSUM_NONE to force checksum validation by the stack. 4612 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4613 * skb->csum_level. 4614 * Return 4615 * 0 on success, or a negative error in case of failure. In the 4616 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4617 * is returned or the error code -EACCES in case the skb is not 4618 * subject to CHECKSUM_UNNECESSARY. 4619 * 4620 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4621 * Description 4622 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4623 * Return 4624 * *sk* if casting is valid, or **NULL** otherwise. 4625 * 4626 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4627 * Description 4628 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4629 * Return 4630 * *sk* if casting is valid, or **NULL** otherwise. 4631 * 4632 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4633 * Description 4634 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4635 * Return 4636 * *sk* if casting is valid, or **NULL** otherwise. 4637 * 4638 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4639 * Description 4640 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4641 * Return 4642 * *sk* if casting is valid, or **NULL** otherwise. 4643 * 4644 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4645 * Description 4646 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4647 * Return 4648 * *sk* if casting is valid, or **NULL** otherwise. 4649 * 4650 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4651 * Description 4652 * Return a user or a kernel stack in bpf program provided buffer. 4653 * Note: the user stack will only be populated if the *task* is 4654 * the current task; all other tasks will return -EOPNOTSUPP. 4655 * To achieve this, the helper needs *task*, which is a valid 4656 * pointer to **struct task_struct**. To store the stacktrace, the 4657 * bpf program provides *buf* with a nonnegative *size*. 4658 * 4659 * The last argument, *flags*, holds the number of stack frames to 4660 * skip (from 0 to 255), masked with 4661 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4662 * the following flags: 4663 * 4664 * **BPF_F_USER_STACK** 4665 * Collect a user space stack instead of a kernel stack. 4666 * The *task* must be the current task. 4667 * **BPF_F_USER_BUILD_ID** 4668 * Collect buildid+offset instead of ips for user stack, 4669 * only valid if **BPF_F_USER_STACK** is also specified. 4670 * 4671 * **bpf_get_task_stack**\ () can collect up to 4672 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4673 * to sufficient large buffer size. Note that 4674 * this limit can be controlled with the **sysctl** program, and 4675 * that it should be manually increased in order to profile long 4676 * user stacks (such as stacks for Java programs). To do so, use: 4677 * 4678 * :: 4679 * 4680 * # sysctl kernel.perf_event_max_stack=<new value> 4681 * Return 4682 * The non-negative copied *buf* length equal to or less than 4683 * *size* on success, or a negative error in case of failure. 4684 * 4685 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4686 * Description 4687 * Load header option. Support reading a particular TCP header 4688 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4689 * 4690 * If *flags* is 0, it will search the option from the 4691 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4692 * has details on what skb_data contains under different 4693 * *skops*\ **->op**. 4694 * 4695 * The first byte of the *searchby_res* specifies the 4696 * kind that it wants to search. 4697 * 4698 * If the searching kind is an experimental kind 4699 * (i.e. 253 or 254 according to RFC6994). It also 4700 * needs to specify the "magic" which is either 4701 * 2 bytes or 4 bytes. It then also needs to 4702 * specify the size of the magic by using 4703 * the 2nd byte which is "kind-length" of a TCP 4704 * header option and the "kind-length" also 4705 * includes the first 2 bytes "kind" and "kind-length" 4706 * itself as a normal TCP header option also does. 4707 * 4708 * For example, to search experimental kind 254 with 4709 * 2 byte magic 0xeB9F, the searchby_res should be 4710 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4711 * 4712 * To search for the standard window scale option (3), 4713 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4714 * Note, kind-length must be 0 for regular option. 4715 * 4716 * Searching for No-Op (0) and End-of-Option-List (1) are 4717 * not supported. 4718 * 4719 * *len* must be at least 2 bytes which is the minimal size 4720 * of a header option. 4721 * 4722 * Supported flags: 4723 * 4724 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4725 * saved_syn packet or the just-received syn packet. 4726 * 4727 * Return 4728 * > 0 when found, the header option is copied to *searchby_res*. 4729 * The return value is the total length copied. On failure, a 4730 * negative error code is returned: 4731 * 4732 * **-EINVAL** if a parameter is invalid. 4733 * 4734 * **-ENOMSG** if the option is not found. 4735 * 4736 * **-ENOENT** if no syn packet is available when 4737 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4738 * 4739 * **-ENOSPC** if there is not enough space. Only *len* number of 4740 * bytes are copied. 4741 * 4742 * **-EFAULT** on failure to parse the header options in the 4743 * packet. 4744 * 4745 * **-EPERM** if the helper cannot be used under the current 4746 * *skops*\ **->op**. 4747 * 4748 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4749 * Description 4750 * Store header option. The data will be copied 4751 * from buffer *from* with length *len* to the TCP header. 4752 * 4753 * The buffer *from* should have the whole option that 4754 * includes the kind, kind-length, and the actual 4755 * option data. The *len* must be at least kind-length 4756 * long. The kind-length does not have to be 4 byte 4757 * aligned. The kernel will take care of the padding 4758 * and setting the 4 bytes aligned value to th->doff. 4759 * 4760 * This helper will check for duplicated option 4761 * by searching the same option in the outgoing skb. 4762 * 4763 * This helper can only be called during 4764 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4765 * 4766 * Return 4767 * 0 on success, or negative error in case of failure: 4768 * 4769 * **-EINVAL** If param is invalid. 4770 * 4771 * **-ENOSPC** if there is not enough space in the header. 4772 * Nothing has been written 4773 * 4774 * **-EEXIST** if the option already exists. 4775 * 4776 * **-EFAULT** on failure to parse the existing header options. 4777 * 4778 * **-EPERM** if the helper cannot be used under the current 4779 * *skops*\ **->op**. 4780 * 4781 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4782 * Description 4783 * Reserve *len* bytes for the bpf header option. The 4784 * space will be used by **bpf_store_hdr_opt**\ () later in 4785 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4786 * 4787 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4788 * the total number of bytes will be reserved. 4789 * 4790 * This helper can only be called during 4791 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4792 * 4793 * Return 4794 * 0 on success, or negative error in case of failure: 4795 * 4796 * **-EINVAL** if a parameter is invalid. 4797 * 4798 * **-ENOSPC** if there is not enough space in the header. 4799 * 4800 * **-EPERM** if the helper cannot be used under the current 4801 * *skops*\ **->op**. 4802 * 4803 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4804 * Description 4805 * Get a bpf_local_storage from an *inode*. 4806 * 4807 * Logically, it could be thought of as getting the value from 4808 * a *map* with *inode* as the **key**. From this 4809 * perspective, the usage is not much different from 4810 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4811 * helper enforces the key must be an inode and the map must also 4812 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4813 * 4814 * Underneath, the value is stored locally at *inode* instead of 4815 * the *map*. The *map* is used as the bpf-local-storage 4816 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4817 * searched against all bpf_local_storage residing at *inode*. 4818 * 4819 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4820 * used such that a new bpf_local_storage will be 4821 * created if one does not exist. *value* can be used 4822 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4823 * the initial value of a bpf_local_storage. If *value* is 4824 * **NULL**, the new bpf_local_storage will be zero initialized. 4825 * Return 4826 * A bpf_local_storage pointer is returned on success. 4827 * 4828 * **NULL** if not found or there was an error in adding 4829 * a new bpf_local_storage. 4830 * 4831 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4832 * Description 4833 * Delete a bpf_local_storage from an *inode*. 4834 * Return 4835 * 0 on success. 4836 * 4837 * **-ENOENT** if the bpf_local_storage cannot be found. 4838 * 4839 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4840 * Description 4841 * Return full path for given **struct path** object, which 4842 * needs to be the kernel BTF *path* object. The path is 4843 * returned in the provided buffer *buf* of size *sz* and 4844 * is zero terminated. 4845 * 4846 * Return 4847 * On success, the strictly positive length of the string, 4848 * including the trailing NUL character. On error, a negative 4849 * value. 4850 * 4851 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4852 * Description 4853 * Read *size* bytes from user space address *user_ptr* and store 4854 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4855 * Return 4856 * 0 on success, or a negative error in case of failure. 4857 * 4858 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4859 * Description 4860 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4861 * using *ptr*->type_id. This value should specify the type 4862 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4863 * can be used to look up vmlinux BTF type ids. Traversing the 4864 * data structure using BTF, the type information and values are 4865 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4866 * the pointer data is carried out to avoid kernel crashes during 4867 * operation. Smaller types can use string space on the stack; 4868 * larger programs can use map data to store the string 4869 * representation. 4870 * 4871 * The string can be subsequently shared with userspace via 4872 * bpf_perf_event_output() or ring buffer interfaces. 4873 * bpf_trace_printk() is to be avoided as it places too small 4874 * a limit on string size to be useful. 4875 * 4876 * *flags* is a combination of 4877 * 4878 * **BTF_F_COMPACT** 4879 * no formatting around type information 4880 * **BTF_F_NONAME** 4881 * no struct/union member names/types 4882 * **BTF_F_PTR_RAW** 4883 * show raw (unobfuscated) pointer values; 4884 * equivalent to printk specifier %px. 4885 * **BTF_F_ZERO** 4886 * show zero-valued struct/union members; they 4887 * are not displayed by default 4888 * 4889 * Return 4890 * The number of bytes that were written (or would have been 4891 * written if output had to be truncated due to string size), 4892 * or a negative error in cases of failure. 4893 * 4894 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4895 * Description 4896 * Use BTF to write to seq_write a string representation of 4897 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4898 * *flags* are identical to those used for bpf_snprintf_btf. 4899 * Return 4900 * 0 on success or a negative error in case of failure. 4901 * 4902 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4903 * Description 4904 * See **bpf_get_cgroup_classid**\ () for the main description. 4905 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4906 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4907 * associated socket instead of the current process. 4908 * Return 4909 * The id is returned or 0 in case the id could not be retrieved. 4910 * 4911 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4912 * Description 4913 * Redirect the packet to another net device of index *ifindex* 4914 * and fill in L2 addresses from neighboring subsystem. This helper 4915 * is somewhat similar to **bpf_redirect**\ (), except that it 4916 * populates L2 addresses as well, meaning, internally, the helper 4917 * relies on the neighbor lookup for the L2 address of the nexthop. 4918 * 4919 * The helper will perform a FIB lookup based on the skb's 4920 * networking header to get the address of the next hop, unless 4921 * this is supplied by the caller in the *params* argument. The 4922 * *plen* argument indicates the len of *params* and should be set 4923 * to 0 if *params* is NULL. 4924 * 4925 * The *flags* argument is reserved and must be 0. The helper is 4926 * currently only supported for tc BPF program types, and enabled 4927 * for IPv4 and IPv6 protocols. 4928 * Return 4929 * The helper returns **TC_ACT_REDIRECT** on success or 4930 * **TC_ACT_SHOT** on error. 4931 * 4932 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4933 * Description 4934 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4935 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4936 * extern variable decorated with '__ksym'. For ksym, there is a 4937 * global var (either static or global) defined of the same name 4938 * in the kernel. The ksym is percpu if the global var is percpu. 4939 * The returned pointer points to the global percpu var on *cpu*. 4940 * 4941 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4942 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4943 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4944 * bpf_per_cpu_ptr() must check the returned value. 4945 * Return 4946 * A pointer pointing to the kernel percpu variable on *cpu*, or 4947 * NULL, if *cpu* is invalid. 4948 * 4949 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4950 * Description 4951 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4952 * pointer to the percpu kernel variable on this cpu. See the 4953 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4954 * 4955 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4956 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4957 * never return NULL. 4958 * Return 4959 * A pointer pointing to the kernel percpu variable on this cpu. 4960 * 4961 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4962 * Description 4963 * Redirect the packet to another net device of index *ifindex*. 4964 * This helper is somewhat similar to **bpf_redirect**\ (), except 4965 * that the redirection happens to the *ifindex*' peer device and 4966 * the netns switch takes place from ingress to ingress without 4967 * going through the CPU's backlog queue. 4968 * 4969 * The *flags* argument is reserved and must be 0. The helper is 4970 * currently only supported for tc BPF program types at the 4971 * ingress hook and for veth and netkit target device types. The 4972 * peer device must reside in a different network namespace. 4973 * Return 4974 * The helper returns **TC_ACT_REDIRECT** on success or 4975 * **TC_ACT_SHOT** on error. 4976 * 4977 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4978 * Description 4979 * Get a bpf_local_storage from the *task*. 4980 * 4981 * Logically, it could be thought of as getting the value from 4982 * a *map* with *task* as the **key**. From this 4983 * perspective, the usage is not much different from 4984 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4985 * helper enforces the key must be a task_struct and the map must also 4986 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4987 * 4988 * Underneath, the value is stored locally at *task* instead of 4989 * the *map*. The *map* is used as the bpf-local-storage 4990 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4991 * searched against all bpf_local_storage residing at *task*. 4992 * 4993 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4994 * used such that a new bpf_local_storage will be 4995 * created if one does not exist. *value* can be used 4996 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4997 * the initial value of a bpf_local_storage. If *value* is 4998 * **NULL**, the new bpf_local_storage will be zero initialized. 4999 * Return 5000 * A bpf_local_storage pointer is returned on success. 5001 * 5002 * **NULL** if not found or there was an error in adding 5003 * a new bpf_local_storage. 5004 * 5005 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 5006 * Description 5007 * Delete a bpf_local_storage from a *task*. 5008 * Return 5009 * 0 on success. 5010 * 5011 * **-ENOENT** if the bpf_local_storage cannot be found. 5012 * 5013 * struct task_struct *bpf_get_current_task_btf(void) 5014 * Description 5015 * Return a BTF pointer to the "current" task. 5016 * This pointer can also be used in helpers that accept an 5017 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 5018 * Return 5019 * Pointer to the current task. 5020 * 5021 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 5022 * Description 5023 * Set or clear certain options on *bprm*: 5024 * 5025 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 5026 * which sets the **AT_SECURE** auxv for glibc. The bit 5027 * is cleared if the flag is not specified. 5028 * Return 5029 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 5030 * 5031 * u64 bpf_ktime_get_coarse_ns(void) 5032 * Description 5033 * Return a coarse-grained version of the time elapsed since 5034 * system boot, in nanoseconds. Does not include time the system 5035 * was suspended. 5036 * 5037 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 5038 * Return 5039 * Current *ktime*. 5040 * 5041 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 5042 * Description 5043 * Returns the stored IMA hash of the *inode* (if it's available). 5044 * If the hash is larger than *size*, then only *size* 5045 * bytes will be copied to *dst* 5046 * Return 5047 * The **hash_algo** is returned on success, 5048 * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if 5049 * invalid arguments are passed. 5050 * 5051 * struct socket *bpf_sock_from_file(struct file *file) 5052 * Description 5053 * If the given file represents a socket, returns the associated 5054 * socket. 5055 * Return 5056 * A pointer to a struct socket on success or NULL if the file is 5057 * not a socket. 5058 * 5059 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 5060 * Description 5061 * Check packet size against exceeding MTU of net device (based 5062 * on *ifindex*). This helper will likely be used in combination 5063 * with helpers that adjust/change the packet size. 5064 * 5065 * The argument *len_diff* can be used for querying with a planned 5066 * size change. This allows to check MTU prior to changing packet 5067 * ctx. Providing a *len_diff* adjustment that is larger than the 5068 * actual packet size (resulting in negative packet size) will in 5069 * principle not exceed the MTU, which is why it is not considered 5070 * a failure. Other BPF helpers are needed for performing the 5071 * planned size change; therefore the responsibility for catching 5072 * a negative packet size belongs in those helpers. 5073 * 5074 * Specifying *ifindex* zero means the MTU check is performed 5075 * against the current net device. This is practical if this isn't 5076 * used prior to redirect. 5077 * 5078 * On input *mtu_len* must be a valid pointer, else verifier will 5079 * reject BPF program. If the value *mtu_len* is initialized to 5080 * zero then the ctx packet size is use. When value *mtu_len* is 5081 * provided as input this specify the L3 length that the MTU check 5082 * is done against. Remember XDP and TC length operate at L2, but 5083 * this value is L3 as this correlate to MTU and IP-header tot_len 5084 * values which are L3 (similar behavior as bpf_fib_lookup). 5085 * 5086 * The Linux kernel route table can configure MTUs on a more 5087 * specific per route level, which is not provided by this helper. 5088 * For route level MTU checks use the **bpf_fib_lookup**\ () 5089 * helper. 5090 * 5091 * *ctx* is either **struct xdp_md** for XDP programs or 5092 * **struct sk_buff** for tc cls_act programs. 5093 * 5094 * The *flags* argument can be a combination of one or more of the 5095 * following values: 5096 * 5097 * **BPF_MTU_CHK_SEGS** 5098 * This flag will only works for *ctx* **struct sk_buff**. 5099 * If packet context contains extra packet segment buffers 5100 * (often knows as GSO skb), then MTU check is harder to 5101 * check at this point, because in transmit path it is 5102 * possible for the skb packet to get re-segmented 5103 * (depending on net device features). This could still be 5104 * a MTU violation, so this flag enables performing MTU 5105 * check against segments, with a different violation 5106 * return code to tell it apart. Check cannot use len_diff. 5107 * 5108 * On return *mtu_len* pointer contains the MTU value of the net 5109 * device. Remember the net device configured MTU is the L3 size, 5110 * which is returned here and XDP and TC length operate at L2. 5111 * Helper take this into account for you, but remember when using 5112 * MTU value in your BPF-code. 5113 * 5114 * Return 5115 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5116 * 5117 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5118 * 5119 * MTU violations return positive values, but also populate MTU 5120 * value in *mtu_len* pointer, as this can be needed for 5121 * implementing PMTU handing: 5122 * 5123 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5124 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5125 * 5126 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5127 * Description 5128 * For each element in **map**, call **callback_fn** function with 5129 * **map**, **callback_ctx** and other map-specific parameters. 5130 * The **callback_fn** should be a static function and 5131 * the **callback_ctx** should be a pointer to the stack. 5132 * The **flags** is used to control certain aspects of the helper. 5133 * Currently, the **flags** must be 0. 5134 * 5135 * The following are a list of supported map types and their 5136 * respective expected callback signatures: 5137 * 5138 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5139 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5140 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5141 * 5142 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5143 * 5144 * For per_cpu maps, the map_value is the value on the cpu where the 5145 * bpf_prog is running. 5146 * 5147 * If **callback_fn** return 0, the helper will continue to the next 5148 * element. If return value is 1, the helper will skip the rest of 5149 * elements and return. Other return values are not used now. 5150 * 5151 * Return 5152 * The number of traversed map elements for success, **-EINVAL** for 5153 * invalid **flags**. 5154 * 5155 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5156 * Description 5157 * Outputs a string into the **str** buffer of size **str_size** 5158 * based on a format string stored in a read-only map pointed by 5159 * **fmt**. 5160 * 5161 * Each format specifier in **fmt** corresponds to one u64 element 5162 * in the **data** array. For strings and pointers where pointees 5163 * are accessed, only the pointer values are stored in the *data* 5164 * array. The *data_len* is the size of *data* in bytes - must be 5165 * a multiple of 8. 5166 * 5167 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5168 * memory. Reading kernel memory may fail due to either invalid 5169 * address or valid address but requiring a major memory fault. If 5170 * reading kernel memory fails, the string for **%s** will be an 5171 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5172 * Not returning error to bpf program is consistent with what 5173 * **bpf_trace_printk**\ () does for now. 5174 * 5175 * Return 5176 * The strictly positive length of the formatted string, including 5177 * the trailing zero character. If the return value is greater than 5178 * **str_size**, **str** contains a truncated string, guaranteed to 5179 * be zero-terminated except when **str_size** is 0. 5180 * 5181 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5182 * 5183 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5184 * Description 5185 * Execute bpf syscall with given arguments. 5186 * Return 5187 * A syscall result. 5188 * 5189 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5190 * Description 5191 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5192 * Return 5193 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5194 * 5195 * long bpf_sys_close(u32 fd) 5196 * Description 5197 * Execute close syscall for given FD. 5198 * Return 5199 * A syscall result. 5200 * 5201 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5202 * Description 5203 * Initialize the timer. 5204 * First 4 bits of *flags* specify clockid. 5205 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5206 * All other bits of *flags* are reserved. 5207 * The verifier will reject the program if *timer* is not from 5208 * the same *map*. 5209 * Return 5210 * 0 on success. 5211 * **-EBUSY** if *timer* is already initialized. 5212 * **-EINVAL** if invalid *flags* are passed. 5213 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5214 * The user space should either hold a file descriptor to a map with timers 5215 * or pin such map in bpffs. When map is unpinned or file descriptor is 5216 * closed all timers in the map will be cancelled and freed. 5217 * 5218 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5219 * Description 5220 * Configure the timer to call *callback_fn* static function. 5221 * Return 5222 * 0 on success. 5223 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5224 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5225 * The user space should either hold a file descriptor to a map with timers 5226 * or pin such map in bpffs. When map is unpinned or file descriptor is 5227 * closed all timers in the map will be cancelled and freed. 5228 * 5229 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5230 * Description 5231 * Set timer expiration N nanoseconds from the current time. The 5232 * configured callback will be invoked in soft irq context on some cpu 5233 * and will not repeat unless another bpf_timer_start() is made. 5234 * In such case the next invocation can migrate to a different cpu. 5235 * Since struct bpf_timer is a field inside map element the map 5236 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5237 * of BPF program to make sure that callback_fn code stays valid. 5238 * When user space reference to a map reaches zero all timers 5239 * in a map are cancelled and corresponding program's refcnts are 5240 * decremented. This is done to make sure that Ctrl-C of a user 5241 * process doesn't leave any timers running. If map is pinned in 5242 * bpffs the callback_fn can re-arm itself indefinitely. 5243 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5244 * cancel and free the timer in the given map element. 5245 * The map can contain timers that invoke callback_fn-s from different 5246 * programs. The same callback_fn can serve different timers from 5247 * different maps if key/value layout matches across maps. 5248 * Every bpf_timer_set_callback() can have different callback_fn. 5249 * 5250 * *flags* can be one of: 5251 * 5252 * **BPF_F_TIMER_ABS** 5253 * Start the timer in absolute expire value instead of the 5254 * default relative one. 5255 * **BPF_F_TIMER_CPU_PIN** 5256 * Timer will be pinned to the CPU of the caller. 5257 * 5258 * Return 5259 * 0 on success. 5260 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5261 * or invalid *flags* are passed. 5262 * 5263 * long bpf_timer_cancel(struct bpf_timer *timer) 5264 * Description 5265 * Cancel the timer and wait for callback_fn to finish if it was running. 5266 * Return 5267 * 0 if the timer was not active. 5268 * 1 if the timer was active. 5269 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5270 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5271 * own timer which would have led to a deadlock otherwise. 5272 * 5273 * u64 bpf_get_func_ip(void *ctx) 5274 * Description 5275 * Get address of the traced function (for tracing and kprobe programs). 5276 * 5277 * When called for kprobe program attached as uprobe it returns 5278 * probe address for both entry and return uprobe. 5279 * 5280 * Return 5281 * Address of the traced function for kprobe. 5282 * 0 for kprobes placed within the function (not at the entry). 5283 * Address of the probe for uprobe and return uprobe. 5284 * 5285 * u64 bpf_get_attach_cookie(void *ctx) 5286 * Description 5287 * Get bpf_cookie value provided (optionally) during the program 5288 * attachment. It might be different for each individual 5289 * attachment, even if BPF program itself is the same. 5290 * Expects BPF program context *ctx* as a first argument. 5291 * 5292 * Supported for the following program types: 5293 * - kprobe/uprobe; 5294 * - tracepoint; 5295 * - perf_event. 5296 * Return 5297 * Value specified by user at BPF link creation/attachment time 5298 * or 0, if it was not specified. 5299 * 5300 * long bpf_task_pt_regs(struct task_struct *task) 5301 * Description 5302 * Get the struct pt_regs associated with **task**. 5303 * Return 5304 * A pointer to struct pt_regs. 5305 * 5306 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5307 * Description 5308 * Get branch trace from hardware engines like Intel LBR. The 5309 * hardware engine is stopped shortly after the helper is 5310 * called. Therefore, the user need to filter branch entries 5311 * based on the actual use case. To capture branch trace 5312 * before the trigger point of the BPF program, the helper 5313 * should be called at the beginning of the BPF program. 5314 * 5315 * The data is stored as struct perf_branch_entry into output 5316 * buffer *entries*. *size* is the size of *entries* in bytes. 5317 * *flags* is reserved for now and must be zero. 5318 * 5319 * Return 5320 * On success, number of bytes written to *buf*. On error, a 5321 * negative value. 5322 * 5323 * **-EINVAL** if *flags* is not zero. 5324 * 5325 * **-ENOENT** if architecture does not support branch records. 5326 * 5327 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5328 * Description 5329 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5330 * to format and can handle more format args as a result. 5331 * 5332 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5333 * Return 5334 * The number of bytes written to the buffer, or a negative error 5335 * in case of failure. 5336 * 5337 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5338 * Description 5339 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5340 * Return 5341 * *sk* if casting is valid, or **NULL** otherwise. 5342 * 5343 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5344 * Description 5345 * Get the address of a kernel symbol, returned in *res*. *res* is 5346 * set to 0 if the symbol is not found. 5347 * Return 5348 * On success, zero. On error, a negative value. 5349 * 5350 * **-EINVAL** if *flags* is not zero. 5351 * 5352 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5353 * 5354 * **-ENOENT** if symbol is not found. 5355 * 5356 * **-EPERM** if caller does not have permission to obtain kernel address. 5357 * 5358 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5359 * Description 5360 * Find vma of *task* that contains *addr*, call *callback_fn* 5361 * function with *task*, *vma*, and *callback_ctx*. 5362 * The *callback_fn* should be a static function and 5363 * the *callback_ctx* should be a pointer to the stack. 5364 * The *flags* is used to control certain aspects of the helper. 5365 * Currently, the *flags* must be 0. 5366 * 5367 * The expected callback signature is 5368 * 5369 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5370 * 5371 * Return 5372 * 0 on success. 5373 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5374 * **-EBUSY** if failed to try lock mmap_lock. 5375 * **-EINVAL** for invalid **flags**. 5376 * 5377 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5378 * Description 5379 * For **nr_loops**, call **callback_fn** function 5380 * with **callback_ctx** as the context parameter. 5381 * The **callback_fn** should be a static function and 5382 * the **callback_ctx** should be a pointer to the stack. 5383 * The **flags** is used to control certain aspects of the helper. 5384 * Currently, the **flags** must be 0. Currently, nr_loops is 5385 * limited to 1 << 23 (~8 million) loops. 5386 * 5387 * long (\*callback_fn)(u32 index, void \*ctx); 5388 * 5389 * where **index** is the current index in the loop. The index 5390 * is zero-indexed. 5391 * 5392 * If **callback_fn** returns 0, the helper will continue to the next 5393 * loop. If return value is 1, the helper will skip the rest of 5394 * the loops and return. Other return values are not used now, 5395 * and will be rejected by the verifier. 5396 * 5397 * Return 5398 * The number of loops performed, **-EINVAL** for invalid **flags**, 5399 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5400 * 5401 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5402 * Description 5403 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5404 * to be null-terminated and **s1_sz** is the maximum storage 5405 * size of **s1**. **s2** must be a read-only string. 5406 * Return 5407 * An integer less than, equal to, or greater than zero 5408 * if the first **s1_sz** bytes of **s1** is found to be 5409 * less than, to match, or be greater than **s2**. 5410 * 5411 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5412 * Description 5413 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5414 * returned in **value**. 5415 * 5416 * Return 5417 * 0 on success. 5418 * **-EINVAL** if n >= argument register count of traced function. 5419 * 5420 * long bpf_get_func_ret(void *ctx, u64 *value) 5421 * Description 5422 * Get return value of the traced function (for tracing programs) 5423 * in **value**. 5424 * 5425 * Return 5426 * 0 on success. 5427 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5428 * 5429 * long bpf_get_func_arg_cnt(void *ctx) 5430 * Description 5431 * Get number of registers of the traced function (for tracing programs) where 5432 * function arguments are stored in these registers. 5433 * 5434 * Return 5435 * The number of argument registers of the traced function. 5436 * 5437 * int bpf_get_retval(void) 5438 * Description 5439 * Get the BPF program's return value that will be returned to the upper layers. 5440 * 5441 * This helper is currently supported by cgroup programs and only by the hooks 5442 * where BPF program's return value is returned to the userspace via errno. 5443 * Return 5444 * The BPF program's return value. 5445 * 5446 * int bpf_set_retval(int retval) 5447 * Description 5448 * Set the BPF program's return value that will be returned to the upper layers. 5449 * 5450 * This helper is currently supported by cgroup programs and only by the hooks 5451 * where BPF program's return value is returned to the userspace via errno. 5452 * 5453 * Note that there is the following corner case where the program exports an error 5454 * via bpf_set_retval but signals success via 'return 1': 5455 * 5456 * bpf_set_retval(-EPERM); 5457 * return 1; 5458 * 5459 * In this case, the BPF program's return value will use helper's -EPERM. This 5460 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5461 * 5462 * Return 5463 * 0 on success, or a negative error in case of failure. 5464 * 5465 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5466 * Description 5467 * Get the total size of a given xdp buff (linear and paged area) 5468 * Return 5469 * The total size of a given xdp buffer. 5470 * 5471 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5472 * Description 5473 * This helper is provided as an easy way to load data from a 5474 * xdp buffer. It can be used to load *len* bytes from *offset* from 5475 * the frame associated to *xdp_md*, into the buffer pointed by 5476 * *buf*. 5477 * Return 5478 * 0 on success, or a negative error in case of failure. 5479 * 5480 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5481 * Description 5482 * Store *len* bytes from buffer *buf* into the frame 5483 * associated to *xdp_md*, at *offset*. 5484 * Return 5485 * 0 on success, or a negative error in case of failure. 5486 * 5487 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5488 * Description 5489 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5490 * address space, and stores the data in *dst*. *flags* is not 5491 * used yet and is provided for future extensibility. This helper 5492 * can only be used by sleepable programs. 5493 * Return 5494 * 0 on success, or a negative error in case of failure. On error 5495 * *dst* buffer is zeroed out. 5496 * 5497 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5498 * Description 5499 * Change the __sk_buff->tstamp_type to *tstamp_type* 5500 * and set *tstamp* to the __sk_buff->tstamp together. 5501 * 5502 * If there is no need to change the __sk_buff->tstamp_type, 5503 * the tstamp value can be directly written to __sk_buff->tstamp 5504 * instead. 5505 * 5506 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5507 * will be kept during bpf_redirect_*(). A non zero 5508 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5509 * *tstamp_type*. 5510 * 5511 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5512 * with a zero *tstamp*. 5513 * 5514 * Only IPv4 and IPv6 skb->protocol are supported. 5515 * 5516 * This function is most useful when it needs to set a 5517 * mono delivery time to __sk_buff->tstamp and then 5518 * bpf_redirect_*() to the egress of an iface. For example, 5519 * changing the (rcv) timestamp in __sk_buff->tstamp at 5520 * ingress to a mono delivery time and then bpf_redirect_*() 5521 * to sch_fq@phy-dev. 5522 * Return 5523 * 0 on success. 5524 * **-EINVAL** for invalid input 5525 * **-EOPNOTSUPP** for unsupported protocol 5526 * 5527 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5528 * Description 5529 * Returns a calculated IMA hash of the *file*. 5530 * If the hash is larger than *size*, then only *size* 5531 * bytes will be copied to *dst* 5532 * Return 5533 * The **hash_algo** is returned on success, 5534 * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if 5535 * invalid arguments are passed. 5536 * 5537 * void *bpf_kptr_xchg(void *dst, void *ptr) 5538 * Description 5539 * Exchange kptr at pointer *dst* with *ptr*, and return the old value. 5540 * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise 5541 * it must be a referenced pointer which will be released when this helper 5542 * is called. 5543 * Return 5544 * The old value of kptr (which can be NULL). The returned pointer 5545 * if not NULL, is a reference which must be released using its 5546 * corresponding release function, or moved into a BPF map before 5547 * program exit. 5548 * 5549 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5550 * Description 5551 * Perform a lookup in *percpu map* for an entry associated to 5552 * *key* on *cpu*. 5553 * Return 5554 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5555 * was found or *cpu* is invalid. 5556 * 5557 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5558 * Description 5559 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5560 * Return 5561 * *sk* if casting is valid, or **NULL** otherwise. 5562 * 5563 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5564 * Description 5565 * Get a dynptr to local memory *data*. 5566 * 5567 * *data* must be a ptr to a map value. 5568 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5569 * *flags* is currently unused. 5570 * Return 5571 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5572 * -EINVAL if flags is not 0. 5573 * 5574 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5575 * Description 5576 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5577 * through the dynptr interface. *flags* must be 0. 5578 * 5579 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5580 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5581 * reservation fails. This is enforced by the verifier. 5582 * Return 5583 * 0 on success, or a negative error in case of failure. 5584 * 5585 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5586 * Description 5587 * Submit reserved ring buffer sample, pointed to by *data*, 5588 * through the dynptr interface. This is a no-op if the dynptr is 5589 * invalid/null. 5590 * 5591 * For more information on *flags*, please see 5592 * 'bpf_ringbuf_submit'. 5593 * Return 5594 * Nothing. Always succeeds. 5595 * 5596 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5597 * Description 5598 * Discard reserved ring buffer sample through the dynptr 5599 * interface. This is a no-op if the dynptr is invalid/null. 5600 * 5601 * For more information on *flags*, please see 5602 * 'bpf_ringbuf_discard'. 5603 * Return 5604 * Nothing. Always succeeds. 5605 * 5606 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5607 * Description 5608 * Read *len* bytes from *src* into *dst*, starting from *offset* 5609 * into *src*. 5610 * *flags* is currently unused. 5611 * Return 5612 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5613 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5614 * *flags* is not 0. 5615 * 5616 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5617 * Description 5618 * Write *len* bytes from *src* into *dst*, starting from *offset* 5619 * into *dst*. 5620 * 5621 * *flags* must be 0 except for skb-type dynptrs. 5622 * 5623 * For skb-type dynptrs: 5624 * * All data slices of the dynptr are automatically 5625 * invalidated after **bpf_dynptr_write**\ (). This is 5626 * because writing may pull the skb and change the 5627 * underlying packet buffer. 5628 * 5629 * * For *flags*, please see the flags accepted by 5630 * **bpf_skb_store_bytes**\ (). 5631 * Return 5632 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5633 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5634 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5635 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5636 * 5637 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5638 * Description 5639 * Get a pointer to the underlying dynptr data. 5640 * 5641 * *len* must be a statically known value. The returned data slice 5642 * is invalidated whenever the dynptr is invalidated. 5643 * 5644 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5645 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5646 * Return 5647 * Pointer to the underlying dynptr data, NULL if the dynptr is 5648 * read-only, if the dynptr is invalid, or if the offset and length 5649 * is out of bounds. 5650 * 5651 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5652 * Description 5653 * Try to issue a SYN cookie for the packet with corresponding 5654 * IPv4/TCP headers, *iph* and *th*, without depending on a 5655 * listening socket. 5656 * 5657 * *iph* points to the IPv4 header. 5658 * 5659 * *th* points to the start of the TCP header, while *th_len* 5660 * contains the length of the TCP header (at least 5661 * **sizeof**\ (**struct tcphdr**)). 5662 * Return 5663 * On success, lower 32 bits hold the generated SYN cookie in 5664 * followed by 16 bits which hold the MSS value for that cookie, 5665 * and the top 16 bits are unused. 5666 * 5667 * On failure, the returned value is one of the following: 5668 * 5669 * **-EINVAL** if *th_len* is invalid. 5670 * 5671 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5672 * Description 5673 * Try to issue a SYN cookie for the packet with corresponding 5674 * IPv6/TCP headers, *iph* and *th*, without depending on a 5675 * listening socket. 5676 * 5677 * *iph* points to the IPv6 header. 5678 * 5679 * *th* points to the start of the TCP header, while *th_len* 5680 * contains the length of the TCP header (at least 5681 * **sizeof**\ (**struct tcphdr**)). 5682 * Return 5683 * On success, lower 32 bits hold the generated SYN cookie in 5684 * followed by 16 bits which hold the MSS value for that cookie, 5685 * and the top 16 bits are unused. 5686 * 5687 * On failure, the returned value is one of the following: 5688 * 5689 * **-EINVAL** if *th_len* is invalid. 5690 * 5691 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5692 * 5693 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5694 * Description 5695 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5696 * without depending on a listening socket. 5697 * 5698 * *iph* points to the IPv4 header. 5699 * 5700 * *th* points to the TCP header. 5701 * Return 5702 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5703 * 5704 * On failure, the returned value is one of the following: 5705 * 5706 * **-EACCES** if the SYN cookie is not valid. 5707 * 5708 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5709 * Description 5710 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5711 * without depending on a listening socket. 5712 * 5713 * *iph* points to the IPv6 header. 5714 * 5715 * *th* points to the TCP header. 5716 * Return 5717 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5718 * 5719 * On failure, the returned value is one of the following: 5720 * 5721 * **-EACCES** if the SYN cookie is not valid. 5722 * 5723 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5724 * 5725 * u64 bpf_ktime_get_tai_ns(void) 5726 * Description 5727 * A nonsettable system-wide clock derived from wall-clock time but 5728 * ignoring leap seconds. This clock does not experience 5729 * discontinuities and backwards jumps caused by NTP inserting leap 5730 * seconds as CLOCK_REALTIME does. 5731 * 5732 * See: **clock_gettime**\ (**CLOCK_TAI**) 5733 * Return 5734 * Current *ktime*. 5735 * 5736 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5737 * Description 5738 * Drain samples from the specified user ring buffer, and invoke 5739 * the provided callback for each such sample: 5740 * 5741 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5742 * 5743 * If **callback_fn** returns 0, the helper will continue to try 5744 * and drain the next sample, up to a maximum of 5745 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5746 * the helper will skip the rest of the samples and return. Other 5747 * return values are not used now, and will be rejected by the 5748 * verifier. 5749 * Return 5750 * The number of drained samples if no error was encountered while 5751 * draining samples, or 0 if no samples were present in the ring 5752 * buffer. If a user-space producer was epoll-waiting on this map, 5753 * and at least one sample was drained, they will receive an event 5754 * notification notifying them of available space in the ring 5755 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5756 * function, no wakeup notification will be sent. If the 5757 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5758 * be sent even if no sample was drained. 5759 * 5760 * On failure, the returned value is one of the following: 5761 * 5762 * **-EBUSY** if the ring buffer is contended, and another calling 5763 * context was concurrently draining the ring buffer. 5764 * 5765 * **-EINVAL** if user-space is not properly tracking the ring 5766 * buffer due to the producer position not being aligned to 8 5767 * bytes, a sample not being aligned to 8 bytes, or the producer 5768 * position not matching the advertised length of a sample. 5769 * 5770 * **-E2BIG** if user-space has tried to publish a sample which is 5771 * larger than the size of the ring buffer, or which cannot fit 5772 * within a struct bpf_dynptr. 5773 * 5774 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5775 * Description 5776 * Get a bpf_local_storage from the *cgroup*. 5777 * 5778 * Logically, it could be thought of as getting the value from 5779 * a *map* with *cgroup* as the **key**. From this 5780 * perspective, the usage is not much different from 5781 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5782 * helper enforces the key must be a cgroup struct and the map must also 5783 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5784 * 5785 * In reality, the local-storage value is embedded directly inside of the 5786 * *cgroup* object itself, rather than being located in the 5787 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5788 * queried for some *map* on a *cgroup* object, the kernel will perform an 5789 * O(n) iteration over all of the live local-storage values for that 5790 * *cgroup* object until the local-storage value for the *map* is found. 5791 * 5792 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5793 * used such that a new bpf_local_storage will be 5794 * created if one does not exist. *value* can be used 5795 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5796 * the initial value of a bpf_local_storage. If *value* is 5797 * **NULL**, the new bpf_local_storage will be zero initialized. 5798 * Return 5799 * A bpf_local_storage pointer is returned on success. 5800 * 5801 * **NULL** if not found or there was an error in adding 5802 * a new bpf_local_storage. 5803 * 5804 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5805 * Description 5806 * Delete a bpf_local_storage from a *cgroup*. 5807 * Return 5808 * 0 on success. 5809 * 5810 * **-ENOENT** if the bpf_local_storage cannot be found. 5811 */ 5812 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5813 FN(unspec, 0, ##ctx) \ 5814 FN(map_lookup_elem, 1, ##ctx) \ 5815 FN(map_update_elem, 2, ##ctx) \ 5816 FN(map_delete_elem, 3, ##ctx) \ 5817 FN(probe_read, 4, ##ctx) \ 5818 FN(ktime_get_ns, 5, ##ctx) \ 5819 FN(trace_printk, 6, ##ctx) \ 5820 FN(get_prandom_u32, 7, ##ctx) \ 5821 FN(get_smp_processor_id, 8, ##ctx) \ 5822 FN(skb_store_bytes, 9, ##ctx) \ 5823 FN(l3_csum_replace, 10, ##ctx) \ 5824 FN(l4_csum_replace, 11, ##ctx) \ 5825 FN(tail_call, 12, ##ctx) \ 5826 FN(clone_redirect, 13, ##ctx) \ 5827 FN(get_current_pid_tgid, 14, ##ctx) \ 5828 FN(get_current_uid_gid, 15, ##ctx) \ 5829 FN(get_current_comm, 16, ##ctx) \ 5830 FN(get_cgroup_classid, 17, ##ctx) \ 5831 FN(skb_vlan_push, 18, ##ctx) \ 5832 FN(skb_vlan_pop, 19, ##ctx) \ 5833 FN(skb_get_tunnel_key, 20, ##ctx) \ 5834 FN(skb_set_tunnel_key, 21, ##ctx) \ 5835 FN(perf_event_read, 22, ##ctx) \ 5836 FN(redirect, 23, ##ctx) \ 5837 FN(get_route_realm, 24, ##ctx) \ 5838 FN(perf_event_output, 25, ##ctx) \ 5839 FN(skb_load_bytes, 26, ##ctx) \ 5840 FN(get_stackid, 27, ##ctx) \ 5841 FN(csum_diff, 28, ##ctx) \ 5842 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5843 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5844 FN(skb_change_proto, 31, ##ctx) \ 5845 FN(skb_change_type, 32, ##ctx) \ 5846 FN(skb_under_cgroup, 33, ##ctx) \ 5847 FN(get_hash_recalc, 34, ##ctx) \ 5848 FN(get_current_task, 35, ##ctx) \ 5849 FN(probe_write_user, 36, ##ctx) \ 5850 FN(current_task_under_cgroup, 37, ##ctx) \ 5851 FN(skb_change_tail, 38, ##ctx) \ 5852 FN(skb_pull_data, 39, ##ctx) \ 5853 FN(csum_update, 40, ##ctx) \ 5854 FN(set_hash_invalid, 41, ##ctx) \ 5855 FN(get_numa_node_id, 42, ##ctx) \ 5856 FN(skb_change_head, 43, ##ctx) \ 5857 FN(xdp_adjust_head, 44, ##ctx) \ 5858 FN(probe_read_str, 45, ##ctx) \ 5859 FN(get_socket_cookie, 46, ##ctx) \ 5860 FN(get_socket_uid, 47, ##ctx) \ 5861 FN(set_hash, 48, ##ctx) \ 5862 FN(setsockopt, 49, ##ctx) \ 5863 FN(skb_adjust_room, 50, ##ctx) \ 5864 FN(redirect_map, 51, ##ctx) \ 5865 FN(sk_redirect_map, 52, ##ctx) \ 5866 FN(sock_map_update, 53, ##ctx) \ 5867 FN(xdp_adjust_meta, 54, ##ctx) \ 5868 FN(perf_event_read_value, 55, ##ctx) \ 5869 FN(perf_prog_read_value, 56, ##ctx) \ 5870 FN(getsockopt, 57, ##ctx) \ 5871 FN(override_return, 58, ##ctx) \ 5872 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5873 FN(msg_redirect_map, 60, ##ctx) \ 5874 FN(msg_apply_bytes, 61, ##ctx) \ 5875 FN(msg_cork_bytes, 62, ##ctx) \ 5876 FN(msg_pull_data, 63, ##ctx) \ 5877 FN(bind, 64, ##ctx) \ 5878 FN(xdp_adjust_tail, 65, ##ctx) \ 5879 FN(skb_get_xfrm_state, 66, ##ctx) \ 5880 FN(get_stack, 67, ##ctx) \ 5881 FN(skb_load_bytes_relative, 68, ##ctx) \ 5882 FN(fib_lookup, 69, ##ctx) \ 5883 FN(sock_hash_update, 70, ##ctx) \ 5884 FN(msg_redirect_hash, 71, ##ctx) \ 5885 FN(sk_redirect_hash, 72, ##ctx) \ 5886 FN(lwt_push_encap, 73, ##ctx) \ 5887 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5888 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5889 FN(lwt_seg6_action, 76, ##ctx) \ 5890 FN(rc_repeat, 77, ##ctx) \ 5891 FN(rc_keydown, 78, ##ctx) \ 5892 FN(skb_cgroup_id, 79, ##ctx) \ 5893 FN(get_current_cgroup_id, 80, ##ctx) \ 5894 FN(get_local_storage, 81, ##ctx) \ 5895 FN(sk_select_reuseport, 82, ##ctx) \ 5896 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5897 FN(sk_lookup_tcp, 84, ##ctx) \ 5898 FN(sk_lookup_udp, 85, ##ctx) \ 5899 FN(sk_release, 86, ##ctx) \ 5900 FN(map_push_elem, 87, ##ctx) \ 5901 FN(map_pop_elem, 88, ##ctx) \ 5902 FN(map_peek_elem, 89, ##ctx) \ 5903 FN(msg_push_data, 90, ##ctx) \ 5904 FN(msg_pop_data, 91, ##ctx) \ 5905 FN(rc_pointer_rel, 92, ##ctx) \ 5906 FN(spin_lock, 93, ##ctx) \ 5907 FN(spin_unlock, 94, ##ctx) \ 5908 FN(sk_fullsock, 95, ##ctx) \ 5909 FN(tcp_sock, 96, ##ctx) \ 5910 FN(skb_ecn_set_ce, 97, ##ctx) \ 5911 FN(get_listener_sock, 98, ##ctx) \ 5912 FN(skc_lookup_tcp, 99, ##ctx) \ 5913 FN(tcp_check_syncookie, 100, ##ctx) \ 5914 FN(sysctl_get_name, 101, ##ctx) \ 5915 FN(sysctl_get_current_value, 102, ##ctx) \ 5916 FN(sysctl_get_new_value, 103, ##ctx) \ 5917 FN(sysctl_set_new_value, 104, ##ctx) \ 5918 FN(strtol, 105, ##ctx) \ 5919 FN(strtoul, 106, ##ctx) \ 5920 FN(sk_storage_get, 107, ##ctx) \ 5921 FN(sk_storage_delete, 108, ##ctx) \ 5922 FN(send_signal, 109, ##ctx) \ 5923 FN(tcp_gen_syncookie, 110, ##ctx) \ 5924 FN(skb_output, 111, ##ctx) \ 5925 FN(probe_read_user, 112, ##ctx) \ 5926 FN(probe_read_kernel, 113, ##ctx) \ 5927 FN(probe_read_user_str, 114, ##ctx) \ 5928 FN(probe_read_kernel_str, 115, ##ctx) \ 5929 FN(tcp_send_ack, 116, ##ctx) \ 5930 FN(send_signal_thread, 117, ##ctx) \ 5931 FN(jiffies64, 118, ##ctx) \ 5932 FN(read_branch_records, 119, ##ctx) \ 5933 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5934 FN(xdp_output, 121, ##ctx) \ 5935 FN(get_netns_cookie, 122, ##ctx) \ 5936 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5937 FN(sk_assign, 124, ##ctx) \ 5938 FN(ktime_get_boot_ns, 125, ##ctx) \ 5939 FN(seq_printf, 126, ##ctx) \ 5940 FN(seq_write, 127, ##ctx) \ 5941 FN(sk_cgroup_id, 128, ##ctx) \ 5942 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5943 FN(ringbuf_output, 130, ##ctx) \ 5944 FN(ringbuf_reserve, 131, ##ctx) \ 5945 FN(ringbuf_submit, 132, ##ctx) \ 5946 FN(ringbuf_discard, 133, ##ctx) \ 5947 FN(ringbuf_query, 134, ##ctx) \ 5948 FN(csum_level, 135, ##ctx) \ 5949 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5950 FN(skc_to_tcp_sock, 137, ##ctx) \ 5951 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5952 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5953 FN(skc_to_udp6_sock, 140, ##ctx) \ 5954 FN(get_task_stack, 141, ##ctx) \ 5955 FN(load_hdr_opt, 142, ##ctx) \ 5956 FN(store_hdr_opt, 143, ##ctx) \ 5957 FN(reserve_hdr_opt, 144, ##ctx) \ 5958 FN(inode_storage_get, 145, ##ctx) \ 5959 FN(inode_storage_delete, 146, ##ctx) \ 5960 FN(d_path, 147, ##ctx) \ 5961 FN(copy_from_user, 148, ##ctx) \ 5962 FN(snprintf_btf, 149, ##ctx) \ 5963 FN(seq_printf_btf, 150, ##ctx) \ 5964 FN(skb_cgroup_classid, 151, ##ctx) \ 5965 FN(redirect_neigh, 152, ##ctx) \ 5966 FN(per_cpu_ptr, 153, ##ctx) \ 5967 FN(this_cpu_ptr, 154, ##ctx) \ 5968 FN(redirect_peer, 155, ##ctx) \ 5969 FN(task_storage_get, 156, ##ctx) \ 5970 FN(task_storage_delete, 157, ##ctx) \ 5971 FN(get_current_task_btf, 158, ##ctx) \ 5972 FN(bprm_opts_set, 159, ##ctx) \ 5973 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5974 FN(ima_inode_hash, 161, ##ctx) \ 5975 FN(sock_from_file, 162, ##ctx) \ 5976 FN(check_mtu, 163, ##ctx) \ 5977 FN(for_each_map_elem, 164, ##ctx) \ 5978 FN(snprintf, 165, ##ctx) \ 5979 FN(sys_bpf, 166, ##ctx) \ 5980 FN(btf_find_by_name_kind, 167, ##ctx) \ 5981 FN(sys_close, 168, ##ctx) \ 5982 FN(timer_init, 169, ##ctx) \ 5983 FN(timer_set_callback, 170, ##ctx) \ 5984 FN(timer_start, 171, ##ctx) \ 5985 FN(timer_cancel, 172, ##ctx) \ 5986 FN(get_func_ip, 173, ##ctx) \ 5987 FN(get_attach_cookie, 174, ##ctx) \ 5988 FN(task_pt_regs, 175, ##ctx) \ 5989 FN(get_branch_snapshot, 176, ##ctx) \ 5990 FN(trace_vprintk, 177, ##ctx) \ 5991 FN(skc_to_unix_sock, 178, ##ctx) \ 5992 FN(kallsyms_lookup_name, 179, ##ctx) \ 5993 FN(find_vma, 180, ##ctx) \ 5994 FN(loop, 181, ##ctx) \ 5995 FN(strncmp, 182, ##ctx) \ 5996 FN(get_func_arg, 183, ##ctx) \ 5997 FN(get_func_ret, 184, ##ctx) \ 5998 FN(get_func_arg_cnt, 185, ##ctx) \ 5999 FN(get_retval, 186, ##ctx) \ 6000 FN(set_retval, 187, ##ctx) \ 6001 FN(xdp_get_buff_len, 188, ##ctx) \ 6002 FN(xdp_load_bytes, 189, ##ctx) \ 6003 FN(xdp_store_bytes, 190, ##ctx) \ 6004 FN(copy_from_user_task, 191, ##ctx) \ 6005 FN(skb_set_tstamp, 192, ##ctx) \ 6006 FN(ima_file_hash, 193, ##ctx) \ 6007 FN(kptr_xchg, 194, ##ctx) \ 6008 FN(map_lookup_percpu_elem, 195, ##ctx) \ 6009 FN(skc_to_mptcp_sock, 196, ##ctx) \ 6010 FN(dynptr_from_mem, 197, ##ctx) \ 6011 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 6012 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 6013 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 6014 FN(dynptr_read, 201, ##ctx) \ 6015 FN(dynptr_write, 202, ##ctx) \ 6016 FN(dynptr_data, 203, ##ctx) \ 6017 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 6018 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 6019 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 6020 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 6021 FN(ktime_get_tai_ns, 208, ##ctx) \ 6022 FN(user_ringbuf_drain, 209, ##ctx) \ 6023 FN(cgrp_storage_get, 210, ##ctx) \ 6024 FN(cgrp_storage_delete, 211, ##ctx) \ 6025 /* */ 6026 6027 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 6028 * know or care about integer value that is now passed as second argument 6029 */ 6030 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 6031 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 6032 6033 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 6034 * function eBPF program intends to call 6035 */ 6036 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 6037 enum bpf_func_id { 6038 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 6039 __BPF_FUNC_MAX_ID, 6040 }; 6041 #undef __BPF_ENUM_FN 6042 6043 /* All flags used by eBPF helper functions, placed here. */ 6044 6045 /* BPF_FUNC_skb_store_bytes flags. */ 6046 enum { 6047 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 6048 BPF_F_INVALIDATE_HASH = (1ULL << 1), 6049 }; 6050 6051 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 6052 * First 4 bits are for passing the header field size. 6053 */ 6054 enum { 6055 BPF_F_HDR_FIELD_MASK = 0xfULL, 6056 }; 6057 6058 /* BPF_FUNC_l4_csum_replace flags. */ 6059 enum { 6060 BPF_F_PSEUDO_HDR = (1ULL << 4), 6061 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 6062 BPF_F_MARK_ENFORCE = (1ULL << 6), 6063 BPF_F_IPV6 = (1ULL << 7), 6064 }; 6065 6066 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6067 enum { 6068 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6069 }; 6070 6071 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6072 enum { 6073 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6074 BPF_F_USER_STACK = (1ULL << 8), 6075 /* flags used by BPF_FUNC_get_stackid only. */ 6076 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6077 BPF_F_REUSE_STACKID = (1ULL << 10), 6078 /* flags used by BPF_FUNC_get_stack only. */ 6079 BPF_F_USER_BUILD_ID = (1ULL << 11), 6080 }; 6081 6082 /* BPF_FUNC_skb_set_tunnel_key flags. */ 6083 enum { 6084 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6085 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6086 BPF_F_SEQ_NUMBER = (1ULL << 3), 6087 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6088 }; 6089 6090 /* BPF_FUNC_skb_get_tunnel_key flags. */ 6091 enum { 6092 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6093 }; 6094 6095 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6096 * BPF_FUNC_perf_event_read_value flags. 6097 */ 6098 enum { 6099 BPF_F_INDEX_MASK = 0xffffffffULL, 6100 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6101 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 6102 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6103 }; 6104 6105 /* Current network namespace */ 6106 enum { 6107 BPF_F_CURRENT_NETNS = (-1L), 6108 }; 6109 6110 /* BPF_FUNC_csum_level level values. */ 6111 enum { 6112 BPF_CSUM_LEVEL_QUERY, 6113 BPF_CSUM_LEVEL_INC, 6114 BPF_CSUM_LEVEL_DEC, 6115 BPF_CSUM_LEVEL_RESET, 6116 }; 6117 6118 /* BPF_FUNC_skb_adjust_room flags. */ 6119 enum { 6120 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6121 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6122 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6123 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6124 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6125 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6126 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6127 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6128 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6129 }; 6130 6131 enum { 6132 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6133 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6134 }; 6135 6136 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6137 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6138 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6139 6140 /* BPF_FUNC_sysctl_get_name flags. */ 6141 enum { 6142 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6143 }; 6144 6145 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 6146 enum { 6147 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6148 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6149 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6150 */ 6151 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6152 }; 6153 6154 /* BPF_FUNC_read_branch_records flags. */ 6155 enum { 6156 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6157 }; 6158 6159 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6160 * BPF_FUNC_bpf_ringbuf_output flags. 6161 */ 6162 enum { 6163 BPF_RB_NO_WAKEUP = (1ULL << 0), 6164 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6165 }; 6166 6167 /* BPF_FUNC_bpf_ringbuf_query flags */ 6168 enum { 6169 BPF_RB_AVAIL_DATA = 0, 6170 BPF_RB_RING_SIZE = 1, 6171 BPF_RB_CONS_POS = 2, 6172 BPF_RB_PROD_POS = 3, 6173 }; 6174 6175 /* BPF ring buffer constants */ 6176 enum { 6177 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6178 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6179 BPF_RINGBUF_HDR_SZ = 8, 6180 }; 6181 6182 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6183 enum { 6184 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6185 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6186 }; 6187 6188 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6189 enum bpf_adj_room_mode { 6190 BPF_ADJ_ROOM_NET, 6191 BPF_ADJ_ROOM_MAC, 6192 }; 6193 6194 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6195 enum bpf_hdr_start_off { 6196 BPF_HDR_START_MAC, 6197 BPF_HDR_START_NET, 6198 }; 6199 6200 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6201 enum bpf_lwt_encap_mode { 6202 BPF_LWT_ENCAP_SEG6, 6203 BPF_LWT_ENCAP_SEG6_INLINE, 6204 BPF_LWT_ENCAP_IP, 6205 }; 6206 6207 /* Flags for bpf_bprm_opts_set helper */ 6208 enum { 6209 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6210 }; 6211 6212 /* Flags for bpf_redirect and bpf_redirect_map helpers */ 6213 enum { 6214 BPF_F_INGRESS = (1ULL << 0), /* used for skb path */ 6215 BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */ 6216 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */ 6217 #define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS) 6218 }; 6219 6220 #define __bpf_md_ptr(type, name) \ 6221 union { \ 6222 type name; \ 6223 __u64 :64; \ 6224 } __attribute__((aligned(8))) 6225 6226 /* The enum used in skb->tstamp_type. It specifies the clock type 6227 * of the time stored in the skb->tstamp. 6228 */ 6229 enum { 6230 BPF_SKB_TSTAMP_UNSPEC = 0, /* DEPRECATED */ 6231 BPF_SKB_TSTAMP_DELIVERY_MONO = 1, /* DEPRECATED */ 6232 BPF_SKB_CLOCK_REALTIME = 0, 6233 BPF_SKB_CLOCK_MONOTONIC = 1, 6234 BPF_SKB_CLOCK_TAI = 2, 6235 /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle, 6236 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid. 6237 */ 6238 }; 6239 6240 /* user accessible mirror of in-kernel sk_buff. 6241 * new fields can only be added to the end of this structure 6242 */ 6243 struct __sk_buff { 6244 __u32 len; 6245 __u32 pkt_type; 6246 __u32 mark; 6247 __u32 queue_mapping; 6248 __u32 protocol; 6249 __u32 vlan_present; 6250 __u32 vlan_tci; 6251 __u32 vlan_proto; 6252 __u32 priority; 6253 __u32 ingress_ifindex; 6254 __u32 ifindex; 6255 __u32 tc_index; 6256 __u32 cb[5]; 6257 __u32 hash; 6258 __u32 tc_classid; 6259 __u32 data; 6260 __u32 data_end; 6261 __u32 napi_id; 6262 6263 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6264 __u32 family; 6265 __u32 remote_ip4; /* Stored in network byte order */ 6266 __u32 local_ip4; /* Stored in network byte order */ 6267 __u32 remote_ip6[4]; /* Stored in network byte order */ 6268 __u32 local_ip6[4]; /* Stored in network byte order */ 6269 __u32 remote_port; /* Stored in network byte order */ 6270 __u32 local_port; /* stored in host byte order */ 6271 /* ... here. */ 6272 6273 __u32 data_meta; 6274 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6275 __u64 tstamp; 6276 __u32 wire_len; 6277 __u32 gso_segs; 6278 __bpf_md_ptr(struct bpf_sock *, sk); 6279 __u32 gso_size; 6280 __u8 tstamp_type; 6281 __u32 :24; /* Padding, future use. */ 6282 __u64 hwtstamp; 6283 }; 6284 6285 struct bpf_tunnel_key { 6286 __u32 tunnel_id; 6287 union { 6288 __u32 remote_ipv4; 6289 __u32 remote_ipv6[4]; 6290 }; 6291 __u8 tunnel_tos; 6292 __u8 tunnel_ttl; 6293 union { 6294 __u16 tunnel_ext; /* compat */ 6295 __be16 tunnel_flags; 6296 }; 6297 __u32 tunnel_label; 6298 union { 6299 __u32 local_ipv4; 6300 __u32 local_ipv6[4]; 6301 }; 6302 }; 6303 6304 /* user accessible mirror of in-kernel xfrm_state. 6305 * new fields can only be added to the end of this structure 6306 */ 6307 struct bpf_xfrm_state { 6308 __u32 reqid; 6309 __u32 spi; /* Stored in network byte order */ 6310 __u16 family; 6311 __u16 ext; /* Padding, future use. */ 6312 union { 6313 __u32 remote_ipv4; /* Stored in network byte order */ 6314 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6315 }; 6316 }; 6317 6318 /* Generic BPF return codes which all BPF program types may support. 6319 * The values are binary compatible with their TC_ACT_* counter-part to 6320 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6321 * programs. 6322 * 6323 * XDP is handled seprately, see XDP_*. 6324 */ 6325 enum bpf_ret_code { 6326 BPF_OK = 0, 6327 /* 1 reserved */ 6328 BPF_DROP = 2, 6329 /* 3-6 reserved */ 6330 BPF_REDIRECT = 7, 6331 /* >127 are reserved for prog type specific return codes. 6332 * 6333 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6334 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6335 * changed and should be routed based on its new L3 header. 6336 * (This is an L3 redirect, as opposed to L2 redirect 6337 * represented by BPF_REDIRECT above). 6338 */ 6339 BPF_LWT_REROUTE = 128, 6340 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6341 * to indicate that no custom dissection was performed, and 6342 * fallback to standard dissector is requested. 6343 */ 6344 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6345 }; 6346 6347 struct bpf_sock { 6348 __u32 bound_dev_if; 6349 __u32 family; 6350 __u32 type; 6351 __u32 protocol; 6352 __u32 mark; 6353 __u32 priority; 6354 /* IP address also allows 1 and 2 bytes access */ 6355 __u32 src_ip4; 6356 __u32 src_ip6[4]; 6357 __u32 src_port; /* host byte order */ 6358 __be16 dst_port; /* network byte order */ 6359 __u16 :16; /* zero padding */ 6360 __u32 dst_ip4; 6361 __u32 dst_ip6[4]; 6362 __u32 state; 6363 __s32 rx_queue_mapping; 6364 }; 6365 6366 struct bpf_tcp_sock { 6367 __u32 snd_cwnd; /* Sending congestion window */ 6368 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6369 __u32 rtt_min; 6370 __u32 snd_ssthresh; /* Slow start size threshold */ 6371 __u32 rcv_nxt; /* What we want to receive next */ 6372 __u32 snd_nxt; /* Next sequence we send */ 6373 __u32 snd_una; /* First byte we want an ack for */ 6374 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6375 __u32 ecn_flags; /* ECN status bits. */ 6376 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6377 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6378 __u32 packets_out; /* Packets which are "in flight" */ 6379 __u32 retrans_out; /* Retransmitted packets out */ 6380 __u32 total_retrans; /* Total retransmits for entire connection */ 6381 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6382 * total number of segments in. 6383 */ 6384 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6385 * total number of data segments in. 6386 */ 6387 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6388 * The total number of segments sent. 6389 */ 6390 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6391 * total number of data segments sent. 6392 */ 6393 __u32 lost_out; /* Lost packets */ 6394 __u32 sacked_out; /* SACK'd packets */ 6395 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6396 * sum(delta(rcv_nxt)), or how many bytes 6397 * were acked. 6398 */ 6399 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6400 * sum(delta(snd_una)), or how many bytes 6401 * were acked. 6402 */ 6403 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6404 * total number of DSACK blocks received 6405 */ 6406 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6407 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6408 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6409 }; 6410 6411 struct bpf_sock_tuple { 6412 union { 6413 struct { 6414 __be32 saddr; 6415 __be32 daddr; 6416 __be16 sport; 6417 __be16 dport; 6418 } ipv4; 6419 struct { 6420 __be32 saddr[4]; 6421 __be32 daddr[4]; 6422 __be16 sport; 6423 __be16 dport; 6424 } ipv6; 6425 }; 6426 }; 6427 6428 /* (Simplified) user return codes for tcx prog type. 6429 * A valid tcx program must return one of these defined values. All other 6430 * return codes are reserved for future use. Must remain compatible with 6431 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6432 * return codes are mapped to TCX_NEXT. 6433 */ 6434 enum tcx_action_base { 6435 TCX_NEXT = -1, 6436 TCX_PASS = 0, 6437 TCX_DROP = 2, 6438 TCX_REDIRECT = 7, 6439 }; 6440 6441 struct bpf_xdp_sock { 6442 __u32 queue_id; 6443 }; 6444 6445 #define XDP_PACKET_HEADROOM 256 6446 6447 /* User return codes for XDP prog type. 6448 * A valid XDP program must return one of these defined values. All other 6449 * return codes are reserved for future use. Unknown return codes will 6450 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6451 */ 6452 enum xdp_action { 6453 XDP_ABORTED = 0, 6454 XDP_DROP, 6455 XDP_PASS, 6456 XDP_TX, 6457 XDP_REDIRECT, 6458 }; 6459 6460 /* user accessible metadata for XDP packet hook 6461 * new fields must be added to the end of this structure 6462 */ 6463 struct xdp_md { 6464 __u32 data; 6465 __u32 data_end; 6466 __u32 data_meta; 6467 /* Below access go through struct xdp_rxq_info */ 6468 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6469 __u32 rx_queue_index; /* rxq->queue_index */ 6470 6471 __u32 egress_ifindex; /* txq->dev->ifindex */ 6472 }; 6473 6474 /* DEVMAP map-value layout 6475 * 6476 * The struct data-layout of map-value is a configuration interface. 6477 * New members can only be added to the end of this structure. 6478 */ 6479 struct bpf_devmap_val { 6480 __u32 ifindex; /* device index */ 6481 union { 6482 int fd; /* prog fd on map write */ 6483 __u32 id; /* prog id on map read */ 6484 } bpf_prog; 6485 }; 6486 6487 /* CPUMAP map-value layout 6488 * 6489 * The struct data-layout of map-value is a configuration interface. 6490 * New members can only be added to the end of this structure. 6491 */ 6492 struct bpf_cpumap_val { 6493 __u32 qsize; /* queue size to remote target CPU */ 6494 union { 6495 int fd; /* prog fd on map write */ 6496 __u32 id; /* prog id on map read */ 6497 } bpf_prog; 6498 }; 6499 6500 enum sk_action { 6501 SK_DROP = 0, 6502 SK_PASS, 6503 }; 6504 6505 /* user accessible metadata for SK_MSG packet hook, new fields must 6506 * be added to the end of this structure 6507 */ 6508 struct sk_msg_md { 6509 __bpf_md_ptr(void *, data); 6510 __bpf_md_ptr(void *, data_end); 6511 6512 __u32 family; 6513 __u32 remote_ip4; /* Stored in network byte order */ 6514 __u32 local_ip4; /* Stored in network byte order */ 6515 __u32 remote_ip6[4]; /* Stored in network byte order */ 6516 __u32 local_ip6[4]; /* Stored in network byte order */ 6517 __u32 remote_port; /* Stored in network byte order */ 6518 __u32 local_port; /* stored in host byte order */ 6519 __u32 size; /* Total size of sk_msg */ 6520 6521 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6522 }; 6523 6524 struct sk_reuseport_md { 6525 /* 6526 * Start of directly accessible data. It begins from 6527 * the tcp/udp header. 6528 */ 6529 __bpf_md_ptr(void *, data); 6530 /* End of directly accessible data */ 6531 __bpf_md_ptr(void *, data_end); 6532 /* 6533 * Total length of packet (starting from the tcp/udp header). 6534 * Note that the directly accessible bytes (data_end - data) 6535 * could be less than this "len". Those bytes could be 6536 * indirectly read by a helper "bpf_skb_load_bytes()". 6537 */ 6538 __u32 len; 6539 /* 6540 * Eth protocol in the mac header (network byte order). e.g. 6541 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6542 */ 6543 __u32 eth_protocol; 6544 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6545 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6546 __u32 hash; /* A hash of the packet 4 tuples */ 6547 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6548 * new incoming connection request (e.g. selecting a listen sk for 6549 * the received SYN in the TCP case). reuse->sk is one of the sk 6550 * in the reuseport group. The bpf prog can use reuse->sk to learn 6551 * the local listening ip/port without looking into the skb. 6552 * 6553 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6554 * reuse->migrating_sk is the socket that needs to be migrated 6555 * to another listening socket. migrating_sk could be a fullsock 6556 * sk that is fully established or a reqsk that is in-the-middle 6557 * of 3-way handshake. 6558 */ 6559 __bpf_md_ptr(struct bpf_sock *, sk); 6560 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6561 }; 6562 6563 #define BPF_TAG_SIZE 8 6564 6565 struct bpf_prog_info { 6566 __u32 type; 6567 __u32 id; 6568 __u8 tag[BPF_TAG_SIZE]; 6569 __u32 jited_prog_len; 6570 __u32 xlated_prog_len; 6571 __aligned_u64 jited_prog_insns; 6572 __aligned_u64 xlated_prog_insns; 6573 __u64 load_time; /* ns since boottime */ 6574 __u32 created_by_uid; 6575 __u32 nr_map_ids; 6576 __aligned_u64 map_ids; 6577 char name[BPF_OBJ_NAME_LEN]; 6578 __u32 ifindex; 6579 __u32 gpl_compatible:1; 6580 __u32 :31; /* alignment pad */ 6581 __u64 netns_dev; 6582 __u64 netns_ino; 6583 __u32 nr_jited_ksyms; 6584 __u32 nr_jited_func_lens; 6585 __aligned_u64 jited_ksyms; 6586 __aligned_u64 jited_func_lens; 6587 __u32 btf_id; 6588 __u32 func_info_rec_size; 6589 __aligned_u64 func_info; 6590 __u32 nr_func_info; 6591 __u32 nr_line_info; 6592 __aligned_u64 line_info; 6593 __aligned_u64 jited_line_info; 6594 __u32 nr_jited_line_info; 6595 __u32 line_info_rec_size; 6596 __u32 jited_line_info_rec_size; 6597 __u32 nr_prog_tags; 6598 __aligned_u64 prog_tags; 6599 __u64 run_time_ns; 6600 __u64 run_cnt; 6601 __u64 recursion_misses; 6602 __u32 verified_insns; 6603 __u32 attach_btf_obj_id; 6604 __u32 attach_btf_id; 6605 } __attribute__((aligned(8))); 6606 6607 struct bpf_map_info { 6608 __u32 type; 6609 __u32 id; 6610 __u32 key_size; 6611 __u32 value_size; 6612 __u32 max_entries; 6613 __u32 map_flags; 6614 char name[BPF_OBJ_NAME_LEN]; 6615 __u32 ifindex; 6616 __u32 btf_vmlinux_value_type_id; 6617 __u64 netns_dev; 6618 __u64 netns_ino; 6619 __u32 btf_id; 6620 __u32 btf_key_type_id; 6621 __u32 btf_value_type_id; 6622 __u32 btf_vmlinux_id; 6623 __u64 map_extra; 6624 } __attribute__((aligned(8))); 6625 6626 struct bpf_btf_info { 6627 __aligned_u64 btf; 6628 __u32 btf_size; 6629 __u32 id; 6630 __aligned_u64 name; 6631 __u32 name_len; 6632 __u32 kernel_btf; 6633 } __attribute__((aligned(8))); 6634 6635 struct bpf_link_info { 6636 __u32 type; 6637 __u32 id; 6638 __u32 prog_id; 6639 union { 6640 struct { 6641 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6642 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6643 } raw_tracepoint; 6644 struct { 6645 __u32 attach_type; 6646 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6647 __u32 target_btf_id; /* BTF type id inside the object */ 6648 } tracing; 6649 struct { 6650 __u64 cgroup_id; 6651 __u32 attach_type; 6652 } cgroup; 6653 struct { 6654 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6655 __u32 target_name_len; /* in/out: target_name buffer len */ 6656 6657 /* If the iter specific field is 32 bits, it can be put 6658 * in the first or second union. Otherwise it should be 6659 * put in the second union. 6660 */ 6661 union { 6662 struct { 6663 __u32 map_id; 6664 } map; 6665 }; 6666 union { 6667 struct { 6668 __u64 cgroup_id; 6669 __u32 order; 6670 } cgroup; 6671 struct { 6672 __u32 tid; 6673 __u32 pid; 6674 } task; 6675 }; 6676 } iter; 6677 struct { 6678 __u32 netns_ino; 6679 __u32 attach_type; 6680 } netns; 6681 struct { 6682 __u32 ifindex; 6683 } xdp; 6684 struct { 6685 __u32 map_id; 6686 } struct_ops; 6687 struct { 6688 __u32 pf; 6689 __u32 hooknum; 6690 __s32 priority; 6691 __u32 flags; 6692 } netfilter; 6693 struct { 6694 __aligned_u64 addrs; 6695 __u32 count; /* in/out: kprobe_multi function count */ 6696 __u32 flags; 6697 __u64 missed; 6698 __aligned_u64 cookies; 6699 } kprobe_multi; 6700 struct { 6701 __aligned_u64 path; 6702 __aligned_u64 offsets; 6703 __aligned_u64 ref_ctr_offsets; 6704 __aligned_u64 cookies; 6705 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6706 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6707 __u32 flags; 6708 __u32 pid; 6709 } uprobe_multi; 6710 struct { 6711 __u32 type; /* enum bpf_perf_event_type */ 6712 __u32 :32; 6713 union { 6714 struct { 6715 __aligned_u64 file_name; /* in/out */ 6716 __u32 name_len; 6717 __u32 offset; /* offset from file_name */ 6718 __u64 cookie; 6719 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6720 struct { 6721 __aligned_u64 func_name; /* in/out */ 6722 __u32 name_len; 6723 __u32 offset; /* offset from func_name */ 6724 __u64 addr; 6725 __u64 missed; 6726 __u64 cookie; 6727 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6728 struct { 6729 __aligned_u64 tp_name; /* in/out */ 6730 __u32 name_len; 6731 __u32 :32; 6732 __u64 cookie; 6733 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6734 struct { 6735 __u64 config; 6736 __u32 type; 6737 __u32 :32; 6738 __u64 cookie; 6739 } event; /* BPF_PERF_EVENT_EVENT */ 6740 }; 6741 } perf_event; 6742 struct { 6743 __u32 ifindex; 6744 __u32 attach_type; 6745 } tcx; 6746 struct { 6747 __u32 ifindex; 6748 __u32 attach_type; 6749 } netkit; 6750 struct { 6751 __u32 map_id; 6752 __u32 attach_type; 6753 } sockmap; 6754 }; 6755 } __attribute__((aligned(8))); 6756 6757 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6758 * by user and intended to be used by socket (e.g. to bind to, depends on 6759 * attach type). 6760 */ 6761 struct bpf_sock_addr { 6762 __u32 user_family; /* Allows 4-byte read, but no write. */ 6763 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6764 * Stored in network byte order. 6765 */ 6766 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6767 * Stored in network byte order. 6768 */ 6769 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6770 * Stored in network byte order 6771 */ 6772 __u32 family; /* Allows 4-byte read, but no write */ 6773 __u32 type; /* Allows 4-byte read, but no write */ 6774 __u32 protocol; /* Allows 4-byte read, but no write */ 6775 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6776 * Stored in network byte order. 6777 */ 6778 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6779 * Stored in network byte order. 6780 */ 6781 __bpf_md_ptr(struct bpf_sock *, sk); 6782 }; 6783 6784 /* User bpf_sock_ops struct to access socket values and specify request ops 6785 * and their replies. 6786 * Some of this fields are in network (bigendian) byte order and may need 6787 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6788 * New fields can only be added at the end of this structure 6789 */ 6790 struct bpf_sock_ops { 6791 __u32 op; 6792 union { 6793 __u32 args[4]; /* Optionally passed to bpf program */ 6794 __u32 reply; /* Returned by bpf program */ 6795 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6796 }; 6797 __u32 family; 6798 __u32 remote_ip4; /* Stored in network byte order */ 6799 __u32 local_ip4; /* Stored in network byte order */ 6800 __u32 remote_ip6[4]; /* Stored in network byte order */ 6801 __u32 local_ip6[4]; /* Stored in network byte order */ 6802 __u32 remote_port; /* Stored in network byte order */ 6803 __u32 local_port; /* stored in host byte order */ 6804 __u32 is_fullsock; /* Some TCP fields are only valid if 6805 * there is a full socket. If not, the 6806 * fields read as zero. 6807 */ 6808 __u32 snd_cwnd; 6809 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6810 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6811 __u32 state; 6812 __u32 rtt_min; 6813 __u32 snd_ssthresh; 6814 __u32 rcv_nxt; 6815 __u32 snd_nxt; 6816 __u32 snd_una; 6817 __u32 mss_cache; 6818 __u32 ecn_flags; 6819 __u32 rate_delivered; 6820 __u32 rate_interval_us; 6821 __u32 packets_out; 6822 __u32 retrans_out; 6823 __u32 total_retrans; 6824 __u32 segs_in; 6825 __u32 data_segs_in; 6826 __u32 segs_out; 6827 __u32 data_segs_out; 6828 __u32 lost_out; 6829 __u32 sacked_out; 6830 __u32 sk_txhash; 6831 __u64 bytes_received; 6832 __u64 bytes_acked; 6833 __bpf_md_ptr(struct bpf_sock *, sk); 6834 /* [skb_data, skb_data_end) covers the whole TCP header. 6835 * 6836 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6837 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6838 * header has not been written. 6839 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6840 * been written so far. 6841 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6842 * the 3WHS. 6843 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6844 * the 3WHS. 6845 * 6846 * bpf_load_hdr_opt() can also be used to read a particular option. 6847 */ 6848 __bpf_md_ptr(void *, skb_data); 6849 __bpf_md_ptr(void *, skb_data_end); 6850 __u32 skb_len; /* The total length of a packet. 6851 * It includes the header, options, 6852 * and payload. 6853 */ 6854 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6855 * an easy way to check for tcp_flags 6856 * without parsing skb_data. 6857 * 6858 * In particular, the skb_tcp_flags 6859 * will still be available in 6860 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6861 * the outgoing header has not 6862 * been written yet. 6863 */ 6864 __u64 skb_hwtstamp; 6865 }; 6866 6867 /* Definitions for bpf_sock_ops_cb_flags */ 6868 enum { 6869 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6870 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6871 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6872 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6873 /* Call bpf for all received TCP headers. The bpf prog will be 6874 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6875 * 6876 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6877 * for the header option related helpers that will be useful 6878 * to the bpf programs. 6879 * 6880 * It could be used at the client/active side (i.e. connect() side) 6881 * when the server told it that the server was in syncookie 6882 * mode and required the active side to resend the bpf-written 6883 * options. The active side can keep writing the bpf-options until 6884 * it received a valid packet from the server side to confirm 6885 * the earlier packet (and options) has been received. The later 6886 * example patch is using it like this at the active side when the 6887 * server is in syncookie mode. 6888 * 6889 * The bpf prog will usually turn this off in the common cases. 6890 */ 6891 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6892 /* Call bpf when kernel has received a header option that 6893 * the kernel cannot handle. The bpf prog will be called under 6894 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6895 * 6896 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6897 * for the header option related helpers that will be useful 6898 * to the bpf programs. 6899 */ 6900 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6901 /* Call bpf when the kernel is writing header options for the 6902 * outgoing packet. The bpf prog will first be called 6903 * to reserve space in a skb under 6904 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6905 * the bpf prog will be called to write the header option(s) 6906 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6907 * 6908 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6909 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6910 * related helpers that will be useful to the bpf programs. 6911 * 6912 * The kernel gets its chance to reserve space and write 6913 * options first before the BPF program does. 6914 */ 6915 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6916 /* Mask of all currently supported cb flags */ 6917 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6918 }; 6919 6920 /* List of known BPF sock_ops operators. 6921 * New entries can only be added at the end 6922 */ 6923 enum { 6924 BPF_SOCK_OPS_VOID, 6925 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6926 * -1 if default value should be used 6927 */ 6928 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6929 * window (in packets) or -1 if default 6930 * value should be used 6931 */ 6932 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6933 * active connection is initialized 6934 */ 6935 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6936 * active connection is 6937 * established 6938 */ 6939 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6940 * passive connection is 6941 * established 6942 */ 6943 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6944 * needs ECN 6945 */ 6946 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6947 * based on the path and may be 6948 * dependent on the congestion control 6949 * algorithm. In general it indicates 6950 * a congestion threshold. RTTs above 6951 * this indicate congestion 6952 */ 6953 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6954 * Arg1: value of icsk_retransmits 6955 * Arg2: value of icsk_rto 6956 * Arg3: whether RTO has expired 6957 */ 6958 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6959 * Arg1: sequence number of 1st byte 6960 * Arg2: # segments 6961 * Arg3: return value of 6962 * tcp_transmit_skb (0 => success) 6963 */ 6964 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6965 * Arg1: old_state 6966 * Arg2: new_state 6967 */ 6968 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6969 * socket transition to LISTEN state. 6970 */ 6971 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6972 * Arg1: measured RTT input (mrtt) 6973 * Arg2: updated srtt 6974 */ 6975 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6976 * It will be called to handle 6977 * the packets received at 6978 * an already established 6979 * connection. 6980 * 6981 * sock_ops->skb_data: 6982 * Referring to the received skb. 6983 * It covers the TCP header only. 6984 * 6985 * bpf_load_hdr_opt() can also 6986 * be used to search for a 6987 * particular option. 6988 */ 6989 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6990 * header option later in 6991 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6992 * Arg1: bool want_cookie. (in 6993 * writing SYNACK only) 6994 * 6995 * sock_ops->skb_data: 6996 * Not available because no header has 6997 * been written yet. 6998 * 6999 * sock_ops->skb_tcp_flags: 7000 * The tcp_flags of the 7001 * outgoing skb. (e.g. SYN, ACK, FIN). 7002 * 7003 * bpf_reserve_hdr_opt() should 7004 * be used to reserve space. 7005 */ 7006 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 7007 * Arg1: bool want_cookie. (in 7008 * writing SYNACK only) 7009 * 7010 * sock_ops->skb_data: 7011 * Referring to the outgoing skb. 7012 * It covers the TCP header 7013 * that has already been written 7014 * by the kernel and the 7015 * earlier bpf-progs. 7016 * 7017 * sock_ops->skb_tcp_flags: 7018 * The tcp_flags of the outgoing 7019 * skb. (e.g. SYN, ACK, FIN). 7020 * 7021 * bpf_store_hdr_opt() should 7022 * be used to write the 7023 * option. 7024 * 7025 * bpf_load_hdr_opt() can also 7026 * be used to search for a 7027 * particular option that 7028 * has already been written 7029 * by the kernel or the 7030 * earlier bpf-progs. 7031 */ 7032 }; 7033 7034 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 7035 * changes between the TCP and BPF versions. Ideally this should never happen. 7036 * If it does, we need to add code to convert them before calling 7037 * the BPF sock_ops function. 7038 */ 7039 enum { 7040 BPF_TCP_ESTABLISHED = 1, 7041 BPF_TCP_SYN_SENT, 7042 BPF_TCP_SYN_RECV, 7043 BPF_TCP_FIN_WAIT1, 7044 BPF_TCP_FIN_WAIT2, 7045 BPF_TCP_TIME_WAIT, 7046 BPF_TCP_CLOSE, 7047 BPF_TCP_CLOSE_WAIT, 7048 BPF_TCP_LAST_ACK, 7049 BPF_TCP_LISTEN, 7050 BPF_TCP_CLOSING, /* Now a valid state */ 7051 BPF_TCP_NEW_SYN_RECV, 7052 BPF_TCP_BOUND_INACTIVE, 7053 7054 BPF_TCP_MAX_STATES /* Leave at the end! */ 7055 }; 7056 7057 enum { 7058 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 7059 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 7060 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 7061 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 7062 /* Copy the SYN pkt to optval 7063 * 7064 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 7065 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 7066 * to only getting from the saved_syn. It can either get the 7067 * syn packet from: 7068 * 7069 * 1. the just-received SYN packet (only available when writing the 7070 * SYNACK). It will be useful when it is not necessary to 7071 * save the SYN packet for latter use. It is also the only way 7072 * to get the SYN during syncookie mode because the syn 7073 * packet cannot be saved during syncookie. 7074 * 7075 * OR 7076 * 7077 * 2. the earlier saved syn which was done by 7078 * bpf_setsockopt(TCP_SAVE_SYN). 7079 * 7080 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7081 * SYN packet is obtained. 7082 * 7083 * If the bpf-prog does not need the IP[46] header, the 7084 * bpf-prog can avoid parsing the IP header by using 7085 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7086 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7087 * 7088 * >0: Total number of bytes copied 7089 * -ENOSPC: Not enough space in optval. Only optlen number of 7090 * bytes is copied. 7091 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7092 * is not saved by setsockopt(TCP_SAVE_SYN). 7093 */ 7094 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7095 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7096 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7097 TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */ 7098 }; 7099 7100 enum { 7101 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7102 }; 7103 7104 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7105 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7106 */ 7107 enum { 7108 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7109 * total option spaces 7110 * required for an established 7111 * sk in order to calculate the 7112 * MSS. No skb is actually 7113 * sent. 7114 */ 7115 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7116 * when sending a SYN. 7117 */ 7118 }; 7119 7120 struct bpf_perf_event_value { 7121 __u64 counter; 7122 __u64 enabled; 7123 __u64 running; 7124 }; 7125 7126 enum { 7127 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7128 BPF_DEVCG_ACC_READ = (1ULL << 1), 7129 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7130 }; 7131 7132 enum { 7133 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7134 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7135 }; 7136 7137 struct bpf_cgroup_dev_ctx { 7138 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7139 __u32 access_type; 7140 __u32 major; 7141 __u32 minor; 7142 }; 7143 7144 struct bpf_raw_tracepoint_args { 7145 __u64 args[0]; 7146 }; 7147 7148 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7149 * OUTPUT: Do lookup from egress perspective; default is ingress 7150 */ 7151 enum { 7152 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7153 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7154 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7155 BPF_FIB_LOOKUP_TBID = (1U << 3), 7156 BPF_FIB_LOOKUP_SRC = (1U << 4), 7157 BPF_FIB_LOOKUP_MARK = (1U << 5), 7158 }; 7159 7160 enum { 7161 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7162 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7163 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7164 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7165 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7166 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7167 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7168 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7169 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7170 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7171 }; 7172 7173 struct bpf_fib_lookup { 7174 /* input: network family for lookup (AF_INET, AF_INET6) 7175 * output: network family of egress nexthop 7176 */ 7177 __u8 family; 7178 7179 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7180 __u8 l4_protocol; 7181 __be16 sport; 7182 __be16 dport; 7183 7184 union { /* used for MTU check */ 7185 /* input to lookup */ 7186 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7187 7188 /* output: MTU value */ 7189 __u16 mtu_result; 7190 } __attribute__((packed, aligned(2))); 7191 /* input: L3 device index for lookup 7192 * output: device index from FIB lookup 7193 */ 7194 __u32 ifindex; 7195 7196 union { 7197 /* inputs to lookup */ 7198 __u8 tos; /* AF_INET */ 7199 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7200 7201 /* output: metric of fib result (IPv4/IPv6 only) */ 7202 __u32 rt_metric; 7203 }; 7204 7205 /* input: source address to consider for lookup 7206 * output: source address result from lookup 7207 */ 7208 union { 7209 __be32 ipv4_src; 7210 __u32 ipv6_src[4]; /* in6_addr; network order */ 7211 }; 7212 7213 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7214 * network header. output: bpf_fib_lookup sets to gateway address 7215 * if FIB lookup returns gateway route 7216 */ 7217 union { 7218 __be32 ipv4_dst; 7219 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7220 }; 7221 7222 union { 7223 struct { 7224 /* output */ 7225 __be16 h_vlan_proto; 7226 __be16 h_vlan_TCI; 7227 }; 7228 /* input: when accompanied with the 7229 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7230 * specific routing table to use for the fib lookup. 7231 */ 7232 __u32 tbid; 7233 }; 7234 7235 union { 7236 /* input */ 7237 struct { 7238 __u32 mark; /* policy routing */ 7239 /* 2 4-byte holes for input */ 7240 }; 7241 7242 /* output: source and dest mac */ 7243 struct { 7244 __u8 smac[6]; /* ETH_ALEN */ 7245 __u8 dmac[6]; /* ETH_ALEN */ 7246 }; 7247 }; 7248 }; 7249 7250 struct bpf_redir_neigh { 7251 /* network family for lookup (AF_INET, AF_INET6) */ 7252 __u32 nh_family; 7253 /* network address of nexthop; skips fib lookup to find gateway */ 7254 union { 7255 __be32 ipv4_nh; 7256 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7257 }; 7258 }; 7259 7260 /* bpf_check_mtu flags*/ 7261 enum bpf_check_mtu_flags { 7262 BPF_MTU_CHK_SEGS = (1U << 0), 7263 }; 7264 7265 enum bpf_check_mtu_ret { 7266 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7267 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7268 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7269 }; 7270 7271 enum bpf_task_fd_type { 7272 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7273 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7274 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7275 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7276 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7277 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7278 }; 7279 7280 enum { 7281 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7282 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7283 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7284 }; 7285 7286 struct bpf_flow_keys { 7287 __u16 nhoff; 7288 __u16 thoff; 7289 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7290 __u8 is_frag; 7291 __u8 is_first_frag; 7292 __u8 is_encap; 7293 __u8 ip_proto; 7294 __be16 n_proto; 7295 __be16 sport; 7296 __be16 dport; 7297 union { 7298 struct { 7299 __be32 ipv4_src; 7300 __be32 ipv4_dst; 7301 }; 7302 struct { 7303 __u32 ipv6_src[4]; /* in6_addr; network order */ 7304 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7305 }; 7306 }; 7307 __u32 flags; 7308 __be32 flow_label; 7309 }; 7310 7311 struct bpf_func_info { 7312 __u32 insn_off; 7313 __u32 type_id; 7314 }; 7315 7316 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7317 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7318 7319 struct bpf_line_info { 7320 __u32 insn_off; 7321 __u32 file_name_off; 7322 __u32 line_off; 7323 __u32 line_col; 7324 }; 7325 7326 struct bpf_spin_lock { 7327 __u32 val; 7328 }; 7329 7330 struct bpf_timer { 7331 __u64 __opaque[2]; 7332 } __attribute__((aligned(8))); 7333 7334 struct bpf_wq { 7335 __u64 __opaque[2]; 7336 } __attribute__((aligned(8))); 7337 7338 struct bpf_dynptr { 7339 __u64 __opaque[2]; 7340 } __attribute__((aligned(8))); 7341 7342 struct bpf_list_head { 7343 __u64 __opaque[2]; 7344 } __attribute__((aligned(8))); 7345 7346 struct bpf_list_node { 7347 __u64 __opaque[3]; 7348 } __attribute__((aligned(8))); 7349 7350 struct bpf_rb_root { 7351 __u64 __opaque[2]; 7352 } __attribute__((aligned(8))); 7353 7354 struct bpf_rb_node { 7355 __u64 __opaque[4]; 7356 } __attribute__((aligned(8))); 7357 7358 struct bpf_refcount { 7359 __u32 __opaque[1]; 7360 } __attribute__((aligned(4))); 7361 7362 struct bpf_sysctl { 7363 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7364 * Allows 1,2,4-byte read, but no write. 7365 */ 7366 __u32 file_pos; /* Sysctl file position to read from, write to. 7367 * Allows 1,2,4-byte read an 4-byte write. 7368 */ 7369 }; 7370 7371 struct bpf_sockopt { 7372 __bpf_md_ptr(struct bpf_sock *, sk); 7373 __bpf_md_ptr(void *, optval); 7374 __bpf_md_ptr(void *, optval_end); 7375 7376 __s32 level; 7377 __s32 optname; 7378 __s32 optlen; 7379 __s32 retval; 7380 }; 7381 7382 struct bpf_pidns_info { 7383 __u32 pid; 7384 __u32 tgid; 7385 }; 7386 7387 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7388 struct bpf_sk_lookup { 7389 union { 7390 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7391 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7392 }; 7393 7394 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7395 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7396 __u32 remote_ip4; /* Network byte order */ 7397 __u32 remote_ip6[4]; /* Network byte order */ 7398 __be16 remote_port; /* Network byte order */ 7399 __u16 :16; /* Zero padding */ 7400 __u32 local_ip4; /* Network byte order */ 7401 __u32 local_ip6[4]; /* Network byte order */ 7402 __u32 local_port; /* Host byte order */ 7403 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7404 }; 7405 7406 /* 7407 * struct btf_ptr is used for typed pointer representation; the 7408 * type id is used to render the pointer data as the appropriate type 7409 * via the bpf_snprintf_btf() helper described above. A flags field - 7410 * potentially to specify additional details about the BTF pointer 7411 * (rather than its mode of display) - is included for future use. 7412 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7413 */ 7414 struct btf_ptr { 7415 void *ptr; 7416 __u32 type_id; 7417 __u32 flags; /* BTF ptr flags; unused at present. */ 7418 }; 7419 7420 /* 7421 * Flags to control bpf_snprintf_btf() behaviour. 7422 * - BTF_F_COMPACT: no formatting around type information 7423 * - BTF_F_NONAME: no struct/union member names/types 7424 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7425 * equivalent to %px. 7426 * - BTF_F_ZERO: show zero-valued struct/union members; they 7427 * are not displayed by default 7428 */ 7429 enum { 7430 BTF_F_COMPACT = (1ULL << 0), 7431 BTF_F_NONAME = (1ULL << 1), 7432 BTF_F_PTR_RAW = (1ULL << 2), 7433 BTF_F_ZERO = (1ULL << 3), 7434 }; 7435 7436 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7437 * has to be adjusted by relocations. It is emitted by llvm and passed to 7438 * libbpf and later to the kernel. 7439 */ 7440 enum bpf_core_relo_kind { 7441 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7442 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7443 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7444 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7445 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7446 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7447 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7448 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7449 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7450 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7451 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7452 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7453 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7454 }; 7455 7456 /* 7457 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7458 * and from libbpf to the kernel. 7459 * 7460 * CO-RE relocation captures the following data: 7461 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7462 * its insn->imm field to be relocated with actual field info; 7463 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7464 * type or field; 7465 * - access_str_off - offset into corresponding .BTF string section. String 7466 * interpretation depends on specific relocation kind: 7467 * - for field-based relocations, string encodes an accessed field using 7468 * a sequence of field and array indices, separated by colon (:). It's 7469 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7470 * arguments for identifying offset to a field. 7471 * - for type-based relocations, strings is expected to be just "0"; 7472 * - for enum value-based relocations, string contains an index of enum 7473 * value within its enum type; 7474 * - kind - one of enum bpf_core_relo_kind; 7475 * 7476 * Example: 7477 * struct sample { 7478 * int a; 7479 * struct { 7480 * int b[10]; 7481 * }; 7482 * }; 7483 * 7484 * struct sample *s = ...; 7485 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7486 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7487 * // b is field #0 inside anon struct, accessing elem #5) 7488 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7489 * 7490 * type_id for all relocs in this example will capture BTF type id of 7491 * `struct sample`. 7492 * 7493 * Such relocation is emitted when using __builtin_preserve_access_index() 7494 * Clang built-in, passing expression that captures field address, e.g.: 7495 * 7496 * bpf_probe_read(&dst, sizeof(dst), 7497 * __builtin_preserve_access_index(&src->a.b.c)); 7498 * 7499 * In this case Clang will emit field relocation recording necessary data to 7500 * be able to find offset of embedded `a.b.c` field within `src` struct. 7501 * 7502 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7503 */ 7504 struct bpf_core_relo { 7505 __u32 insn_off; 7506 __u32 type_id; 7507 __u32 access_str_off; 7508 enum bpf_core_relo_kind kind; 7509 }; 7510 7511 /* 7512 * Flags to control bpf_timer_start() behaviour. 7513 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7514 * relative to current time. 7515 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7516 */ 7517 enum { 7518 BPF_F_TIMER_ABS = (1ULL << 0), 7519 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7520 }; 7521 7522 /* BPF numbers iterator state */ 7523 struct bpf_iter_num { 7524 /* opaque iterator state; having __u64 here allows to preserve correct 7525 * alignment requirements in vmlinux.h, generated from BTF 7526 */ 7527 __u64 __opaque[1]; 7528 } __attribute__((aligned(8))); 7529 7530 /* 7531 * Flags to control BPF kfunc behaviour. 7532 * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective 7533 * helper documentation for details.) 7534 */ 7535 enum bpf_kfunc_flags { 7536 BPF_F_PAD_ZEROS = (1ULL << 0), 7537 }; 7538 7539 #endif /* _UAPI__LINUX_BPF_H__ */ 7540