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 #define BPF_LOAD_ACQ 0x100 /* load-acquire */ 55 #define BPF_STORE_REL 0x110 /* store-release */ 56 57 enum bpf_cond_pseudo_jmp { 58 BPF_MAY_GOTO = 0, 59 }; 60 61 /* Register numbers */ 62 enum { 63 BPF_REG_0 = 0, 64 BPF_REG_1, 65 BPF_REG_2, 66 BPF_REG_3, 67 BPF_REG_4, 68 BPF_REG_5, 69 BPF_REG_6, 70 BPF_REG_7, 71 BPF_REG_8, 72 BPF_REG_9, 73 BPF_REG_10, 74 __MAX_BPF_REG, 75 }; 76 77 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 78 #define MAX_BPF_REG __MAX_BPF_REG 79 80 struct bpf_insn { 81 __u8 code; /* opcode */ 82 __u8 dst_reg:4; /* dest register */ 83 __u8 src_reg:4; /* source register */ 84 __s16 off; /* signed offset */ 85 __s32 imm; /* signed immediate constant */ 86 }; 87 88 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 89 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 90 * the trailing flexible array member) instead. 91 */ 92 struct bpf_lpm_trie_key { 93 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 94 __u8 data[0]; /* Arbitrary size */ 95 }; 96 97 /* Header for bpf_lpm_trie_key structs */ 98 struct bpf_lpm_trie_key_hdr { 99 __u32 prefixlen; 100 }; 101 102 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 103 struct bpf_lpm_trie_key_u8 { 104 union { 105 struct bpf_lpm_trie_key_hdr hdr; 106 __u32 prefixlen; 107 }; 108 __u8 data[]; /* Arbitrary size */ 109 }; 110 111 struct bpf_cgroup_storage_key { 112 __u64 cgroup_inode_id; /* cgroup inode id */ 113 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 114 }; 115 116 enum bpf_cgroup_iter_order { 117 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 118 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 119 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 120 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 121 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 122 }; 123 124 union bpf_iter_link_info { 125 struct { 126 __u32 map_fd; 127 } map; 128 struct { 129 enum bpf_cgroup_iter_order order; 130 131 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 132 * both are zero, the walk starts from the default cgroup v2 133 * root. For walking v1 hierarchy, one should always explicitly 134 * specify cgroup_fd. 135 */ 136 __u32 cgroup_fd; 137 __u64 cgroup_id; 138 } cgroup; 139 /* Parameters of task iterators. */ 140 struct { 141 __u32 tid; 142 __u32 pid; 143 __u32 pid_fd; 144 } task; 145 }; 146 147 /* BPF syscall commands, see bpf(2) man-page for more details. */ 148 /** 149 * DOC: eBPF Syscall Preamble 150 * 151 * The operation to be performed by the **bpf**\ () system call is determined 152 * by the *cmd* argument. Each operation takes an accompanying argument, 153 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 154 * below). The size argument is the size of the union pointed to by *attr*. 155 */ 156 /** 157 * DOC: eBPF Syscall Commands 158 * 159 * BPF_MAP_CREATE 160 * Description 161 * Create a map and return a file descriptor that refers to the 162 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 163 * is automatically enabled for the new file descriptor. 164 * 165 * Applying **close**\ (2) to the file descriptor returned by 166 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 167 * 168 * Return 169 * A new file descriptor (a nonnegative integer), or -1 if an 170 * error occurred (in which case, *errno* is set appropriately). 171 * 172 * BPF_MAP_LOOKUP_ELEM 173 * Description 174 * Look up an element with a given *key* in the map referred to 175 * by the file descriptor *map_fd*. 176 * 177 * The *flags* argument may be specified as one of the 178 * following: 179 * 180 * **BPF_F_LOCK** 181 * Look up the value of a spin-locked map without 182 * returning the lock. This must be specified if the 183 * elements contain a spinlock. 184 * 185 * Return 186 * Returns zero on success. On error, -1 is returned and *errno* 187 * is set appropriately. 188 * 189 * BPF_MAP_UPDATE_ELEM 190 * Description 191 * Create or update an element (key/value pair) in a specified map. 192 * 193 * The *flags* argument should be specified as one of the 194 * following: 195 * 196 * **BPF_ANY** 197 * Create a new element or update an existing element. 198 * **BPF_NOEXIST** 199 * Create a new element only if it did not exist. 200 * **BPF_EXIST** 201 * Update an existing element. 202 * **BPF_F_LOCK** 203 * Update a spin_lock-ed map element. 204 * 205 * Return 206 * Returns zero on success. On error, -1 is returned and *errno* 207 * is set appropriately. 208 * 209 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 210 * **E2BIG**, **EEXIST**, or **ENOENT**. 211 * 212 * **E2BIG** 213 * The number of elements in the map reached the 214 * *max_entries* limit specified at map creation time. 215 * **EEXIST** 216 * If *flags* specifies **BPF_NOEXIST** and the element 217 * with *key* already exists in the map. 218 * **ENOENT** 219 * If *flags* specifies **BPF_EXIST** and the element with 220 * *key* does not exist in the map. 221 * 222 * BPF_MAP_DELETE_ELEM 223 * Description 224 * Look up and delete an element by key in a specified map. 225 * 226 * Return 227 * Returns zero on success. On error, -1 is returned and *errno* 228 * is set appropriately. 229 * 230 * BPF_MAP_GET_NEXT_KEY 231 * Description 232 * Look up an element by key in a specified map and return the key 233 * of the next element. Can be used to iterate over all elements 234 * in the map. 235 * 236 * Return 237 * Returns zero on success. On error, -1 is returned and *errno* 238 * is set appropriately. 239 * 240 * The following cases can be used to iterate over all elements of 241 * the map: 242 * 243 * * If *key* is not found, the operation returns zero and sets 244 * the *next_key* pointer to the key of the first element. 245 * * If *key* is found, the operation returns zero and sets the 246 * *next_key* pointer to the key of the next element. 247 * * If *key* is the last element, returns -1 and *errno* is set 248 * to **ENOENT**. 249 * 250 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 251 * **EINVAL** on error. 252 * 253 * BPF_PROG_LOAD 254 * Description 255 * Verify and load an eBPF program, returning a new file 256 * descriptor associated with the program. 257 * 258 * Applying **close**\ (2) to the file descriptor returned by 259 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 260 * 261 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 262 * automatically enabled for the new file descriptor. 263 * 264 * Return 265 * A new file descriptor (a nonnegative integer), or -1 if an 266 * error occurred (in which case, *errno* is set appropriately). 267 * 268 * BPF_OBJ_PIN 269 * Description 270 * Pin an eBPF program or map referred by the specified *bpf_fd* 271 * to the provided *pathname* on the filesystem. 272 * 273 * The *pathname* argument must not contain a dot ("."). 274 * 275 * On success, *pathname* retains a reference to the eBPF object, 276 * preventing deallocation of the object when the original 277 * *bpf_fd* is closed. This allow the eBPF object to live beyond 278 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 279 * process. 280 * 281 * Applying **unlink**\ (2) or similar calls to the *pathname* 282 * unpins the object from the filesystem, removing the reference. 283 * If no other file descriptors or filesystem nodes refer to the 284 * same object, it will be deallocated (see NOTES). 285 * 286 * The filesystem type for the parent directory of *pathname* must 287 * be **BPF_FS_MAGIC**. 288 * 289 * Return 290 * Returns zero on success. On error, -1 is returned and *errno* 291 * is set appropriately. 292 * 293 * BPF_OBJ_GET 294 * Description 295 * Open a file descriptor for the eBPF object pinned to the 296 * specified *pathname*. 297 * 298 * Return 299 * A new file descriptor (a nonnegative integer), or -1 if an 300 * error occurred (in which case, *errno* is set appropriately). 301 * 302 * BPF_PROG_ATTACH 303 * Description 304 * Attach an eBPF program to a *target_fd* at the specified 305 * *attach_type* hook. 306 * 307 * The *attach_type* specifies the eBPF attachment point to 308 * attach the program to, and must be one of *bpf_attach_type* 309 * (see below). 310 * 311 * The *attach_bpf_fd* must be a valid file descriptor for a 312 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 313 * or sock_ops type corresponding to the specified *attach_type*. 314 * 315 * The *target_fd* must be a valid file descriptor for a kernel 316 * object which depends on the attach type of *attach_bpf_fd*: 317 * 318 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 319 * **BPF_PROG_TYPE_CGROUP_SKB**, 320 * **BPF_PROG_TYPE_CGROUP_SOCK**, 321 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 322 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 323 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 324 * **BPF_PROG_TYPE_SOCK_OPS** 325 * 326 * Control Group v2 hierarchy with the eBPF controller 327 * enabled. Requires the kernel to be compiled with 328 * **CONFIG_CGROUP_BPF**. 329 * 330 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 331 * 332 * Network namespace (eg /proc/self/ns/net). 333 * 334 * **BPF_PROG_TYPE_LIRC_MODE2** 335 * 336 * LIRC device path (eg /dev/lircN). Requires the kernel 337 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 338 * 339 * **BPF_PROG_TYPE_SK_SKB**, 340 * **BPF_PROG_TYPE_SK_MSG** 341 * 342 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 343 * 344 * Return 345 * Returns zero on success. On error, -1 is returned and *errno* 346 * is set appropriately. 347 * 348 * BPF_PROG_DETACH 349 * Description 350 * Detach the eBPF program associated with the *target_fd* at the 351 * hook specified by *attach_type*. The program must have been 352 * previously attached using **BPF_PROG_ATTACH**. 353 * 354 * Return 355 * Returns zero on success. On error, -1 is returned and *errno* 356 * is set appropriately. 357 * 358 * BPF_PROG_TEST_RUN 359 * Description 360 * Run the eBPF program associated with the *prog_fd* a *repeat* 361 * number of times against a provided program context *ctx_in* and 362 * data *data_in*, and return the modified program context 363 * *ctx_out*, *data_out* (for example, packet data), result of the 364 * execution *retval*, and *duration* of the test run. 365 * 366 * The sizes of the buffers provided as input and output 367 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 368 * be provided in the corresponding variables *ctx_size_in*, 369 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 370 * of these parameters are not provided (ie set to NULL), the 371 * corresponding size field must be zero. 372 * 373 * Some program types have particular requirements: 374 * 375 * **BPF_PROG_TYPE_SK_LOOKUP** 376 * *data_in* and *data_out* must be NULL. 377 * 378 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 379 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 380 * 381 * *ctx_out*, *data_in* and *data_out* must be NULL. 382 * *repeat* must be zero. 383 * 384 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 385 * 386 * Return 387 * Returns zero on success. On error, -1 is returned and *errno* 388 * is set appropriately. 389 * 390 * **ENOSPC** 391 * Either *data_size_out* or *ctx_size_out* is too small. 392 * **ENOTSUPP** 393 * This command is not supported by the program type of 394 * the program referred to by *prog_fd*. 395 * 396 * BPF_PROG_GET_NEXT_ID 397 * Description 398 * Fetch the next eBPF program currently loaded into the kernel. 399 * 400 * Looks for the eBPF program with an id greater than *start_id* 401 * and updates *next_id* on success. If no other eBPF programs 402 * remain with ids higher than *start_id*, returns -1 and sets 403 * *errno* to **ENOENT**. 404 * 405 * Return 406 * Returns zero on success. On error, or when no id remains, -1 407 * is returned and *errno* is set appropriately. 408 * 409 * BPF_MAP_GET_NEXT_ID 410 * Description 411 * Fetch the next eBPF map currently loaded into the kernel. 412 * 413 * Looks for the eBPF map with an id greater than *start_id* 414 * and updates *next_id* on success. If no other eBPF maps 415 * remain with ids higher than *start_id*, returns -1 and sets 416 * *errno* to **ENOENT**. 417 * 418 * Return 419 * Returns zero on success. On error, or when no id remains, -1 420 * is returned and *errno* is set appropriately. 421 * 422 * BPF_PROG_GET_FD_BY_ID 423 * Description 424 * Open a file descriptor for the eBPF program corresponding to 425 * *prog_id*. 426 * 427 * Return 428 * A new file descriptor (a nonnegative integer), or -1 if an 429 * error occurred (in which case, *errno* is set appropriately). 430 * 431 * BPF_MAP_GET_FD_BY_ID 432 * Description 433 * Open a file descriptor for the eBPF map corresponding to 434 * *map_id*. 435 * 436 * Return 437 * A new file descriptor (a nonnegative integer), or -1 if an 438 * error occurred (in which case, *errno* is set appropriately). 439 * 440 * BPF_OBJ_GET_INFO_BY_FD 441 * Description 442 * Obtain information about the eBPF object corresponding to 443 * *bpf_fd*. 444 * 445 * Populates up to *info_len* bytes of *info*, which will be in 446 * one of the following formats depending on the eBPF object type 447 * of *bpf_fd*: 448 * 449 * * **struct bpf_prog_info** 450 * * **struct bpf_map_info** 451 * * **struct bpf_btf_info** 452 * * **struct bpf_link_info** 453 * 454 * Return 455 * Returns zero on success. On error, -1 is returned and *errno* 456 * is set appropriately. 457 * 458 * BPF_PROG_QUERY 459 * Description 460 * Obtain information about eBPF programs associated with the 461 * specified *attach_type* hook. 462 * 463 * The *target_fd* must be a valid file descriptor for a kernel 464 * object which depends on the attach type of *attach_bpf_fd*: 465 * 466 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 467 * **BPF_PROG_TYPE_CGROUP_SKB**, 468 * **BPF_PROG_TYPE_CGROUP_SOCK**, 469 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 470 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 471 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 472 * **BPF_PROG_TYPE_SOCK_OPS** 473 * 474 * Control Group v2 hierarchy with the eBPF controller 475 * enabled. Requires the kernel to be compiled with 476 * **CONFIG_CGROUP_BPF**. 477 * 478 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 479 * 480 * Network namespace (eg /proc/self/ns/net). 481 * 482 * **BPF_PROG_TYPE_LIRC_MODE2** 483 * 484 * LIRC device path (eg /dev/lircN). Requires the kernel 485 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 486 * 487 * **BPF_PROG_QUERY** always fetches the number of programs 488 * attached and the *attach_flags* which were used to attach those 489 * programs. Additionally, if *prog_ids* is nonzero and the number 490 * of attached programs is less than *prog_cnt*, populates 491 * *prog_ids* with the eBPF program ids of the programs attached 492 * at *target_fd*. 493 * 494 * The following flags may alter the result: 495 * 496 * **BPF_F_QUERY_EFFECTIVE** 497 * Only return information regarding programs which are 498 * currently effective at the specified *target_fd*. 499 * 500 * Return 501 * Returns zero on success. On error, -1 is returned and *errno* 502 * is set appropriately. 503 * 504 * BPF_RAW_TRACEPOINT_OPEN 505 * Description 506 * Attach an eBPF program to a tracepoint *name* to access kernel 507 * internal arguments of the tracepoint in their raw form. 508 * 509 * The *prog_fd* must be a valid file descriptor associated with 510 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 511 * 512 * No ABI guarantees are made about the content of tracepoint 513 * arguments exposed to the corresponding eBPF program. 514 * 515 * Applying **close**\ (2) to the file descriptor returned by 516 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 517 * 518 * Return 519 * A new file descriptor (a nonnegative integer), or -1 if an 520 * error occurred (in which case, *errno* is set appropriately). 521 * 522 * BPF_BTF_LOAD 523 * Description 524 * Verify and load BPF Type Format (BTF) metadata into the kernel, 525 * returning a new file descriptor associated with the metadata. 526 * BTF is described in more detail at 527 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 528 * 529 * The *btf* parameter must point to valid memory providing 530 * *btf_size* bytes of BTF binary metadata. 531 * 532 * The returned file descriptor can be passed to other **bpf**\ () 533 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 534 * associate the BTF with those objects. 535 * 536 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 537 * parameters to specify a *btf_log_buf*, *btf_log_size* and 538 * *btf_log_level* which allow the kernel to return freeform log 539 * output regarding the BTF verification process. 540 * 541 * Return 542 * A new file descriptor (a nonnegative integer), or -1 if an 543 * error occurred (in which case, *errno* is set appropriately). 544 * 545 * BPF_BTF_GET_FD_BY_ID 546 * Description 547 * Open a file descriptor for the BPF Type Format (BTF) 548 * corresponding to *btf_id*. 549 * 550 * Return 551 * A new file descriptor (a nonnegative integer), or -1 if an 552 * error occurred (in which case, *errno* is set appropriately). 553 * 554 * BPF_TASK_FD_QUERY 555 * Description 556 * Obtain information about eBPF programs associated with the 557 * target process identified by *pid* and *fd*. 558 * 559 * If the *pid* and *fd* are associated with a tracepoint, kprobe 560 * or uprobe perf event, then the *prog_id* and *fd_type* will 561 * be populated with the eBPF program id and file descriptor type 562 * of type **bpf_task_fd_type**. If associated with a kprobe or 563 * uprobe, the *probe_offset* and *probe_addr* will also be 564 * populated. Optionally, if *buf* is provided, then up to 565 * *buf_len* bytes of *buf* will be populated with the name of 566 * the tracepoint, kprobe or uprobe. 567 * 568 * The resulting *prog_id* may be introspected in deeper detail 569 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 570 * 571 * Return 572 * Returns zero on success. On error, -1 is returned and *errno* 573 * is set appropriately. 574 * 575 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 576 * Description 577 * Look up an element with the given *key* in the map referred to 578 * by the file descriptor *fd*, and if found, delete the element. 579 * 580 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 581 * types, the *flags* argument needs to be set to 0, but for other 582 * map types, it may be specified as: 583 * 584 * **BPF_F_LOCK** 585 * Look up and delete the value of a spin-locked map 586 * without returning the lock. This must be specified if 587 * the elements contain a spinlock. 588 * 589 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 590 * implement this command as a "pop" operation, deleting the top 591 * element rather than one corresponding to *key*. 592 * The *key* and *key_len* parameters should be zeroed when 593 * issuing this operation for these map types. 594 * 595 * This command is only valid for the following map types: 596 * * **BPF_MAP_TYPE_QUEUE** 597 * * **BPF_MAP_TYPE_STACK** 598 * * **BPF_MAP_TYPE_HASH** 599 * * **BPF_MAP_TYPE_PERCPU_HASH** 600 * * **BPF_MAP_TYPE_LRU_HASH** 601 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 602 * 603 * Return 604 * Returns zero on success. On error, -1 is returned and *errno* 605 * is set appropriately. 606 * 607 * BPF_MAP_FREEZE 608 * Description 609 * Freeze the permissions of the specified map. 610 * 611 * Write permissions may be frozen by passing zero *flags*. 612 * Upon success, no future syscall invocations may alter the 613 * map state of *map_fd*. Write operations from eBPF programs 614 * are still possible for a frozen map. 615 * 616 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 617 * 618 * Return 619 * Returns zero on success. On error, -1 is returned and *errno* 620 * is set appropriately. 621 * 622 * BPF_BTF_GET_NEXT_ID 623 * Description 624 * Fetch the next BPF Type Format (BTF) object currently loaded 625 * into the kernel. 626 * 627 * Looks for the BTF object with an id greater than *start_id* 628 * and updates *next_id* on success. If no other BTF objects 629 * remain with ids higher than *start_id*, returns -1 and sets 630 * *errno* to **ENOENT**. 631 * 632 * Return 633 * Returns zero on success. On error, or when no id remains, -1 634 * is returned and *errno* is set appropriately. 635 * 636 * BPF_MAP_LOOKUP_BATCH 637 * Description 638 * Iterate and fetch multiple elements in a map. 639 * 640 * Two opaque values are used to manage batch operations, 641 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 642 * to NULL to begin the batched operation. After each subsequent 643 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 644 * *out_batch* as the *in_batch* for the next operation to 645 * continue iteration from the current point. Both *in_batch* and 646 * *out_batch* must point to memory large enough to hold a key, 647 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH, 648 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters 649 * must be at least 4 bytes wide regardless of key size. 650 * 651 * The *keys* and *values* are output parameters which must point 652 * to memory large enough to hold *count* items based on the key 653 * and value size of the map *map_fd*. The *keys* buffer must be 654 * of *key_size* * *count*. The *values* buffer must be of 655 * *value_size* * *count*. 656 * 657 * The *elem_flags* argument may be specified as one of the 658 * following: 659 * 660 * **BPF_F_LOCK** 661 * Look up the value of a spin-locked map without 662 * returning the lock. This must be specified if the 663 * elements contain a spinlock. 664 * 665 * On success, *count* elements from the map are copied into the 666 * user buffer, with the keys copied into *keys* and the values 667 * copied into the corresponding indices in *values*. 668 * 669 * If an error is returned and *errno* is not **EFAULT**, *count* 670 * is set to the number of successfully processed elements. 671 * 672 * Return 673 * Returns zero on success. On error, -1 is returned and *errno* 674 * is set appropriately. 675 * 676 * May set *errno* to **ENOSPC** to indicate that *keys* or 677 * *values* is too small to dump an entire bucket during 678 * iteration of a hash-based map type. 679 * 680 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 681 * Description 682 * Iterate and delete all elements in a map. 683 * 684 * This operation has the same behavior as 685 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 686 * 687 * * Every element that is successfully returned is also deleted 688 * from the map. This is at least *count* elements. Note that 689 * *count* is both an input and an output parameter. 690 * * Upon returning with *errno* set to **EFAULT**, up to 691 * *count* elements may be deleted without returning the keys 692 * and values of the deleted elements. 693 * 694 * Return 695 * Returns zero on success. On error, -1 is returned and *errno* 696 * is set appropriately. 697 * 698 * BPF_MAP_UPDATE_BATCH 699 * Description 700 * Update multiple elements in a map by *key*. 701 * 702 * The *keys* and *values* are input parameters which must point 703 * to memory large enough to hold *count* items based on the key 704 * and value size of the map *map_fd*. The *keys* buffer must be 705 * of *key_size* * *count*. The *values* buffer must be of 706 * *value_size* * *count*. 707 * 708 * Each element specified in *keys* is sequentially updated to the 709 * value in the corresponding index in *values*. The *in_batch* 710 * and *out_batch* parameters are ignored and should be zeroed. 711 * 712 * The *elem_flags* argument should be specified as one of the 713 * following: 714 * 715 * **BPF_ANY** 716 * Create new elements or update a existing elements. 717 * **BPF_NOEXIST** 718 * Create new elements only if they do not exist. 719 * **BPF_EXIST** 720 * Update existing elements. 721 * **BPF_F_LOCK** 722 * Update spin_lock-ed map elements. This must be 723 * specified if the map value contains a spinlock. 724 * 725 * On success, *count* elements from the map are updated. 726 * 727 * If an error is returned and *errno* is not **EFAULT**, *count* 728 * is set to the number of successfully processed elements. 729 * 730 * Return 731 * Returns zero on success. On error, -1 is returned and *errno* 732 * is set appropriately. 733 * 734 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 735 * **E2BIG**. **E2BIG** indicates that the number of elements in 736 * the map reached the *max_entries* limit specified at map 737 * creation time. 738 * 739 * May set *errno* to one of the following error codes under 740 * specific circumstances: 741 * 742 * **EEXIST** 743 * If *flags* specifies **BPF_NOEXIST** and the element 744 * with *key* already exists in the map. 745 * **ENOENT** 746 * If *flags* specifies **BPF_EXIST** and the element with 747 * *key* does not exist in the map. 748 * 749 * BPF_MAP_DELETE_BATCH 750 * Description 751 * Delete multiple elements in a map by *key*. 752 * 753 * The *keys* parameter is an input parameter which must point 754 * to memory large enough to hold *count* items based on the key 755 * size of the map *map_fd*, that is, *key_size* * *count*. 756 * 757 * Each element specified in *keys* is sequentially deleted. The 758 * *in_batch*, *out_batch*, and *values* parameters are ignored 759 * and should be zeroed. 760 * 761 * The *elem_flags* argument may be specified as one of the 762 * following: 763 * 764 * **BPF_F_LOCK** 765 * Look up the value of a spin-locked map without 766 * returning the lock. This must be specified if the 767 * elements contain a spinlock. 768 * 769 * On success, *count* elements from the map are updated. 770 * 771 * If an error is returned and *errno* is not **EFAULT**, *count* 772 * is set to the number of successfully processed elements. If 773 * *errno* is **EFAULT**, up to *count* elements may be been 774 * deleted. 775 * 776 * Return 777 * Returns zero on success. On error, -1 is returned and *errno* 778 * is set appropriately. 779 * 780 * BPF_LINK_CREATE 781 * Description 782 * Attach an eBPF program to a *target_fd* at the specified 783 * *attach_type* hook and return a file descriptor handle for 784 * managing the link. 785 * 786 * Return 787 * A new file descriptor (a nonnegative integer), or -1 if an 788 * error occurred (in which case, *errno* is set appropriately). 789 * 790 * BPF_LINK_UPDATE 791 * Description 792 * Update the eBPF program in the specified *link_fd* to 793 * *new_prog_fd*. 794 * 795 * Return 796 * Returns zero on success. On error, -1 is returned and *errno* 797 * is set appropriately. 798 * 799 * BPF_LINK_GET_FD_BY_ID 800 * Description 801 * Open a file descriptor for the eBPF Link corresponding to 802 * *link_id*. 803 * 804 * Return 805 * A new file descriptor (a nonnegative integer), or -1 if an 806 * error occurred (in which case, *errno* is set appropriately). 807 * 808 * BPF_LINK_GET_NEXT_ID 809 * Description 810 * Fetch the next eBPF link currently loaded into the kernel. 811 * 812 * Looks for the eBPF link with an id greater than *start_id* 813 * and updates *next_id* on success. If no other eBPF links 814 * remain with ids higher than *start_id*, returns -1 and sets 815 * *errno* to **ENOENT**. 816 * 817 * Return 818 * Returns zero on success. On error, or when no id remains, -1 819 * is returned and *errno* is set appropriately. 820 * 821 * BPF_ENABLE_STATS 822 * Description 823 * Enable eBPF runtime statistics gathering. 824 * 825 * Runtime statistics gathering for the eBPF runtime is disabled 826 * by default to minimize the corresponding performance overhead. 827 * This command enables statistics globally. 828 * 829 * Multiple programs may independently enable statistics. 830 * After gathering the desired statistics, eBPF runtime statistics 831 * may be disabled again by calling **close**\ (2) for the file 832 * descriptor returned by this function. Statistics will only be 833 * disabled system-wide when all outstanding file descriptors 834 * returned by prior calls for this subcommand are closed. 835 * 836 * Return 837 * A new file descriptor (a nonnegative integer), or -1 if an 838 * error occurred (in which case, *errno* is set appropriately). 839 * 840 * BPF_ITER_CREATE 841 * Description 842 * Create an iterator on top of the specified *link_fd* (as 843 * previously created using **BPF_LINK_CREATE**) and return a 844 * file descriptor that can be used to trigger the iteration. 845 * 846 * If the resulting file descriptor is pinned to the filesystem 847 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 848 * for that path will trigger the iterator to read kernel state 849 * using the eBPF program attached to *link_fd*. 850 * 851 * Return 852 * A new file descriptor (a nonnegative integer), or -1 if an 853 * error occurred (in which case, *errno* is set appropriately). 854 * 855 * BPF_LINK_DETACH 856 * Description 857 * Forcefully detach the specified *link_fd* from its 858 * corresponding attachment point. 859 * 860 * Return 861 * Returns zero on success. On error, -1 is returned and *errno* 862 * is set appropriately. 863 * 864 * BPF_PROG_BIND_MAP 865 * Description 866 * Bind a map to the lifetime of an eBPF program. 867 * 868 * The map identified by *map_fd* is bound to the program 869 * identified by *prog_fd* and only released when *prog_fd* is 870 * released. This may be used in cases where metadata should be 871 * associated with a program which otherwise does not contain any 872 * references to the map (for example, embedded in the eBPF 873 * program instructions). 874 * 875 * Return 876 * Returns zero on success. On error, -1 is returned and *errno* 877 * is set appropriately. 878 * 879 * BPF_TOKEN_CREATE 880 * Description 881 * Create BPF token with embedded information about what 882 * BPF-related functionality it allows: 883 * - a set of allowed bpf() syscall commands; 884 * - a set of allowed BPF map types to be created with 885 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 886 * - a set of allowed BPF program types and BPF program attach 887 * types to be loaded with BPF_PROG_LOAD command, if 888 * BPF_PROG_LOAD itself is allowed. 889 * 890 * BPF token is created (derived) from an instance of BPF FS, 891 * assuming it has necessary delegation mount options specified. 892 * This BPF token can be passed as an extra parameter to various 893 * bpf() syscall commands to grant BPF subsystem functionality to 894 * unprivileged processes. 895 * 896 * When created, BPF token is "associated" with the owning 897 * user namespace of BPF FS instance (super block) that it was 898 * derived from, and subsequent BPF operations performed with 899 * BPF token would be performing capabilities checks (i.e., 900 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 901 * that user namespace. Without BPF token, such capabilities 902 * have to be granted in init user namespace, making bpf() 903 * syscall incompatible with user namespace, for the most part. 904 * 905 * Return 906 * A new file descriptor (a nonnegative integer), or -1 if an 907 * error occurred (in which case, *errno* is set appropriately). 908 * 909 * NOTES 910 * eBPF objects (maps and programs) can be shared between processes. 911 * 912 * * After **fork**\ (2), the child inherits file descriptors 913 * referring to the same eBPF objects. 914 * * File descriptors referring to eBPF objects can be transferred over 915 * **unix**\ (7) domain sockets. 916 * * File descriptors referring to eBPF objects can be duplicated in the 917 * usual way, using **dup**\ (2) and similar calls. 918 * * File descriptors referring to eBPF objects can be pinned to the 919 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 920 * 921 * An eBPF object is deallocated only after all file descriptors referring 922 * to the object have been closed and no references remain pinned to the 923 * filesystem or attached (for example, bound to a program or device). 924 */ 925 enum bpf_cmd { 926 BPF_MAP_CREATE, 927 BPF_MAP_LOOKUP_ELEM, 928 BPF_MAP_UPDATE_ELEM, 929 BPF_MAP_DELETE_ELEM, 930 BPF_MAP_GET_NEXT_KEY, 931 BPF_PROG_LOAD, 932 BPF_OBJ_PIN, 933 BPF_OBJ_GET, 934 BPF_PROG_ATTACH, 935 BPF_PROG_DETACH, 936 BPF_PROG_TEST_RUN, 937 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 938 BPF_PROG_GET_NEXT_ID, 939 BPF_MAP_GET_NEXT_ID, 940 BPF_PROG_GET_FD_BY_ID, 941 BPF_MAP_GET_FD_BY_ID, 942 BPF_OBJ_GET_INFO_BY_FD, 943 BPF_PROG_QUERY, 944 BPF_RAW_TRACEPOINT_OPEN, 945 BPF_BTF_LOAD, 946 BPF_BTF_GET_FD_BY_ID, 947 BPF_TASK_FD_QUERY, 948 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 949 BPF_MAP_FREEZE, 950 BPF_BTF_GET_NEXT_ID, 951 BPF_MAP_LOOKUP_BATCH, 952 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 953 BPF_MAP_UPDATE_BATCH, 954 BPF_MAP_DELETE_BATCH, 955 BPF_LINK_CREATE, 956 BPF_LINK_UPDATE, 957 BPF_LINK_GET_FD_BY_ID, 958 BPF_LINK_GET_NEXT_ID, 959 BPF_ENABLE_STATS, 960 BPF_ITER_CREATE, 961 BPF_LINK_DETACH, 962 BPF_PROG_BIND_MAP, 963 BPF_TOKEN_CREATE, 964 __MAX_BPF_CMD, 965 }; 966 967 enum bpf_map_type { 968 BPF_MAP_TYPE_UNSPEC, 969 BPF_MAP_TYPE_HASH, 970 BPF_MAP_TYPE_ARRAY, 971 BPF_MAP_TYPE_PROG_ARRAY, 972 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 973 BPF_MAP_TYPE_PERCPU_HASH, 974 BPF_MAP_TYPE_PERCPU_ARRAY, 975 BPF_MAP_TYPE_STACK_TRACE, 976 BPF_MAP_TYPE_CGROUP_ARRAY, 977 BPF_MAP_TYPE_LRU_HASH, 978 BPF_MAP_TYPE_LRU_PERCPU_HASH, 979 BPF_MAP_TYPE_LPM_TRIE, 980 BPF_MAP_TYPE_ARRAY_OF_MAPS, 981 BPF_MAP_TYPE_HASH_OF_MAPS, 982 BPF_MAP_TYPE_DEVMAP, 983 BPF_MAP_TYPE_SOCKMAP, 984 BPF_MAP_TYPE_CPUMAP, 985 BPF_MAP_TYPE_XSKMAP, 986 BPF_MAP_TYPE_SOCKHASH, 987 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 988 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 989 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 990 * both cgroup-attached and other progs and supports all functionality 991 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 992 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 993 */ 994 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 995 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 996 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 997 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 998 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 999 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 1000 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 1001 * deprecated. 1002 */ 1003 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 1004 BPF_MAP_TYPE_QUEUE, 1005 BPF_MAP_TYPE_STACK, 1006 BPF_MAP_TYPE_SK_STORAGE, 1007 BPF_MAP_TYPE_DEVMAP_HASH, 1008 BPF_MAP_TYPE_STRUCT_OPS, 1009 BPF_MAP_TYPE_RINGBUF, 1010 BPF_MAP_TYPE_INODE_STORAGE, 1011 BPF_MAP_TYPE_TASK_STORAGE, 1012 BPF_MAP_TYPE_BLOOM_FILTER, 1013 BPF_MAP_TYPE_USER_RINGBUF, 1014 BPF_MAP_TYPE_CGRP_STORAGE, 1015 BPF_MAP_TYPE_ARENA, 1016 __MAX_BPF_MAP_TYPE 1017 }; 1018 1019 /* Note that tracing related programs such as 1020 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1021 * are not subject to a stable API since kernel internal data 1022 * structures can change from release to release and may 1023 * therefore break existing tracing BPF programs. Tracing BPF 1024 * programs correspond to /a/ specific kernel which is to be 1025 * analyzed, and not /a/ specific kernel /and/ all future ones. 1026 */ 1027 enum bpf_prog_type { 1028 BPF_PROG_TYPE_UNSPEC, 1029 BPF_PROG_TYPE_SOCKET_FILTER, 1030 BPF_PROG_TYPE_KPROBE, 1031 BPF_PROG_TYPE_SCHED_CLS, 1032 BPF_PROG_TYPE_SCHED_ACT, 1033 BPF_PROG_TYPE_TRACEPOINT, 1034 BPF_PROG_TYPE_XDP, 1035 BPF_PROG_TYPE_PERF_EVENT, 1036 BPF_PROG_TYPE_CGROUP_SKB, 1037 BPF_PROG_TYPE_CGROUP_SOCK, 1038 BPF_PROG_TYPE_LWT_IN, 1039 BPF_PROG_TYPE_LWT_OUT, 1040 BPF_PROG_TYPE_LWT_XMIT, 1041 BPF_PROG_TYPE_SOCK_OPS, 1042 BPF_PROG_TYPE_SK_SKB, 1043 BPF_PROG_TYPE_CGROUP_DEVICE, 1044 BPF_PROG_TYPE_SK_MSG, 1045 BPF_PROG_TYPE_RAW_TRACEPOINT, 1046 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1047 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1048 BPF_PROG_TYPE_LIRC_MODE2, 1049 BPF_PROG_TYPE_SK_REUSEPORT, 1050 BPF_PROG_TYPE_FLOW_DISSECTOR, 1051 BPF_PROG_TYPE_CGROUP_SYSCTL, 1052 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1053 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1054 BPF_PROG_TYPE_TRACING, 1055 BPF_PROG_TYPE_STRUCT_OPS, 1056 BPF_PROG_TYPE_EXT, 1057 BPF_PROG_TYPE_LSM, 1058 BPF_PROG_TYPE_SK_LOOKUP, 1059 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1060 BPF_PROG_TYPE_NETFILTER, 1061 __MAX_BPF_PROG_TYPE 1062 }; 1063 1064 enum bpf_attach_type { 1065 BPF_CGROUP_INET_INGRESS, 1066 BPF_CGROUP_INET_EGRESS, 1067 BPF_CGROUP_INET_SOCK_CREATE, 1068 BPF_CGROUP_SOCK_OPS, 1069 BPF_SK_SKB_STREAM_PARSER, 1070 BPF_SK_SKB_STREAM_VERDICT, 1071 BPF_CGROUP_DEVICE, 1072 BPF_SK_MSG_VERDICT, 1073 BPF_CGROUP_INET4_BIND, 1074 BPF_CGROUP_INET6_BIND, 1075 BPF_CGROUP_INET4_CONNECT, 1076 BPF_CGROUP_INET6_CONNECT, 1077 BPF_CGROUP_INET4_POST_BIND, 1078 BPF_CGROUP_INET6_POST_BIND, 1079 BPF_CGROUP_UDP4_SENDMSG, 1080 BPF_CGROUP_UDP6_SENDMSG, 1081 BPF_LIRC_MODE2, 1082 BPF_FLOW_DISSECTOR, 1083 BPF_CGROUP_SYSCTL, 1084 BPF_CGROUP_UDP4_RECVMSG, 1085 BPF_CGROUP_UDP6_RECVMSG, 1086 BPF_CGROUP_GETSOCKOPT, 1087 BPF_CGROUP_SETSOCKOPT, 1088 BPF_TRACE_RAW_TP, 1089 BPF_TRACE_FENTRY, 1090 BPF_TRACE_FEXIT, 1091 BPF_MODIFY_RETURN, 1092 BPF_LSM_MAC, 1093 BPF_TRACE_ITER, 1094 BPF_CGROUP_INET4_GETPEERNAME, 1095 BPF_CGROUP_INET6_GETPEERNAME, 1096 BPF_CGROUP_INET4_GETSOCKNAME, 1097 BPF_CGROUP_INET6_GETSOCKNAME, 1098 BPF_XDP_DEVMAP, 1099 BPF_CGROUP_INET_SOCK_RELEASE, 1100 BPF_XDP_CPUMAP, 1101 BPF_SK_LOOKUP, 1102 BPF_XDP, 1103 BPF_SK_SKB_VERDICT, 1104 BPF_SK_REUSEPORT_SELECT, 1105 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1106 BPF_PERF_EVENT, 1107 BPF_TRACE_KPROBE_MULTI, 1108 BPF_LSM_CGROUP, 1109 BPF_STRUCT_OPS, 1110 BPF_NETFILTER, 1111 BPF_TCX_INGRESS, 1112 BPF_TCX_EGRESS, 1113 BPF_TRACE_UPROBE_MULTI, 1114 BPF_CGROUP_UNIX_CONNECT, 1115 BPF_CGROUP_UNIX_SENDMSG, 1116 BPF_CGROUP_UNIX_RECVMSG, 1117 BPF_CGROUP_UNIX_GETPEERNAME, 1118 BPF_CGROUP_UNIX_GETSOCKNAME, 1119 BPF_NETKIT_PRIMARY, 1120 BPF_NETKIT_PEER, 1121 BPF_TRACE_KPROBE_SESSION, 1122 BPF_TRACE_UPROBE_SESSION, 1123 __MAX_BPF_ATTACH_TYPE 1124 }; 1125 1126 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1127 1128 /* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[] 1129 * in sync with the definitions below. 1130 */ 1131 enum bpf_link_type { 1132 BPF_LINK_TYPE_UNSPEC = 0, 1133 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1134 BPF_LINK_TYPE_TRACING = 2, 1135 BPF_LINK_TYPE_CGROUP = 3, 1136 BPF_LINK_TYPE_ITER = 4, 1137 BPF_LINK_TYPE_NETNS = 5, 1138 BPF_LINK_TYPE_XDP = 6, 1139 BPF_LINK_TYPE_PERF_EVENT = 7, 1140 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1141 BPF_LINK_TYPE_STRUCT_OPS = 9, 1142 BPF_LINK_TYPE_NETFILTER = 10, 1143 BPF_LINK_TYPE_TCX = 11, 1144 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1145 BPF_LINK_TYPE_NETKIT = 13, 1146 BPF_LINK_TYPE_SOCKMAP = 14, 1147 __MAX_BPF_LINK_TYPE, 1148 }; 1149 1150 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1151 1152 enum bpf_perf_event_type { 1153 BPF_PERF_EVENT_UNSPEC = 0, 1154 BPF_PERF_EVENT_UPROBE = 1, 1155 BPF_PERF_EVENT_URETPROBE = 2, 1156 BPF_PERF_EVENT_KPROBE = 3, 1157 BPF_PERF_EVENT_KRETPROBE = 4, 1158 BPF_PERF_EVENT_TRACEPOINT = 5, 1159 BPF_PERF_EVENT_EVENT = 6, 1160 }; 1161 1162 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1163 * 1164 * NONE(default): No further bpf programs allowed in the subtree. 1165 * 1166 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1167 * the program in this cgroup yields to sub-cgroup program. 1168 * 1169 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1170 * that cgroup program gets run in addition to the program in this cgroup. 1171 * 1172 * Only one program is allowed to be attached to a cgroup with 1173 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1174 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1175 * release old program and attach the new one. Attach flags has to match. 1176 * 1177 * Multiple programs are allowed to be attached to a cgroup with 1178 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1179 * (those that were attached first, run first) 1180 * The programs of sub-cgroup are executed first, then programs of 1181 * this cgroup and then programs of parent cgroup. 1182 * When children program makes decision (like picking TCP CA or sock bind) 1183 * parent program has a chance to override it. 1184 * 1185 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1186 * programs for a cgroup. Though it's possible to replace an old program at 1187 * any position by also specifying BPF_F_REPLACE flag and position itself in 1188 * replace_bpf_fd attribute. Old program at this position will be released. 1189 * 1190 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1191 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1192 * Ex1: 1193 * cgrp1 (MULTI progs A, B) -> 1194 * cgrp2 (OVERRIDE prog C) -> 1195 * cgrp3 (MULTI prog D) -> 1196 * cgrp4 (OVERRIDE prog E) -> 1197 * cgrp5 (NONE prog F) 1198 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1199 * if prog F is detached, the execution is E,D,A,B 1200 * if prog F and D are detached, the execution is E,A,B 1201 * if prog F, E and D are detached, the execution is C,A,B 1202 * 1203 * All eligible programs are executed regardless of return code from 1204 * earlier programs. 1205 */ 1206 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1207 #define BPF_F_ALLOW_MULTI (1U << 1) 1208 /* Generic attachment flags. */ 1209 #define BPF_F_REPLACE (1U << 2) 1210 #define BPF_F_BEFORE (1U << 3) 1211 #define BPF_F_AFTER (1U << 4) 1212 #define BPF_F_ID (1U << 5) 1213 #define BPF_F_PREORDER (1U << 6) 1214 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1215 1216 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1217 * verifier will perform strict alignment checking as if the kernel 1218 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1219 * and NET_IP_ALIGN defined to 2. 1220 */ 1221 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1222 1223 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1224 * verifier will allow any alignment whatsoever. On platforms 1225 * with strict alignment requirements for loads ands stores (such 1226 * as sparc and mips) the verifier validates that all loads and 1227 * stores provably follow this requirement. This flag turns that 1228 * checking and enforcement off. 1229 * 1230 * It is mostly used for testing when we want to validate the 1231 * context and memory access aspects of the verifier, but because 1232 * of an unaligned access the alignment check would trigger before 1233 * the one we are interested in. 1234 */ 1235 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1236 1237 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1238 * Verifier does sub-register def/use analysis and identifies instructions whose 1239 * def only matters for low 32-bit, high 32-bit is never referenced later 1240 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1241 * that it is safe to ignore clearing high 32-bit for these instructions. This 1242 * saves some back-ends a lot of code-gen. However such optimization is not 1243 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1244 * hence hasn't used verifier's analysis result. But, we really want to have a 1245 * way to be able to verify the correctness of the described optimization on 1246 * x86_64 on which testsuites are frequently exercised. 1247 * 1248 * So, this flag is introduced. Once it is set, verifier will randomize high 1249 * 32-bit for those instructions who has been identified as safe to ignore them. 1250 * Then, if verifier is not doing correct analysis, such randomization will 1251 * regress tests to expose bugs. 1252 */ 1253 #define BPF_F_TEST_RND_HI32 (1U << 2) 1254 1255 /* The verifier internal test flag. Behavior is undefined */ 1256 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1257 1258 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1259 * restrict map and helper usage for such programs. Sleepable BPF programs can 1260 * only be attached to hooks where kernel execution context allows sleeping. 1261 * Such programs are allowed to use helpers that may sleep like 1262 * bpf_copy_from_user(). 1263 */ 1264 #define BPF_F_SLEEPABLE (1U << 4) 1265 1266 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1267 * fully support xdp frags. 1268 */ 1269 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1270 1271 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1272 * program becomes device-bound but can access XDP metadata. 1273 */ 1274 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1275 1276 /* The verifier internal test flag. Behavior is undefined */ 1277 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1278 1279 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1280 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1281 */ 1282 enum { 1283 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1284 }; 1285 1286 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1287 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1288 */ 1289 enum { 1290 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1291 }; 1292 1293 /* link_create.netfilter.flags used in LINK_CREATE command for 1294 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1295 */ 1296 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1297 1298 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1299 * the following extensions: 1300 * 1301 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1302 * insn[0].imm: map fd or fd_idx 1303 * insn[1].imm: 0 1304 * insn[0].off: 0 1305 * insn[1].off: 0 1306 * ldimm64 rewrite: address of map 1307 * verifier type: CONST_PTR_TO_MAP 1308 */ 1309 #define BPF_PSEUDO_MAP_FD 1 1310 #define BPF_PSEUDO_MAP_IDX 5 1311 1312 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1313 * insn[0].imm: map fd or fd_idx 1314 * insn[1].imm: offset into value 1315 * insn[0].off: 0 1316 * insn[1].off: 0 1317 * ldimm64 rewrite: address of map[0]+offset 1318 * verifier type: PTR_TO_MAP_VALUE 1319 */ 1320 #define BPF_PSEUDO_MAP_VALUE 2 1321 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1322 1323 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1324 * insn[0].imm: kernel btd id of VAR 1325 * insn[1].imm: 0 1326 * insn[0].off: 0 1327 * insn[1].off: 0 1328 * ldimm64 rewrite: address of the kernel variable 1329 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1330 * is struct/union. 1331 */ 1332 #define BPF_PSEUDO_BTF_ID 3 1333 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1334 * insn[0].imm: insn offset to the func 1335 * insn[1].imm: 0 1336 * insn[0].off: 0 1337 * insn[1].off: 0 1338 * ldimm64 rewrite: address of the function 1339 * verifier type: PTR_TO_FUNC. 1340 */ 1341 #define BPF_PSEUDO_FUNC 4 1342 1343 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1344 * offset to another bpf function 1345 */ 1346 #define BPF_PSEUDO_CALL 1 1347 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1348 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1349 */ 1350 #define BPF_PSEUDO_KFUNC_CALL 2 1351 1352 enum bpf_addr_space_cast { 1353 BPF_ADDR_SPACE_CAST = 1, 1354 }; 1355 1356 /* flags for BPF_MAP_UPDATE_ELEM command */ 1357 enum { 1358 BPF_ANY = 0, /* create new element or update existing */ 1359 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1360 BPF_EXIST = 2, /* update existing element */ 1361 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1362 }; 1363 1364 /* flags for BPF_MAP_CREATE command */ 1365 enum { 1366 BPF_F_NO_PREALLOC = (1U << 0), 1367 /* Instead of having one common LRU list in the 1368 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1369 * which can scale and perform better. 1370 * Note, the LRU nodes (including free nodes) cannot be moved 1371 * across different LRU lists. 1372 */ 1373 BPF_F_NO_COMMON_LRU = (1U << 1), 1374 /* Specify numa node during map creation */ 1375 BPF_F_NUMA_NODE = (1U << 2), 1376 1377 /* Flags for accessing BPF object from syscall side. */ 1378 BPF_F_RDONLY = (1U << 3), 1379 BPF_F_WRONLY = (1U << 4), 1380 1381 /* Flag for stack_map, store build_id+offset instead of pointer */ 1382 BPF_F_STACK_BUILD_ID = (1U << 5), 1383 1384 /* Zero-initialize hash function seed. This should only be used for testing. */ 1385 BPF_F_ZERO_SEED = (1U << 6), 1386 1387 /* Flags for accessing BPF object from program side. */ 1388 BPF_F_RDONLY_PROG = (1U << 7), 1389 BPF_F_WRONLY_PROG = (1U << 8), 1390 1391 /* Clone map from listener for newly accepted socket */ 1392 BPF_F_CLONE = (1U << 9), 1393 1394 /* Enable memory-mapping BPF map */ 1395 BPF_F_MMAPABLE = (1U << 10), 1396 1397 /* Share perf_event among processes */ 1398 BPF_F_PRESERVE_ELEMS = (1U << 11), 1399 1400 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1401 BPF_F_INNER_MAP = (1U << 12), 1402 1403 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1404 BPF_F_LINK = (1U << 13), 1405 1406 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1407 BPF_F_PATH_FD = (1U << 14), 1408 1409 /* Flag for value_type_btf_obj_fd, the fd is available */ 1410 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1411 1412 /* BPF token FD is passed in a corresponding command's token_fd field */ 1413 BPF_F_TOKEN_FD = (1U << 16), 1414 1415 /* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */ 1416 BPF_F_SEGV_ON_FAULT = (1U << 17), 1417 1418 /* Do not translate kernel bpf_arena pointers to user pointers */ 1419 BPF_F_NO_USER_CONV = (1U << 18), 1420 }; 1421 1422 /* Flags for BPF_PROG_QUERY. */ 1423 1424 /* Query effective (directly attached + inherited from ancestor cgroups) 1425 * programs that will be executed for events within a cgroup. 1426 * attach_flags with this flag are always returned 0. 1427 */ 1428 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1429 1430 /* Flags for BPF_PROG_TEST_RUN */ 1431 1432 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1433 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1434 /* If set, XDP frames will be transmitted after processing */ 1435 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1436 /* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */ 1437 #define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2) 1438 1439 /* type for BPF_ENABLE_STATS */ 1440 enum bpf_stats_type { 1441 /* enabled run_time_ns and run_cnt */ 1442 BPF_STATS_RUN_TIME = 0, 1443 }; 1444 1445 enum bpf_stack_build_id_status { 1446 /* user space need an empty entry to identify end of a trace */ 1447 BPF_STACK_BUILD_ID_EMPTY = 0, 1448 /* with valid build_id and offset */ 1449 BPF_STACK_BUILD_ID_VALID = 1, 1450 /* couldn't get build_id, fallback to ip */ 1451 BPF_STACK_BUILD_ID_IP = 2, 1452 }; 1453 1454 #define BPF_BUILD_ID_SIZE 20 1455 struct bpf_stack_build_id { 1456 __s32 status; 1457 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1458 union { 1459 __u64 offset; 1460 __u64 ip; 1461 }; 1462 }; 1463 1464 #define BPF_OBJ_NAME_LEN 16U 1465 1466 union bpf_attr { 1467 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1468 __u32 map_type; /* one of enum bpf_map_type */ 1469 __u32 key_size; /* size of key in bytes */ 1470 __u32 value_size; /* size of value in bytes */ 1471 __u32 max_entries; /* max number of entries in a map */ 1472 __u32 map_flags; /* BPF_MAP_CREATE related 1473 * flags defined above. 1474 */ 1475 __u32 inner_map_fd; /* fd pointing to the inner map */ 1476 __u32 numa_node; /* numa node (effective only if 1477 * BPF_F_NUMA_NODE is set). 1478 */ 1479 char map_name[BPF_OBJ_NAME_LEN]; 1480 __u32 map_ifindex; /* ifindex of netdev to create on */ 1481 __u32 btf_fd; /* fd pointing to a BTF type data */ 1482 __u32 btf_key_type_id; /* BTF type_id of the key */ 1483 __u32 btf_value_type_id; /* BTF type_id of the value */ 1484 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1485 * struct stored as the 1486 * map value 1487 */ 1488 /* Any per-map-type extra fields 1489 * 1490 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1491 * number of hash functions (if 0, the bloom filter will default 1492 * to using 5 hash functions). 1493 * 1494 * BPF_MAP_TYPE_ARENA - contains the address where user space 1495 * is going to mmap() the arena. It has to be page aligned. 1496 */ 1497 __u64 map_extra; 1498 1499 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1500 * type data for 1501 * btf_vmlinux_value_type_id. 1502 */ 1503 /* BPF token FD to use with BPF_MAP_CREATE operation. 1504 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1505 */ 1506 __s32 map_token_fd; 1507 }; 1508 1509 struct { /* anonymous struct used by BPF_MAP_*_ELEM and BPF_MAP_FREEZE commands */ 1510 __u32 map_fd; 1511 __aligned_u64 key; 1512 union { 1513 __aligned_u64 value; 1514 __aligned_u64 next_key; 1515 }; 1516 __u64 flags; 1517 }; 1518 1519 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1520 __aligned_u64 in_batch; /* start batch, 1521 * NULL to start from beginning 1522 */ 1523 __aligned_u64 out_batch; /* output: next start batch */ 1524 __aligned_u64 keys; 1525 __aligned_u64 values; 1526 __u32 count; /* input/output: 1527 * input: # of key/value 1528 * elements 1529 * output: # of filled elements 1530 */ 1531 __u32 map_fd; 1532 __u64 elem_flags; 1533 __u64 flags; 1534 } batch; 1535 1536 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1537 __u32 prog_type; /* one of enum bpf_prog_type */ 1538 __u32 insn_cnt; 1539 __aligned_u64 insns; 1540 __aligned_u64 license; 1541 __u32 log_level; /* verbosity level of verifier */ 1542 __u32 log_size; /* size of user buffer */ 1543 __aligned_u64 log_buf; /* user supplied buffer */ 1544 __u32 kern_version; /* not used */ 1545 __u32 prog_flags; 1546 char prog_name[BPF_OBJ_NAME_LEN]; 1547 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1548 /* For some prog types expected attach type must be known at 1549 * load time to verify attach type specific parts of prog 1550 * (context accesses, allowed helpers, etc). 1551 */ 1552 __u32 expected_attach_type; 1553 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1554 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1555 __aligned_u64 func_info; /* func info */ 1556 __u32 func_info_cnt; /* number of bpf_func_info records */ 1557 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1558 __aligned_u64 line_info; /* line info */ 1559 __u32 line_info_cnt; /* number of bpf_line_info records */ 1560 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1561 union { 1562 /* valid prog_fd to attach to bpf prog */ 1563 __u32 attach_prog_fd; 1564 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1565 __u32 attach_btf_obj_fd; 1566 }; 1567 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1568 __aligned_u64 fd_array; /* array of FDs */ 1569 __aligned_u64 core_relos; 1570 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1571 /* output: actual total log contents size (including termintaing zero). 1572 * It could be both larger than original log_size (if log was 1573 * truncated), or smaller (if log buffer wasn't filled completely). 1574 */ 1575 __u32 log_true_size; 1576 /* BPF token FD to use with BPF_PROG_LOAD operation. 1577 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1578 */ 1579 __s32 prog_token_fd; 1580 /* The fd_array_cnt can be used to pass the length of the 1581 * fd_array array. In this case all the [map] file descriptors 1582 * passed in this array will be bound to the program, even if 1583 * the maps are not referenced directly. The functionality is 1584 * similar to the BPF_PROG_BIND_MAP syscall, but maps can be 1585 * used by the verifier during the program load. If provided, 1586 * then the fd_array[0,...,fd_array_cnt-1] is expected to be 1587 * continuous. 1588 */ 1589 __u32 fd_array_cnt; 1590 }; 1591 1592 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1593 __aligned_u64 pathname; 1594 __u32 bpf_fd; 1595 __u32 file_flags; 1596 /* Same as dirfd in openat() syscall; see openat(2) 1597 * manpage for details of path FD and pathname semantics; 1598 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1599 * file_flags field, otherwise it should be set to zero; 1600 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1601 */ 1602 __s32 path_fd; 1603 }; 1604 1605 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1606 union { 1607 __u32 target_fd; /* target object to attach to or ... */ 1608 __u32 target_ifindex; /* target ifindex */ 1609 }; 1610 __u32 attach_bpf_fd; 1611 __u32 attach_type; 1612 __u32 attach_flags; 1613 __u32 replace_bpf_fd; 1614 union { 1615 __u32 relative_fd; 1616 __u32 relative_id; 1617 }; 1618 __u64 expected_revision; 1619 }; 1620 1621 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1622 __u32 prog_fd; 1623 __u32 retval; 1624 __u32 data_size_in; /* input: len of data_in */ 1625 __u32 data_size_out; /* input/output: len of data_out 1626 * returns ENOSPC if data_out 1627 * is too small. 1628 */ 1629 __aligned_u64 data_in; 1630 __aligned_u64 data_out; 1631 __u32 repeat; 1632 __u32 duration; 1633 __u32 ctx_size_in; /* input: len of ctx_in */ 1634 __u32 ctx_size_out; /* input/output: len of ctx_out 1635 * returns ENOSPC if ctx_out 1636 * is too small. 1637 */ 1638 __aligned_u64 ctx_in; 1639 __aligned_u64 ctx_out; 1640 __u32 flags; 1641 __u32 cpu; 1642 __u32 batch_size; 1643 } test; 1644 1645 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1646 union { 1647 __u32 start_id; 1648 __u32 prog_id; 1649 __u32 map_id; 1650 __u32 btf_id; 1651 __u32 link_id; 1652 }; 1653 __u32 next_id; 1654 __u32 open_flags; 1655 __s32 fd_by_id_token_fd; 1656 }; 1657 1658 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1659 __u32 bpf_fd; 1660 __u32 info_len; 1661 __aligned_u64 info; 1662 } info; 1663 1664 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1665 union { 1666 __u32 target_fd; /* target object to query or ... */ 1667 __u32 target_ifindex; /* target ifindex */ 1668 }; 1669 __u32 attach_type; 1670 __u32 query_flags; 1671 __u32 attach_flags; 1672 __aligned_u64 prog_ids; 1673 union { 1674 __u32 prog_cnt; 1675 __u32 count; 1676 }; 1677 __u32 :32; 1678 /* output: per-program attach_flags. 1679 * not allowed to be set during effective query. 1680 */ 1681 __aligned_u64 prog_attach_flags; 1682 __aligned_u64 link_ids; 1683 __aligned_u64 link_attach_flags; 1684 __u64 revision; 1685 } query; 1686 1687 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1688 __u64 name; 1689 __u32 prog_fd; 1690 __u32 :32; 1691 __aligned_u64 cookie; 1692 } raw_tracepoint; 1693 1694 struct { /* anonymous struct for BPF_BTF_LOAD */ 1695 __aligned_u64 btf; 1696 __aligned_u64 btf_log_buf; 1697 __u32 btf_size; 1698 __u32 btf_log_size; 1699 __u32 btf_log_level; 1700 /* output: actual total log contents size (including termintaing zero). 1701 * It could be both larger than original log_size (if log was 1702 * truncated), or smaller (if log buffer wasn't filled completely). 1703 */ 1704 __u32 btf_log_true_size; 1705 __u32 btf_flags; 1706 /* BPF token FD to use with BPF_BTF_LOAD operation. 1707 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1708 */ 1709 __s32 btf_token_fd; 1710 }; 1711 1712 struct { 1713 __u32 pid; /* input: pid */ 1714 __u32 fd; /* input: fd */ 1715 __u32 flags; /* input: flags */ 1716 __u32 buf_len; /* input/output: buf len */ 1717 __aligned_u64 buf; /* input/output: 1718 * tp_name for tracepoint 1719 * symbol for kprobe 1720 * filename for uprobe 1721 */ 1722 __u32 prog_id; /* output: prod_id */ 1723 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1724 __u64 probe_offset; /* output: probe_offset */ 1725 __u64 probe_addr; /* output: probe_addr */ 1726 } task_fd_query; 1727 1728 struct { /* struct used by BPF_LINK_CREATE command */ 1729 union { 1730 __u32 prog_fd; /* eBPF program to attach */ 1731 __u32 map_fd; /* struct_ops to attach */ 1732 }; 1733 union { 1734 __u32 target_fd; /* target object to attach to or ... */ 1735 __u32 target_ifindex; /* target ifindex */ 1736 }; 1737 __u32 attach_type; /* attach type */ 1738 __u32 flags; /* extra flags */ 1739 union { 1740 __u32 target_btf_id; /* btf_id of target to attach to */ 1741 struct { 1742 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1743 __u32 iter_info_len; /* iter_info length */ 1744 }; 1745 struct { 1746 /* black box user-provided value passed through 1747 * to BPF program at the execution time and 1748 * accessible through bpf_get_attach_cookie() BPF helper 1749 */ 1750 __u64 bpf_cookie; 1751 } perf_event; 1752 struct { 1753 __u32 flags; 1754 __u32 cnt; 1755 __aligned_u64 syms; 1756 __aligned_u64 addrs; 1757 __aligned_u64 cookies; 1758 } kprobe_multi; 1759 struct { 1760 /* this is overlaid with the target_btf_id above. */ 1761 __u32 target_btf_id; 1762 /* black box user-provided value passed through 1763 * to BPF program at the execution time and 1764 * accessible through bpf_get_attach_cookie() BPF helper 1765 */ 1766 __u64 cookie; 1767 } tracing; 1768 struct { 1769 __u32 pf; 1770 __u32 hooknum; 1771 __s32 priority; 1772 __u32 flags; 1773 } netfilter; 1774 struct { 1775 union { 1776 __u32 relative_fd; 1777 __u32 relative_id; 1778 }; 1779 __u64 expected_revision; 1780 } tcx; 1781 struct { 1782 __aligned_u64 path; 1783 __aligned_u64 offsets; 1784 __aligned_u64 ref_ctr_offsets; 1785 __aligned_u64 cookies; 1786 __u32 cnt; 1787 __u32 flags; 1788 __u32 pid; 1789 } uprobe_multi; 1790 struct { 1791 union { 1792 __u32 relative_fd; 1793 __u32 relative_id; 1794 }; 1795 __u64 expected_revision; 1796 } netkit; 1797 }; 1798 } link_create; 1799 1800 struct { /* struct used by BPF_LINK_UPDATE command */ 1801 __u32 link_fd; /* link fd */ 1802 union { 1803 /* new program fd to update link with */ 1804 __u32 new_prog_fd; 1805 /* new struct_ops map fd to update link with */ 1806 __u32 new_map_fd; 1807 }; 1808 __u32 flags; /* extra flags */ 1809 union { 1810 /* expected link's program fd; is specified only if 1811 * BPF_F_REPLACE flag is set in flags. 1812 */ 1813 __u32 old_prog_fd; 1814 /* expected link's map fd; is specified only 1815 * if BPF_F_REPLACE flag is set. 1816 */ 1817 __u32 old_map_fd; 1818 }; 1819 } link_update; 1820 1821 struct { 1822 __u32 link_fd; 1823 } link_detach; 1824 1825 struct { /* struct used by BPF_ENABLE_STATS command */ 1826 __u32 type; 1827 } enable_stats; 1828 1829 struct { /* struct used by BPF_ITER_CREATE command */ 1830 __u32 link_fd; 1831 __u32 flags; 1832 } iter_create; 1833 1834 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1835 __u32 prog_fd; 1836 __u32 map_fd; 1837 __u32 flags; /* extra flags */ 1838 } prog_bind_map; 1839 1840 struct { /* struct used by BPF_TOKEN_CREATE command */ 1841 __u32 flags; 1842 __u32 bpffs_fd; 1843 } token_create; 1844 1845 } __attribute__((aligned(8))); 1846 1847 /* The description below is an attempt at providing documentation to eBPF 1848 * developers about the multiple available eBPF helper functions. It can be 1849 * parsed and used to produce a manual page. The workflow is the following, 1850 * and requires the rst2man utility: 1851 * 1852 * $ ./scripts/bpf_doc.py \ 1853 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1854 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1855 * $ man /tmp/bpf-helpers.7 1856 * 1857 * Note that in order to produce this external documentation, some RST 1858 * formatting is used in the descriptions to get "bold" and "italics" in 1859 * manual pages. Also note that the few trailing white spaces are 1860 * intentional, removing them would break paragraphs for rst2man. 1861 * 1862 * Start of BPF helper function descriptions: 1863 * 1864 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1865 * Description 1866 * Perform a lookup in *map* for an entry associated to *key*. 1867 * Return 1868 * Map value associated to *key*, or **NULL** if no entry was 1869 * found. 1870 * 1871 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1872 * Description 1873 * Add or update the value of the entry associated to *key* in 1874 * *map* with *value*. *flags* is one of: 1875 * 1876 * **BPF_NOEXIST** 1877 * The entry for *key* must not exist in the map. 1878 * **BPF_EXIST** 1879 * The entry for *key* must already exist in the map. 1880 * **BPF_ANY** 1881 * No condition on the existence of the entry for *key*. 1882 * 1883 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1884 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1885 * elements always exist), the helper would return an error. 1886 * Return 1887 * 0 on success, or a negative error in case of failure. 1888 * 1889 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1890 * Description 1891 * Delete entry with *key* from *map*. 1892 * Return 1893 * 0 on success, or a negative error in case of failure. 1894 * 1895 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1896 * Description 1897 * For tracing programs, safely attempt to read *size* bytes from 1898 * kernel space address *unsafe_ptr* and store the data in *dst*. 1899 * 1900 * Generally, use **bpf_probe_read_user**\ () or 1901 * **bpf_probe_read_kernel**\ () instead. 1902 * Return 1903 * 0 on success, or a negative error in case of failure. 1904 * 1905 * u64 bpf_ktime_get_ns(void) 1906 * Description 1907 * Return the time elapsed since system boot, in nanoseconds. 1908 * Does not include time the system was suspended. 1909 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1910 * Return 1911 * Current *ktime*. 1912 * 1913 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1914 * Description 1915 * This helper is a "printk()-like" facility for debugging. It 1916 * prints a message defined by format *fmt* (of size *fmt_size*) 1917 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1918 * available. It can take up to three additional **u64** 1919 * arguments (as an eBPF helpers, the total number of arguments is 1920 * limited to five). 1921 * 1922 * Each time the helper is called, it appends a line to the trace. 1923 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1924 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1925 * The format of the trace is customizable, and the exact output 1926 * one will get depends on the options set in 1927 * *\/sys/kernel/tracing/trace_options* (see also the 1928 * *README* file under the same directory). However, it usually 1929 * defaults to something like: 1930 * 1931 * :: 1932 * 1933 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1934 * 1935 * In the above: 1936 * 1937 * * ``telnet`` is the name of the current task. 1938 * * ``470`` is the PID of the current task. 1939 * * ``001`` is the CPU number on which the task is 1940 * running. 1941 * * In ``.N..``, each character refers to a set of 1942 * options (whether irqs are enabled, scheduling 1943 * options, whether hard/softirqs are running, level of 1944 * preempt_disabled respectively). **N** means that 1945 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1946 * are set. 1947 * * ``419421.045894`` is a timestamp. 1948 * * ``0x00000001`` is a fake value used by BPF for the 1949 * instruction pointer register. 1950 * * ``<formatted msg>`` is the message formatted with 1951 * *fmt*. 1952 * 1953 * The conversion specifiers supported by *fmt* are similar, but 1954 * more limited than for printk(). They are **%d**, **%i**, 1955 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1956 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1957 * of field, padding with zeroes, etc.) is available, and the 1958 * helper will return **-EINVAL** (but print nothing) if it 1959 * encounters an unknown specifier. 1960 * 1961 * Also, note that **bpf_trace_printk**\ () is slow, and should 1962 * only be used for debugging purposes. For this reason, a notice 1963 * block (spanning several lines) is printed to kernel logs and 1964 * states that the helper should not be used "for production use" 1965 * the first time this helper is used (or more precisely, when 1966 * **trace_printk**\ () buffers are allocated). For passing values 1967 * to user space, perf events should be preferred. 1968 * Return 1969 * The number of bytes written to the buffer, or a negative error 1970 * in case of failure. 1971 * 1972 * u32 bpf_get_prandom_u32(void) 1973 * Description 1974 * Get a pseudo-random number. 1975 * 1976 * From a security point of view, this helper uses its own 1977 * pseudo-random internal state, and cannot be used to infer the 1978 * seed of other random functions in the kernel. However, it is 1979 * essential to note that the generator used by the helper is not 1980 * cryptographically secure. 1981 * Return 1982 * A random 32-bit unsigned value. 1983 * 1984 * u32 bpf_get_smp_processor_id(void) 1985 * Description 1986 * Get the SMP (symmetric multiprocessing) processor id. Note that 1987 * all programs run with migration disabled, which means that the 1988 * SMP processor id is stable during all the execution of the 1989 * program. 1990 * Return 1991 * The SMP id of the processor running the program. 1992 * Attributes 1993 * __bpf_fastcall 1994 * 1995 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1996 * Description 1997 * Store *len* bytes from address *from* into the packet 1998 * associated to *skb*, at *offset*. The *flags* are a combination 1999 * of the following values: 2000 * 2001 * **BPF_F_RECOMPUTE_CSUM** 2002 * Automatically update *skb*\ **->csum** after storing the 2003 * bytes. 2004 * **BPF_F_INVALIDATE_HASH** 2005 * Set *skb*\ **->hash**, *skb*\ **->swhash** and *skb*\ 2006 * **->l4hash** to 0. 2007 * 2008 * A call to this helper is susceptible to change the underlying 2009 * packet buffer. Therefore, at load time, all checks on pointers 2010 * previously done by the verifier are invalidated and must be 2011 * performed again, if the helper is used in combination with 2012 * direct packet access. 2013 * Return 2014 * 0 on success, or a negative error in case of failure. 2015 * 2016 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 2017 * Description 2018 * Recompute the layer 3 (e.g. IP) checksum for the packet 2019 * associated to *skb*. Computation is incremental, so the helper 2020 * must know the former value of the header field that was 2021 * modified (*from*), the new value of this field (*to*), and the 2022 * number of bytes (2 or 4) for this field, stored in *size*. 2023 * Alternatively, it is possible to store the difference between 2024 * the previous and the new values of the header field in *to*, by 2025 * setting *from* and *size* to 0. For both methods, *offset* 2026 * indicates the location of the IP checksum within the packet. 2027 * 2028 * This helper works in combination with **bpf_csum_diff**\ (), 2029 * which does not update the checksum in-place, but offers more 2030 * flexibility and can handle sizes larger than 2 or 4 for the 2031 * checksum to update. 2032 * 2033 * A call to this helper is susceptible to change the underlying 2034 * packet buffer. Therefore, at load time, all checks on pointers 2035 * previously done by the verifier are invalidated and must be 2036 * performed again, if the helper is used in combination with 2037 * direct packet access. 2038 * Return 2039 * 0 on success, or a negative error in case of failure. 2040 * 2041 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 2042 * Description 2043 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 2044 * packet associated to *skb*. Computation is incremental, so the 2045 * helper must know the former value of the header field that was 2046 * modified (*from*), the new value of this field (*to*), and the 2047 * number of bytes (2 or 4) for this field, stored on the lowest 2048 * four bits of *flags*. Alternatively, it is possible to store 2049 * the difference between the previous and the new values of the 2050 * header field in *to*, by setting *from* and the four lowest 2051 * bits of *flags* to 0. For both methods, *offset* indicates the 2052 * location of the IP checksum within the packet. In addition to 2053 * the size of the field, *flags* can be added (bitwise OR) actual 2054 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 2055 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 2056 * for updates resulting in a null checksum the value is set to 2057 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 2058 * that the modified header field is part of the pseudo-header. 2059 * Flag **BPF_F_IPV6** should be set for IPv6 packets. 2060 * 2061 * This helper works in combination with **bpf_csum_diff**\ (), 2062 * which does not update the checksum in-place, but offers more 2063 * flexibility and can handle sizes larger than 2 or 4 for the 2064 * checksum to update. 2065 * 2066 * A call to this helper is susceptible to change the underlying 2067 * packet buffer. Therefore, at load time, all checks on pointers 2068 * previously done by the verifier are invalidated and must be 2069 * performed again, if the helper is used in combination with 2070 * direct packet access. 2071 * Return 2072 * 0 on success, or a negative error in case of failure. 2073 * 2074 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2075 * Description 2076 * This special helper is used to trigger a "tail call", or in 2077 * other words, to jump into another eBPF program. The same stack 2078 * frame is used (but values on stack and in registers for the 2079 * caller are not accessible to the callee). This mechanism allows 2080 * for program chaining, either for raising the maximum number of 2081 * available eBPF instructions, or to execute given programs in 2082 * conditional blocks. For security reasons, there is an upper 2083 * limit to the number of successive tail calls that can be 2084 * performed. 2085 * 2086 * Upon call of this helper, the program attempts to jump into a 2087 * program referenced at index *index* in *prog_array_map*, a 2088 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2089 * *ctx*, a pointer to the context. 2090 * 2091 * If the call succeeds, the kernel immediately runs the first 2092 * instruction of the new program. This is not a function call, 2093 * and it never returns to the previous program. If the call 2094 * fails, then the helper has no effect, and the caller continues 2095 * to run its subsequent instructions. A call can fail if the 2096 * destination program for the jump does not exist (i.e. *index* 2097 * is superior to the number of entries in *prog_array_map*), or 2098 * if the maximum number of tail calls has been reached for this 2099 * chain of programs. This limit is defined in the kernel by the 2100 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2101 * which is currently set to 33. 2102 * Return 2103 * 0 on success, or a negative error in case of failure. 2104 * 2105 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2106 * Description 2107 * Clone and redirect the packet associated to *skb* to another 2108 * net device of index *ifindex*. Both ingress and egress 2109 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2110 * value in *flags* is used to make the distinction (ingress path 2111 * is selected if the flag is present, egress path otherwise). 2112 * This is the only flag supported for now. 2113 * 2114 * In comparison with **bpf_redirect**\ () helper, 2115 * **bpf_clone_redirect**\ () has the associated cost of 2116 * duplicating the packet buffer, but this can be executed out of 2117 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2118 * efficient, but it is handled through an action code where the 2119 * redirection happens only after the eBPF program has returned. 2120 * 2121 * A call to this helper is susceptible to change the underlying 2122 * packet buffer. Therefore, at load time, all checks on pointers 2123 * previously done by the verifier are invalidated and must be 2124 * performed again, if the helper is used in combination with 2125 * direct packet access. 2126 * Return 2127 * 0 on success, or a negative error in case of failure. Positive 2128 * error indicates a potential drop or congestion in the target 2129 * device. The particular positive error codes are not defined. 2130 * 2131 * u64 bpf_get_current_pid_tgid(void) 2132 * Description 2133 * Get the current pid and tgid. 2134 * Return 2135 * A 64-bit integer containing the current tgid and pid, and 2136 * created as such: 2137 * *current_task*\ **->tgid << 32 \|** 2138 * *current_task*\ **->pid**. 2139 * 2140 * u64 bpf_get_current_uid_gid(void) 2141 * Description 2142 * Get the current uid and gid. 2143 * Return 2144 * A 64-bit integer containing the current GID and UID, and 2145 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2146 * 2147 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2148 * Description 2149 * Copy the **comm** attribute of the current task into *buf* of 2150 * *size_of_buf*. The **comm** attribute contains the name of 2151 * the executable (excluding the path) for the current task. The 2152 * *size_of_buf* must be strictly positive. On success, the 2153 * helper makes sure that the *buf* is NUL-terminated. On failure, 2154 * it is filled with zeroes. 2155 * Return 2156 * 0 on success, or a negative error in case of failure. 2157 * 2158 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2159 * Description 2160 * Retrieve the classid for the current task, i.e. for the net_cls 2161 * cgroup to which *skb* belongs. 2162 * 2163 * This helper can be used on TC egress path, but not on ingress. 2164 * 2165 * The net_cls cgroup provides an interface to tag network packets 2166 * based on a user-provided identifier for all traffic coming from 2167 * the tasks belonging to the related cgroup. See also the related 2168 * kernel documentation, available from the Linux sources in file 2169 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2170 * 2171 * The Linux kernel has two versions for cgroups: there are 2172 * cgroups v1 and cgroups v2. Both are available to users, who can 2173 * use a mixture of them, but note that the net_cls cgroup is for 2174 * cgroup v1 only. This makes it incompatible with BPF programs 2175 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2176 * only hold data for one version of cgroups at a time). 2177 * 2178 * This helper is only available is the kernel was compiled with 2179 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2180 * "**y**" or to "**m**". 2181 * Return 2182 * The classid, or 0 for the default unconfigured classid. 2183 * 2184 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2185 * Description 2186 * Push a *vlan_tci* (VLAN tag control information) of protocol 2187 * *vlan_proto* to the packet associated to *skb*, then update 2188 * the checksum. Note that if *vlan_proto* is different from 2189 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2190 * be **ETH_P_8021Q**. 2191 * 2192 * A call to this helper is susceptible to change the underlying 2193 * packet buffer. Therefore, at load time, all checks on pointers 2194 * previously done by the verifier are invalidated and must be 2195 * performed again, if the helper is used in combination with 2196 * direct packet access. 2197 * Return 2198 * 0 on success, or a negative error in case of failure. 2199 * 2200 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2201 * Description 2202 * Pop a VLAN header from the packet associated to *skb*. 2203 * 2204 * A call to this helper is susceptible to change the underlying 2205 * packet buffer. Therefore, at load time, all checks on pointers 2206 * previously done by the verifier are invalidated and must be 2207 * performed again, if the helper is used in combination with 2208 * direct packet access. 2209 * Return 2210 * 0 on success, or a negative error in case of failure. 2211 * 2212 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2213 * Description 2214 * Get tunnel metadata. This helper takes a pointer *key* to an 2215 * empty **struct bpf_tunnel_key** of **size**, that will be 2216 * filled with tunnel metadata for the packet associated to *skb*. 2217 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2218 * indicates that the tunnel is based on IPv6 protocol instead of 2219 * IPv4. 2220 * 2221 * The **struct bpf_tunnel_key** is an object that generalizes the 2222 * principal parameters used by various tunneling protocols into a 2223 * single struct. This way, it can be used to easily make a 2224 * decision based on the contents of the encapsulation header, 2225 * "summarized" in this struct. In particular, it holds the IP 2226 * address of the remote end (IPv4 or IPv6, depending on the case) 2227 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2228 * this struct exposes the *key*\ **->tunnel_id**, which is 2229 * generally mapped to a VNI (Virtual Network Identifier), making 2230 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2231 * () helper. 2232 * 2233 * Let's imagine that the following code is part of a program 2234 * attached to the TC ingress interface, on one end of a GRE 2235 * tunnel, and is supposed to filter out all messages coming from 2236 * remote ends with IPv4 address other than 10.0.0.1: 2237 * 2238 * :: 2239 * 2240 * int ret; 2241 * struct bpf_tunnel_key key = {}; 2242 * 2243 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2244 * if (ret < 0) 2245 * return TC_ACT_SHOT; // drop packet 2246 * 2247 * if (key.remote_ipv4 != 0x0a000001) 2248 * return TC_ACT_SHOT; // drop packet 2249 * 2250 * return TC_ACT_OK; // accept packet 2251 * 2252 * This interface can also be used with all encapsulation devices 2253 * that can operate in "collect metadata" mode: instead of having 2254 * one network device per specific configuration, the "collect 2255 * metadata" mode only requires a single device where the 2256 * configuration can be extracted from this helper. 2257 * 2258 * This can be used together with various tunnels such as VXLan, 2259 * Geneve, GRE or IP in IP (IPIP). 2260 * Return 2261 * 0 on success, or a negative error in case of failure. 2262 * 2263 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2264 * Description 2265 * Populate tunnel metadata for packet associated to *skb.* The 2266 * tunnel metadata is set to the contents of *key*, of *size*. The 2267 * *flags* can be set to a combination of the following values: 2268 * 2269 * **BPF_F_TUNINFO_IPV6** 2270 * Indicate that the tunnel is based on IPv6 protocol 2271 * instead of IPv4. 2272 * **BPF_F_ZERO_CSUM_TX** 2273 * For IPv4 packets, add a flag to tunnel metadata 2274 * indicating that checksum computation should be skipped 2275 * and checksum set to zeroes. 2276 * **BPF_F_DONT_FRAGMENT** 2277 * Add a flag to tunnel metadata indicating that the 2278 * packet should not be fragmented. 2279 * **BPF_F_SEQ_NUMBER** 2280 * Add a flag to tunnel metadata indicating that a 2281 * sequence number should be added to tunnel header before 2282 * sending the packet. This flag was added for GRE 2283 * encapsulation, but might be used with other protocols 2284 * as well in the future. 2285 * **BPF_F_NO_TUNNEL_KEY** 2286 * Add a flag to tunnel metadata indicating that no tunnel 2287 * key should be set in the resulting tunnel header. 2288 * 2289 * Here is a typical usage on the transmit path: 2290 * 2291 * :: 2292 * 2293 * struct bpf_tunnel_key key; 2294 * populate key ... 2295 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2296 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2297 * 2298 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2299 * helper for additional information. 2300 * Return 2301 * 0 on success, or a negative error in case of failure. 2302 * 2303 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2304 * Description 2305 * Read the value of a perf event counter. This helper relies on a 2306 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2307 * the perf event counter is selected when *map* is updated with 2308 * perf event file descriptors. The *map* is an array whose size 2309 * is the number of available CPUs, and each cell contains a value 2310 * relative to one CPU. The value to retrieve is indicated by 2311 * *flags*, that contains the index of the CPU to look up, masked 2312 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2313 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2314 * current CPU should be retrieved. 2315 * 2316 * Note that before Linux 4.13, only hardware perf event can be 2317 * retrieved. 2318 * 2319 * Also, be aware that the newer helper 2320 * **bpf_perf_event_read_value**\ () is recommended over 2321 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2322 * quirks where error and counter value are used as a return code 2323 * (which is wrong to do since ranges may overlap). This issue is 2324 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2325 * time provides more features over the **bpf_perf_event_read**\ 2326 * () interface. Please refer to the description of 2327 * **bpf_perf_event_read_value**\ () for details. 2328 * Return 2329 * The value of the perf event counter read from the map, or a 2330 * negative error code in case of failure. 2331 * 2332 * long bpf_redirect(u32 ifindex, u64 flags) 2333 * Description 2334 * Redirect the packet to another net device of index *ifindex*. 2335 * This helper is somewhat similar to **bpf_clone_redirect**\ 2336 * (), except that the packet is not cloned, which provides 2337 * increased performance. 2338 * 2339 * Except for XDP, both ingress and egress interfaces can be used 2340 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2341 * to make the distinction (ingress path is selected if the flag 2342 * is present, egress path otherwise). Currently, XDP only 2343 * supports redirection to the egress interface, and accepts no 2344 * flag at all. 2345 * 2346 * The same effect can also be attained with the more generic 2347 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2348 * redirect target instead of providing it directly to the helper. 2349 * Return 2350 * For XDP, the helper returns **XDP_REDIRECT** on success or 2351 * **XDP_ABORTED** on error. For other program types, the values 2352 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2353 * error. 2354 * 2355 * u32 bpf_get_route_realm(struct sk_buff *skb) 2356 * Description 2357 * Retrieve the realm or the route, that is to say the 2358 * **tclassid** field of the destination for the *skb*. The 2359 * identifier retrieved is a user-provided tag, similar to the 2360 * one used with the net_cls cgroup (see description for 2361 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2362 * held by a route (a destination entry), not by a task. 2363 * 2364 * Retrieving this identifier works with the clsact TC egress hook 2365 * (see also **tc-bpf(8)**), or alternatively on conventional 2366 * classful egress qdiscs, but not on TC ingress path. In case of 2367 * clsact TC egress hook, this has the advantage that, internally, 2368 * the destination entry has not been dropped yet in the transmit 2369 * path. Therefore, the destination entry does not need to be 2370 * artificially held via **netif_keep_dst**\ () for a classful 2371 * qdisc until the *skb* is freed. 2372 * 2373 * This helper is available only if the kernel was compiled with 2374 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2375 * Return 2376 * The realm of the route for the packet associated to *skb*, or 0 2377 * if none was found. 2378 * 2379 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2380 * Description 2381 * Write raw *data* blob into a special BPF perf event held by 2382 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2383 * event must have the following attributes: **PERF_SAMPLE_RAW** 2384 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2385 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2386 * 2387 * The *flags* are used to indicate the index in *map* for which 2388 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2389 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2390 * to indicate that the index of the current CPU core should be 2391 * used. 2392 * 2393 * The value to write, of *size*, is passed through eBPF stack and 2394 * pointed by *data*. 2395 * 2396 * The context of the program *ctx* needs also be passed to the 2397 * helper. 2398 * 2399 * On user space, a program willing to read the values needs to 2400 * call **perf_event_open**\ () on the perf event (either for 2401 * one or for all CPUs) and to store the file descriptor into the 2402 * *map*. This must be done before the eBPF program can send data 2403 * into it. An example is available in file 2404 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2405 * tree (the eBPF program counterpart is in 2406 * *samples/bpf/trace_output_kern.c*). 2407 * 2408 * **bpf_perf_event_output**\ () achieves better performance 2409 * than **bpf_trace_printk**\ () for sharing data with user 2410 * space, and is much better suitable for streaming data from eBPF 2411 * programs. 2412 * 2413 * Note that this helper is not restricted to tracing use cases 2414 * and can be used with programs attached to TC or XDP as well, 2415 * where it allows for passing data to user space listeners. Data 2416 * can be: 2417 * 2418 * * Only custom structs, 2419 * * Only the packet payload, or 2420 * * A combination of both. 2421 * Return 2422 * 0 on success, or a negative error in case of failure. 2423 * 2424 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2425 * Description 2426 * This helper was provided as an easy way to load data from a 2427 * packet. It can be used to load *len* bytes from *offset* from 2428 * the packet associated to *skb*, into the buffer pointed by 2429 * *to*. 2430 * 2431 * Since Linux 4.7, usage of this helper has mostly been replaced 2432 * by "direct packet access", enabling packet data to be 2433 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2434 * pointing respectively to the first byte of packet data and to 2435 * the byte after the last byte of packet data. However, it 2436 * remains useful if one wishes to read large quantities of data 2437 * at once from a packet into the eBPF stack. 2438 * Return 2439 * 0 on success, or a negative error in case of failure. 2440 * 2441 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2442 * Description 2443 * Walk a user or a kernel stack and return its id. To achieve 2444 * this, the helper needs *ctx*, which is a pointer to the context 2445 * on which the tracing program is executed, and a pointer to a 2446 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2447 * 2448 * The last argument, *flags*, holds the number of stack frames to 2449 * skip (from 0 to 255), masked with 2450 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2451 * a combination of the following flags: 2452 * 2453 * **BPF_F_USER_STACK** 2454 * Collect a user space stack instead of a kernel stack. 2455 * **BPF_F_FAST_STACK_CMP** 2456 * Compare stacks by hash only. 2457 * **BPF_F_REUSE_STACKID** 2458 * If two different stacks hash into the same *stackid*, 2459 * discard the old one. 2460 * 2461 * The stack id retrieved is a 32 bit long integer handle which 2462 * can be further combined with other data (including other stack 2463 * ids) and used as a key into maps. This can be useful for 2464 * generating a variety of graphs (such as flame graphs or off-cpu 2465 * graphs). 2466 * 2467 * For walking a stack, this helper is an improvement over 2468 * **bpf_probe_read**\ (), which can be used with unrolled loops 2469 * but is not efficient and consumes a lot of eBPF instructions. 2470 * Instead, **bpf_get_stackid**\ () can collect up to 2471 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2472 * this limit can be controlled with the **sysctl** program, and 2473 * that it should be manually increased in order to profile long 2474 * user stacks (such as stacks for Java programs). To do so, use: 2475 * 2476 * :: 2477 * 2478 * # sysctl kernel.perf_event_max_stack=<new value> 2479 * Return 2480 * The positive or null stack id on success, or a negative error 2481 * in case of failure. 2482 * 2483 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2484 * Description 2485 * Compute a checksum difference, from the raw buffer pointed by 2486 * *from*, of length *from_size* (that must be a multiple of 4), 2487 * towards the raw buffer pointed by *to*, of size *to_size* 2488 * (same remark). An optional *seed* can be added to the value 2489 * (this can be cascaded, the seed may come from a previous call 2490 * to the helper). 2491 * 2492 * This is flexible enough to be used in several ways: 2493 * 2494 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2495 * checksum, it can be used when pushing new data. 2496 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2497 * checksum, it can be used when removing data from a packet. 2498 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2499 * can be used to compute a diff. Note that *from_size* and 2500 * *to_size* do not need to be equal. 2501 * 2502 * This helper can be used in combination with 2503 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2504 * which one can feed in the difference computed with 2505 * **bpf_csum_diff**\ (). 2506 * Return 2507 * The checksum result, or a negative error code in case of 2508 * failure. 2509 * 2510 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2511 * Description 2512 * Retrieve tunnel options metadata for the packet associated to 2513 * *skb*, and store the raw tunnel option data to the buffer *opt* 2514 * of *size*. 2515 * 2516 * This helper can be used with encapsulation devices that can 2517 * operate in "collect metadata" mode (please refer to the related 2518 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2519 * more details). A particular example where this can be used is 2520 * in combination with the Geneve encapsulation protocol, where it 2521 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2522 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2523 * the eBPF program. This allows for full customization of these 2524 * headers. 2525 * Return 2526 * The size of the option data retrieved. 2527 * 2528 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2529 * Description 2530 * Set tunnel options metadata for the packet associated to *skb* 2531 * to the option data contained in the raw buffer *opt* of *size*. 2532 * 2533 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2534 * helper for additional information. 2535 * Return 2536 * 0 on success, or a negative error in case of failure. 2537 * 2538 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2539 * Description 2540 * Change the protocol of the *skb* to *proto*. Currently 2541 * supported are transition from IPv4 to IPv6, and from IPv6 to 2542 * IPv4. The helper takes care of the groundwork for the 2543 * transition, including resizing the socket buffer. The eBPF 2544 * program is expected to fill the new headers, if any, via 2545 * **skb_store_bytes**\ () and to recompute the checksums with 2546 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2547 * (). The main case for this helper is to perform NAT64 2548 * operations out of an eBPF program. 2549 * 2550 * Internally, the GSO type is marked as dodgy so that headers are 2551 * checked and segments are recalculated by the GSO/GRO engine. 2552 * The size for GSO target is adapted as well. 2553 * 2554 * All values for *flags* are reserved for future usage, and must 2555 * be left at zero. 2556 * 2557 * A call to this helper is susceptible to change the underlying 2558 * packet buffer. Therefore, at load time, all checks on pointers 2559 * previously done by the verifier are invalidated and must be 2560 * performed again, if the helper is used in combination with 2561 * direct packet access. 2562 * Return 2563 * 0 on success, or a negative error in case of failure. 2564 * 2565 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2566 * Description 2567 * Change the packet type for the packet associated to *skb*. This 2568 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2569 * the eBPF program does not have a write access to *skb*\ 2570 * **->pkt_type** beside this helper. Using a helper here allows 2571 * for graceful handling of errors. 2572 * 2573 * The major use case is to change incoming *skb*s to 2574 * **PACKET_HOST** in a programmatic way instead of having to 2575 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2576 * example. 2577 * 2578 * Note that *type* only allows certain values. At this time, they 2579 * are: 2580 * 2581 * **PACKET_HOST** 2582 * Packet is for us. 2583 * **PACKET_BROADCAST** 2584 * Send packet to all. 2585 * **PACKET_MULTICAST** 2586 * Send packet to group. 2587 * **PACKET_OTHERHOST** 2588 * Send packet to someone else. 2589 * Return 2590 * 0 on success, or a negative error in case of failure. 2591 * 2592 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2593 * Description 2594 * Check whether *skb* is a descendant of the cgroup2 held by 2595 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2596 * Return 2597 * The return value depends on the result of the test, and can be: 2598 * 2599 * * 0, if the *skb* failed the cgroup2 descendant test. 2600 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2601 * * A negative error code, if an error occurred. 2602 * 2603 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2604 * Description 2605 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2606 * not set, in particular if the hash was cleared due to mangling, 2607 * recompute this hash. Later accesses to the hash can be done 2608 * directly with *skb*\ **->hash**. 2609 * 2610 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2611 * prototype with **bpf_skb_change_proto**\ (), or calling 2612 * **bpf_skb_store_bytes**\ () with the 2613 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2614 * the hash and to trigger a new computation for the next call to 2615 * **bpf_get_hash_recalc**\ (). 2616 * Return 2617 * The 32-bit hash. 2618 * 2619 * u64 bpf_get_current_task(void) 2620 * Description 2621 * Get the current task. 2622 * Return 2623 * A pointer to the current task struct. 2624 * 2625 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2626 * Description 2627 * Attempt in a safe way to write *len* bytes from the buffer 2628 * *src* to *dst* in memory. It only works for threads that are in 2629 * user context, and *dst* must be a valid user space address. 2630 * 2631 * This helper should not be used to implement any kind of 2632 * security mechanism because of TOC-TOU attacks, but rather to 2633 * debug, divert, and manipulate execution of semi-cooperative 2634 * processes. 2635 * 2636 * Keep in mind that this feature is meant for experiments, and it 2637 * has a risk of crashing the system and running programs. 2638 * Therefore, when an eBPF program using this helper is attached, 2639 * a warning including PID and process name is printed to kernel 2640 * logs. 2641 * Return 2642 * 0 on success, or a negative error in case of failure. 2643 * 2644 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2645 * Description 2646 * Check whether the probe is being run is the context of a given 2647 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2648 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2649 * Return 2650 * The return value depends on the result of the test, and can be: 2651 * 2652 * * 1, if current task belongs to the cgroup2. 2653 * * 0, if current task does not belong to the cgroup2. 2654 * * A negative error code, if an error occurred. 2655 * 2656 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2657 * Description 2658 * Resize (trim or grow) the packet associated to *skb* to the 2659 * new *len*. The *flags* are reserved for future usage, and must 2660 * be left at zero. 2661 * 2662 * The basic idea is that the helper performs the needed work to 2663 * change the size of the packet, then the eBPF program rewrites 2664 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2665 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2666 * and others. This helper is a slow path utility intended for 2667 * replies with control messages. And because it is targeted for 2668 * slow path, the helper itself can afford to be slow: it 2669 * implicitly linearizes, unclones and drops offloads from the 2670 * *skb*. 2671 * 2672 * A call to this helper is susceptible to change the underlying 2673 * packet buffer. Therefore, at load time, all checks on pointers 2674 * previously done by the verifier are invalidated and must be 2675 * performed again, if the helper is used in combination with 2676 * direct packet access. 2677 * Return 2678 * 0 on success, or a negative error in case of failure. 2679 * 2680 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2681 * Description 2682 * Pull in non-linear data in case the *skb* is non-linear and not 2683 * all of *len* are part of the linear section. Make *len* bytes 2684 * from *skb* readable and writable. If a zero value is passed for 2685 * *len*, then all bytes in the linear part of *skb* will be made 2686 * readable and writable. 2687 * 2688 * This helper is only needed for reading and writing with direct 2689 * packet access. 2690 * 2691 * For direct packet access, testing that offsets to access 2692 * are within packet boundaries (test on *skb*\ **->data_end**) is 2693 * susceptible to fail if offsets are invalid, or if the requested 2694 * data is in non-linear parts of the *skb*. On failure the 2695 * program can just bail out, or in the case of a non-linear 2696 * buffer, use a helper to make the data available. The 2697 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2698 * the data. Another one consists in using **bpf_skb_pull_data** 2699 * to pull in once the non-linear parts, then retesting and 2700 * eventually access the data. 2701 * 2702 * At the same time, this also makes sure the *skb* is uncloned, 2703 * which is a necessary condition for direct write. As this needs 2704 * to be an invariant for the write part only, the verifier 2705 * detects writes and adds a prologue that is calling 2706 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2707 * the very beginning in case it is indeed cloned. 2708 * 2709 * A call to this helper is susceptible to change the underlying 2710 * packet buffer. Therefore, at load time, all checks on pointers 2711 * previously done by the verifier are invalidated and must be 2712 * performed again, if the helper is used in combination with 2713 * direct packet access. 2714 * Return 2715 * 0 on success, or a negative error in case of failure. 2716 * 2717 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2718 * Description 2719 * Add the checksum *csum* into *skb*\ **->csum** in case the 2720 * driver has supplied a checksum for the entire packet into that 2721 * field. Return an error otherwise. This helper is intended to be 2722 * used in combination with **bpf_csum_diff**\ (), in particular 2723 * when the checksum needs to be updated after data has been 2724 * written into the packet through direct packet access. 2725 * Return 2726 * The checksum on success, or a negative error code in case of 2727 * failure. 2728 * 2729 * void bpf_set_hash_invalid(struct sk_buff *skb) 2730 * Description 2731 * Invalidate the current *skb*\ **->hash**. It can be used after 2732 * mangling on headers through direct packet access, in order to 2733 * indicate that the hash is outdated and to trigger a 2734 * recalculation the next time the kernel tries to access this 2735 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2736 * Return 2737 * void. 2738 * 2739 * long bpf_get_numa_node_id(void) 2740 * Description 2741 * Return the id of the current NUMA node. The primary use case 2742 * for this helper is the selection of sockets for the local NUMA 2743 * node, when the program is attached to sockets using the 2744 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2745 * but the helper is also available to other eBPF program types, 2746 * similarly to **bpf_get_smp_processor_id**\ (). 2747 * Return 2748 * The id of current NUMA node. 2749 * 2750 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2751 * Description 2752 * Grows headroom of packet associated to *skb* and adjusts the 2753 * offset of the MAC header accordingly, adding *len* bytes of 2754 * space. It automatically extends and reallocates memory as 2755 * required. 2756 * 2757 * This helper can be used on a layer 3 *skb* to push a MAC header 2758 * for redirection into a layer 2 device. 2759 * 2760 * All values for *flags* are reserved for future usage, and must 2761 * be left at zero. 2762 * 2763 * A call to this helper is susceptible to change the underlying 2764 * packet buffer. Therefore, at load time, all checks on pointers 2765 * previously done by the verifier are invalidated and must be 2766 * performed again, if the helper is used in combination with 2767 * direct packet access. 2768 * Return 2769 * 0 on success, or a negative error in case of failure. 2770 * 2771 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2772 * Description 2773 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2774 * it is possible to use a negative value for *delta*. This helper 2775 * can be used to prepare the packet for pushing or popping 2776 * headers. 2777 * 2778 * A call to this helper is susceptible to change the underlying 2779 * packet buffer. Therefore, at load time, all checks on pointers 2780 * previously done by the verifier are invalidated and must be 2781 * performed again, if the helper is used in combination with 2782 * direct packet access. 2783 * Return 2784 * 0 on success, or a negative error in case of failure. 2785 * 2786 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2787 * Description 2788 * Copy a NUL terminated string from an unsafe kernel address 2789 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2790 * more details. 2791 * 2792 * Generally, use **bpf_probe_read_user_str**\ () or 2793 * **bpf_probe_read_kernel_str**\ () instead. 2794 * Return 2795 * On success, the strictly positive length of the string, 2796 * including the trailing NUL character. On error, a negative 2797 * value. 2798 * 2799 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2800 * Description 2801 * If the **struct sk_buff** pointed by *skb* has a known socket, 2802 * retrieve the cookie (generated by the kernel) of this socket. 2803 * If no cookie has been set yet, generate a new cookie. Once 2804 * generated, the socket cookie remains stable for the life of the 2805 * socket. This helper can be useful for monitoring per socket 2806 * networking traffic statistics as it provides a global socket 2807 * identifier that can be assumed unique. 2808 * Return 2809 * A 8-byte long unique number on success, or 0 if the socket 2810 * field is missing inside *skb*. 2811 * 2812 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2813 * Description 2814 * Equivalent to bpf_get_socket_cookie() helper that accepts 2815 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2816 * Return 2817 * A 8-byte long unique number. 2818 * 2819 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2820 * Description 2821 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2822 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2823 * Return 2824 * A 8-byte long unique number. 2825 * 2826 * u64 bpf_get_socket_cookie(struct sock *sk) 2827 * Description 2828 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2829 * *sk*, but gets socket from a BTF **struct sock**. This helper 2830 * also works for sleepable programs. 2831 * Return 2832 * A 8-byte long unique number or 0 if *sk* is NULL. 2833 * 2834 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2835 * Description 2836 * Get the owner UID of the socked associated to *skb*. 2837 * Return 2838 * The owner UID of the socket associated to *skb*. If the socket 2839 * is **NULL**, or if it is not a full socket (i.e. if it is a 2840 * time-wait or a request socket instead), **overflowuid** value 2841 * is returned (note that **overflowuid** might also be the actual 2842 * UID value for the socket). 2843 * 2844 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2845 * Description 2846 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2847 * to value *hash*. 2848 * Return 2849 * 0 2850 * 2851 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2852 * Description 2853 * Emulate a call to **setsockopt()** on the socket associated to 2854 * *bpf_socket*, which must be a full socket. The *level* at 2855 * which the option resides and the name *optname* of the option 2856 * must be specified, see **setsockopt(2)** for more information. 2857 * The option value of length *optlen* is pointed by *optval*. 2858 * 2859 * *bpf_socket* should be one of the following: 2860 * 2861 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2862 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2863 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2864 * 2865 * This helper actually implements a subset of **setsockopt()**. 2866 * It supports the following *level*\ s: 2867 * 2868 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2869 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2870 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2871 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2872 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2873 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2874 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2875 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2876 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2877 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2878 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2879 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2880 * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**. 2881 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2882 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2883 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2884 * Return 2885 * 0 on success, or a negative error in case of failure. 2886 * 2887 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2888 * Description 2889 * Grow or shrink the room for data in the packet associated to 2890 * *skb* by *len_diff*, and according to the selected *mode*. 2891 * 2892 * By default, the helper will reset any offloaded checksum 2893 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2894 * by the following flag: 2895 * 2896 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2897 * checksum data of the skb to CHECKSUM_NONE. 2898 * 2899 * There are two supported modes at this time: 2900 * 2901 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2902 * (room space is added or removed between the layer 2 and 2903 * layer 3 headers). 2904 * 2905 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2906 * (room space is added or removed between the layer 3 and 2907 * layer 4 headers). 2908 * 2909 * The following flags are supported at this time: 2910 * 2911 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2912 * Adjusting mss in this way is not allowed for datagrams. 2913 * 2914 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2915 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2916 * Any new space is reserved to hold a tunnel header. 2917 * Configure skb offsets and other fields accordingly. 2918 * 2919 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2920 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2921 * Use with ENCAP_L3 flags to further specify the tunnel type. 2922 * 2923 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2924 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2925 * type; *len* is the length of the inner MAC header. 2926 * 2927 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2928 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2929 * L2 type as Ethernet. 2930 * 2931 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2932 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2933 * Indicate the new IP header version after decapsulating the outer 2934 * IP header. Used when the inner and outer IP versions are different. 2935 * 2936 * A call to this helper is susceptible to change the underlying 2937 * packet buffer. Therefore, at load time, all checks on pointers 2938 * previously done by the verifier are invalidated and must be 2939 * performed again, if the helper is used in combination with 2940 * direct packet access. 2941 * Return 2942 * 0 on success, or a negative error in case of failure. 2943 * 2944 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2945 * Description 2946 * Redirect the packet to the endpoint referenced by *map* at 2947 * index *key*. Depending on its type, this *map* can contain 2948 * references to net devices (for forwarding packets through other 2949 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2950 * but this is only implemented for native XDP (with driver 2951 * support) as of this writing). 2952 * 2953 * The lower two bits of *flags* are used as the return code if 2954 * the map lookup fails. This is so that the return value can be 2955 * one of the XDP program return codes up to **XDP_TX**, as chosen 2956 * by the caller. The higher bits of *flags* can be set to 2957 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2958 * 2959 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2960 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2961 * interface will be excluded when do broadcasting. 2962 * 2963 * See also **bpf_redirect**\ (), which only supports redirecting 2964 * to an ifindex, but doesn't require a map to do so. 2965 * Return 2966 * **XDP_REDIRECT** on success, or the value of the two lower bits 2967 * of the *flags* argument on error. 2968 * 2969 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2970 * Description 2971 * Redirect the packet to the socket referenced by *map* (of type 2972 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2973 * egress interfaces can be used for redirection. The 2974 * **BPF_F_INGRESS** value in *flags* is used to make the 2975 * distinction (ingress path is selected if the flag is present, 2976 * egress path otherwise). This is the only flag supported for now. 2977 * Return 2978 * **SK_PASS** on success, or **SK_DROP** on error. 2979 * 2980 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2981 * Description 2982 * Add an entry to, or update a *map* referencing sockets. The 2983 * *skops* is used as a new value for the entry associated to 2984 * *key*. *flags* is one of: 2985 * 2986 * **BPF_NOEXIST** 2987 * The entry for *key* must not exist in the map. 2988 * **BPF_EXIST** 2989 * The entry for *key* must already exist in the map. 2990 * **BPF_ANY** 2991 * No condition on the existence of the entry for *key*. 2992 * 2993 * If the *map* has eBPF programs (parser and verdict), those will 2994 * be inherited by the socket being added. If the socket is 2995 * already attached to eBPF programs, this results in an error. 2996 * Return 2997 * 0 on success, or a negative error in case of failure. 2998 * 2999 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 3000 * Description 3001 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 3002 * *delta* (which can be positive or negative). Note that this 3003 * operation modifies the address stored in *xdp_md*\ **->data**, 3004 * so the latter must be loaded only after the helper has been 3005 * called. 3006 * 3007 * The use of *xdp_md*\ **->data_meta** is optional and programs 3008 * are not required to use it. The rationale is that when the 3009 * packet is processed with XDP (e.g. as DoS filter), it is 3010 * possible to push further meta data along with it before passing 3011 * to the stack, and to give the guarantee that an ingress eBPF 3012 * program attached as a TC classifier on the same device can pick 3013 * this up for further post-processing. Since TC works with socket 3014 * buffers, it remains possible to set from XDP the **mark** or 3015 * **priority** pointers, or other pointers for the socket buffer. 3016 * Having this scratch space generic and programmable allows for 3017 * more flexibility as the user is free to store whatever meta 3018 * data they need. 3019 * 3020 * A call to this helper is susceptible to change the underlying 3021 * packet buffer. Therefore, at load time, all checks on pointers 3022 * previously done by the verifier are invalidated and must be 3023 * performed again, if the helper is used in combination with 3024 * direct packet access. 3025 * Return 3026 * 0 on success, or a negative error in case of failure. 3027 * 3028 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 3029 * Description 3030 * Read the value of a perf event counter, and store it into *buf* 3031 * of size *buf_size*. This helper relies on a *map* of type 3032 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 3033 * counter is selected when *map* is updated with perf event file 3034 * descriptors. The *map* is an array whose size is the number of 3035 * available CPUs, and each cell contains a value relative to one 3036 * CPU. The value to retrieve is indicated by *flags*, that 3037 * contains the index of the CPU to look up, masked with 3038 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 3039 * **BPF_F_CURRENT_CPU** to indicate that the value for the 3040 * current CPU should be retrieved. 3041 * 3042 * This helper behaves in a way close to 3043 * **bpf_perf_event_read**\ () helper, save that instead of 3044 * just returning the value observed, it fills the *buf* 3045 * structure. This allows for additional data to be retrieved: in 3046 * particular, the enabled and running times (in *buf*\ 3047 * **->enabled** and *buf*\ **->running**, respectively) are 3048 * copied. In general, **bpf_perf_event_read_value**\ () is 3049 * recommended over **bpf_perf_event_read**\ (), which has some 3050 * ABI issues and provides fewer functionalities. 3051 * 3052 * These values are interesting, because hardware PMU (Performance 3053 * Monitoring Unit) counters are limited resources. When there are 3054 * more PMU based perf events opened than available counters, 3055 * kernel will multiplex these events so each event gets certain 3056 * percentage (but not all) of the PMU time. In case that 3057 * multiplexing happens, the number of samples or counter value 3058 * will not reflect the case compared to when no multiplexing 3059 * occurs. This makes comparison between different runs difficult. 3060 * Typically, the counter value should be normalized before 3061 * comparing to other experiments. The usual normalization is done 3062 * as follows. 3063 * 3064 * :: 3065 * 3066 * normalized_counter = counter * t_enabled / t_running 3067 * 3068 * Where t_enabled is the time enabled for event and t_running is 3069 * the time running for event since last normalization. The 3070 * enabled and running times are accumulated since the perf event 3071 * open. To achieve scaling factor between two invocations of an 3072 * eBPF program, users can use CPU id as the key (which is 3073 * typical for perf array usage model) to remember the previous 3074 * value and do the calculation inside the eBPF program. 3075 * Return 3076 * 0 on success, or a negative error in case of failure. 3077 * 3078 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3079 * Description 3080 * For an eBPF program attached to a perf event, retrieve the 3081 * value of the event counter associated to *ctx* and store it in 3082 * the structure pointed by *buf* and of size *buf_size*. Enabled 3083 * and running times are also stored in the structure (see 3084 * description of helper **bpf_perf_event_read_value**\ () for 3085 * more details). 3086 * Return 3087 * 0 on success, or a negative error in case of failure. 3088 * 3089 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3090 * Description 3091 * Emulate a call to **getsockopt()** on the socket associated to 3092 * *bpf_socket*, which must be a full socket. The *level* at 3093 * which the option resides and the name *optname* of the option 3094 * must be specified, see **getsockopt(2)** for more information. 3095 * The retrieved value is stored in the structure pointed by 3096 * *opval* and of length *optlen*. 3097 * 3098 * *bpf_socket* should be one of the following: 3099 * 3100 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3101 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3102 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3103 * 3104 * This helper actually implements a subset of **getsockopt()**. 3105 * It supports the same set of *optname*\ s that is supported by 3106 * the **bpf_setsockopt**\ () helper. The exceptions are 3107 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3108 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3109 * Return 3110 * 0 on success, or a negative error in case of failure. 3111 * 3112 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3113 * Description 3114 * Used for error injection, this helper uses kprobes to override 3115 * the return value of the probed function, and to set it to *rc*. 3116 * The first argument is the context *regs* on which the kprobe 3117 * works. 3118 * 3119 * This helper works by setting the PC (program counter) 3120 * to an override function which is run in place of the original 3121 * probed function. This means the probed function is not run at 3122 * all. The replacement function just returns with the required 3123 * value. 3124 * 3125 * This helper has security implications, and thus is subject to 3126 * restrictions. It is only available if the kernel was compiled 3127 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3128 * option, and in this case it only works on functions tagged with 3129 * **ALLOW_ERROR_INJECTION** in the kernel code. 3130 * Return 3131 * 0 3132 * 3133 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3134 * Description 3135 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3136 * for the full TCP socket associated to *bpf_sock_ops* to 3137 * *argval*. 3138 * 3139 * The primary use of this field is to determine if there should 3140 * be calls to eBPF programs of type 3141 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3142 * code. A program of the same type can change its value, per 3143 * connection and as necessary, when the connection is 3144 * established. This field is directly accessible for reading, but 3145 * this helper must be used for updates in order to return an 3146 * error if an eBPF program tries to set a callback that is not 3147 * supported in the current kernel. 3148 * 3149 * *argval* is a flag array which can combine these flags: 3150 * 3151 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3152 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3153 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3154 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3155 * 3156 * Therefore, this function can be used to clear a callback flag by 3157 * setting the appropriate bit to zero. e.g. to disable the RTO 3158 * callback: 3159 * 3160 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3161 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3162 * 3163 * Here are some examples of where one could call such eBPF 3164 * program: 3165 * 3166 * * When RTO fires. 3167 * * When a packet is retransmitted. 3168 * * When the connection terminates. 3169 * * When a packet is sent. 3170 * * When a packet is received. 3171 * Return 3172 * Code **-EINVAL** if the socket is not a full TCP socket; 3173 * otherwise, a positive number containing the bits that could not 3174 * be set is returned (which comes down to 0 if all bits were set 3175 * as required). 3176 * 3177 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3178 * Description 3179 * This helper is used in programs implementing policies at the 3180 * socket level. If the message *msg* is allowed to pass (i.e. if 3181 * the verdict eBPF program returns **SK_PASS**), redirect it to 3182 * the socket referenced by *map* (of type 3183 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3184 * egress interfaces can be used for redirection. The 3185 * **BPF_F_INGRESS** value in *flags* is used to make the 3186 * distinction (ingress path is selected if the flag is present, 3187 * egress path otherwise). This is the only flag supported for now. 3188 * Return 3189 * **SK_PASS** on success, or **SK_DROP** on error. 3190 * 3191 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3192 * Description 3193 * For socket policies, apply the verdict of the eBPF program to 3194 * the next *bytes* (number of bytes) of message *msg*. 3195 * 3196 * For example, this helper can be used in the following cases: 3197 * 3198 * * A single **sendmsg**\ () or **sendfile**\ () system call 3199 * contains multiple logical messages that the eBPF program is 3200 * supposed to read and for which it should apply a verdict. 3201 * * An eBPF program only cares to read the first *bytes* of a 3202 * *msg*. If the message has a large payload, then setting up 3203 * and calling the eBPF program repeatedly for all bytes, even 3204 * though the verdict is already known, would create unnecessary 3205 * overhead. 3206 * 3207 * When called from within an eBPF program, the helper sets a 3208 * counter internal to the BPF infrastructure, that is used to 3209 * apply the last verdict to the next *bytes*. If *bytes* is 3210 * smaller than the current data being processed from a 3211 * **sendmsg**\ () or **sendfile**\ () system call, the first 3212 * *bytes* will be sent and the eBPF program will be re-run with 3213 * the pointer for start of data pointing to byte number *bytes* 3214 * **+ 1**. If *bytes* is larger than the current data being 3215 * processed, then the eBPF verdict will be applied to multiple 3216 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3217 * consumed. 3218 * 3219 * Note that if a socket closes with the internal counter holding 3220 * a non-zero value, this is not a problem because data is not 3221 * being buffered for *bytes* and is sent as it is received. 3222 * Return 3223 * 0 3224 * 3225 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3226 * Description 3227 * For socket policies, prevent the execution of the verdict eBPF 3228 * program for message *msg* until *bytes* (byte number) have been 3229 * accumulated. 3230 * 3231 * This can be used when one needs a specific number of bytes 3232 * before a verdict can be assigned, even if the data spans 3233 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3234 * case would be a user calling **sendmsg**\ () repeatedly with 3235 * 1-byte long message segments. Obviously, this is bad for 3236 * performance, but it is still valid. If the eBPF program needs 3237 * *bytes* bytes to validate a header, this helper can be used to 3238 * prevent the eBPF program to be called again until *bytes* have 3239 * been accumulated. 3240 * Return 3241 * 0 3242 * 3243 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3244 * Description 3245 * For socket policies, pull in non-linear data from user space 3246 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3247 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3248 * respectively. 3249 * 3250 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3251 * *msg* it can only parse data that the (**data**, **data_end**) 3252 * pointers have already consumed. For **sendmsg**\ () hooks this 3253 * is likely the first scatterlist element. But for calls relying 3254 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3255 * be the range (**0**, **0**) because the data is shared with 3256 * user space and by default the objective is to avoid allowing 3257 * user space to modify data while (or after) eBPF verdict is 3258 * being decided. This helper can be used to pull in data and to 3259 * set the start and end pointer to given values. Data will be 3260 * copied if necessary (i.e. if data was not linear and if start 3261 * and end pointers do not point to the same chunk). 3262 * 3263 * A call to this helper is susceptible to change the underlying 3264 * packet buffer. Therefore, at load time, all checks on pointers 3265 * previously done by the verifier are invalidated and must be 3266 * performed again, if the helper is used in combination with 3267 * direct packet access. 3268 * 3269 * All values for *flags* are reserved for future usage, and must 3270 * be left at zero. 3271 * Return 3272 * 0 on success, or a negative error in case of failure. 3273 * 3274 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3275 * Description 3276 * Bind the socket associated to *ctx* to the address pointed by 3277 * *addr*, of length *addr_len*. This allows for making outgoing 3278 * connection from the desired IP address, which can be useful for 3279 * example when all processes inside a cgroup should use one 3280 * single IP address on a host that has multiple IP configured. 3281 * 3282 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3283 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3284 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3285 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3286 * behavior and lets the kernel efficiently pick up an unused 3287 * port as long as 4-tuple is unique. Passing non-zero port might 3288 * lead to degraded performance. 3289 * Return 3290 * 0 on success, or a negative error in case of failure. 3291 * 3292 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3293 * Description 3294 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3295 * possible to both shrink and grow the packet tail. 3296 * Shrink done via *delta* being a negative integer. 3297 * 3298 * A call to this helper is susceptible to change the underlying 3299 * packet buffer. Therefore, at load time, all checks on pointers 3300 * previously done by the verifier are invalidated and must be 3301 * performed again, if the helper is used in combination with 3302 * direct packet access. 3303 * Return 3304 * 0 on success, or a negative error in case of failure. 3305 * 3306 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3307 * Description 3308 * Retrieve the XFRM state (IP transform framework, see also 3309 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3310 * 3311 * The retrieved value is stored in the **struct bpf_xfrm_state** 3312 * pointed by *xfrm_state* and of length *size*. 3313 * 3314 * All values for *flags* are reserved for future usage, and must 3315 * be left at zero. 3316 * 3317 * This helper is available only if the kernel was compiled with 3318 * **CONFIG_XFRM** configuration option. 3319 * Return 3320 * 0 on success, or a negative error in case of failure. 3321 * 3322 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3323 * Description 3324 * Return a user or a kernel stack in bpf program provided buffer. 3325 * To achieve this, the helper needs *ctx*, which is a pointer 3326 * to the context on which the tracing program is executed. 3327 * To store the stacktrace, the bpf program provides *buf* with 3328 * a nonnegative *size*. 3329 * 3330 * The last argument, *flags*, holds the number of stack frames to 3331 * skip (from 0 to 255), masked with 3332 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3333 * the following flags: 3334 * 3335 * **BPF_F_USER_STACK** 3336 * Collect a user space stack instead of a kernel stack. 3337 * **BPF_F_USER_BUILD_ID** 3338 * Collect (build_id, file_offset) instead of ips for user 3339 * stack, only valid if **BPF_F_USER_STACK** is also 3340 * specified. 3341 * 3342 * *file_offset* is an offset relative to the beginning 3343 * of the executable or shared object file backing the vma 3344 * which the *ip* falls in. It is *not* an offset relative 3345 * to that object's base address. Accordingly, it must be 3346 * adjusted by adding (sh_addr - sh_offset), where 3347 * sh_{addr,offset} correspond to the executable section 3348 * containing *file_offset* in the object, for comparisons 3349 * to symbols' st_value to be valid. 3350 * 3351 * **bpf_get_stack**\ () can collect up to 3352 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3353 * to sufficient large buffer size. Note that 3354 * this limit can be controlled with the **sysctl** program, and 3355 * that it should be manually increased in order to profile long 3356 * user stacks (such as stacks for Java programs). To do so, use: 3357 * 3358 * :: 3359 * 3360 * # sysctl kernel.perf_event_max_stack=<new value> 3361 * Return 3362 * The non-negative copied *buf* length equal to or less than 3363 * *size* on success, or a negative error in case of failure. 3364 * 3365 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3366 * Description 3367 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3368 * it provides an easy way to load *len* bytes from *offset* 3369 * from the packet associated to *skb*, into the buffer pointed 3370 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3371 * a fifth argument *start_header* exists in order to select a 3372 * base offset to start from. *start_header* can be one of: 3373 * 3374 * **BPF_HDR_START_MAC** 3375 * Base offset to load data from is *skb*'s mac header. 3376 * **BPF_HDR_START_NET** 3377 * Base offset to load data from is *skb*'s network header. 3378 * 3379 * In general, "direct packet access" is the preferred method to 3380 * access packet data, however, this helper is in particular useful 3381 * in socket filters where *skb*\ **->data** does not always point 3382 * to the start of the mac header and where "direct packet access" 3383 * is not available. 3384 * Return 3385 * 0 on success, or a negative error in case of failure. 3386 * 3387 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3388 * Description 3389 * Do FIB lookup in kernel tables using parameters in *params*. 3390 * If lookup is successful and result shows packet is to be 3391 * forwarded, the neighbor tables are searched for the nexthop. 3392 * If successful (ie., FIB lookup shows forwarding and nexthop 3393 * is resolved), the nexthop address is returned in ipv4_dst 3394 * or ipv6_dst based on family, smac is set to mac address of 3395 * egress device, dmac is set to nexthop mac address, rt_metric 3396 * is set to metric from route (IPv4/IPv6 only), and ifindex 3397 * is set to the device index of the nexthop from the FIB lookup. 3398 * 3399 * *plen* argument is the size of the passed in struct. 3400 * *flags* argument can be a combination of one or more of the 3401 * following values: 3402 * 3403 * **BPF_FIB_LOOKUP_DIRECT** 3404 * Do a direct table lookup vs full lookup using FIB 3405 * rules. 3406 * **BPF_FIB_LOOKUP_TBID** 3407 * Used with BPF_FIB_LOOKUP_DIRECT. 3408 * Use the routing table ID present in *params*->tbid 3409 * for the fib lookup. 3410 * **BPF_FIB_LOOKUP_OUTPUT** 3411 * Perform lookup from an egress perspective (default is 3412 * ingress). 3413 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3414 * Skip the neighbour table lookup. *params*->dmac 3415 * and *params*->smac will not be set as output. A common 3416 * use case is to call **bpf_redirect_neigh**\ () after 3417 * doing **bpf_fib_lookup**\ (). 3418 * **BPF_FIB_LOOKUP_SRC** 3419 * Derive and set source IP addr in *params*->ipv{4,6}_src 3420 * for the nexthop. If the src addr cannot be derived, 3421 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3422 * case, *params*->dmac and *params*->smac are not set either. 3423 * **BPF_FIB_LOOKUP_MARK** 3424 * Use the mark present in *params*->mark for the fib lookup. 3425 * This option should not be used with BPF_FIB_LOOKUP_DIRECT, 3426 * as it only has meaning for full lookups. 3427 * 3428 * *ctx* is either **struct xdp_md** for XDP programs or 3429 * **struct sk_buff** tc cls_act programs. 3430 * Return 3431 * * < 0 if any input argument is invalid 3432 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3433 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3434 * packet is not forwarded or needs assist from full stack 3435 * 3436 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3437 * was exceeded and output params->mtu_result contains the MTU. 3438 * 3439 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3440 * Description 3441 * Add an entry to, or update a sockhash *map* referencing sockets. 3442 * The *skops* is used as a new value for the entry associated to 3443 * *key*. *flags* is one of: 3444 * 3445 * **BPF_NOEXIST** 3446 * The entry for *key* must not exist in the map. 3447 * **BPF_EXIST** 3448 * The entry for *key* must already exist in the map. 3449 * **BPF_ANY** 3450 * No condition on the existence of the entry for *key*. 3451 * 3452 * If the *map* has eBPF programs (parser and verdict), those will 3453 * be inherited by the socket being added. If the socket is 3454 * already attached to eBPF programs, this results in an error. 3455 * Return 3456 * 0 on success, or a negative error in case of failure. 3457 * 3458 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3459 * Description 3460 * This helper is used in programs implementing policies at the 3461 * socket level. If the message *msg* is allowed to pass (i.e. if 3462 * the verdict eBPF program returns **SK_PASS**), redirect it to 3463 * the socket referenced by *map* (of type 3464 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3465 * egress interfaces can be used for redirection. The 3466 * **BPF_F_INGRESS** value in *flags* is used to make the 3467 * distinction (ingress path is selected if the flag is present, 3468 * egress path otherwise). This is the only flag supported for now. 3469 * Return 3470 * **SK_PASS** on success, or **SK_DROP** on error. 3471 * 3472 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3473 * Description 3474 * This helper is used in programs implementing policies at the 3475 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3476 * if the verdict eBPF program returns **SK_PASS**), redirect it 3477 * to the socket referenced by *map* (of type 3478 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3479 * egress interfaces can be used for redirection. The 3480 * **BPF_F_INGRESS** value in *flags* is used to make the 3481 * distinction (ingress path is selected if the flag is present, 3482 * egress otherwise). This is the only flag supported for now. 3483 * Return 3484 * **SK_PASS** on success, or **SK_DROP** on error. 3485 * 3486 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3487 * Description 3488 * Encapsulate the packet associated to *skb* within a Layer 3 3489 * protocol header. This header is provided in the buffer at 3490 * address *hdr*, with *len* its size in bytes. *type* indicates 3491 * the protocol of the header and can be one of: 3492 * 3493 * **BPF_LWT_ENCAP_SEG6** 3494 * IPv6 encapsulation with Segment Routing Header 3495 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3496 * the IPv6 header is computed by the kernel. 3497 * **BPF_LWT_ENCAP_SEG6_INLINE** 3498 * Only works if *skb* contains an IPv6 packet. Insert a 3499 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3500 * the IPv6 header. 3501 * **BPF_LWT_ENCAP_IP** 3502 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3503 * must be IPv4 or IPv6, followed by zero or more 3504 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3505 * total bytes in all prepended headers. Please note that 3506 * if **skb_is_gso**\ (*skb*) is true, no more than two 3507 * headers can be prepended, and the inner header, if 3508 * present, should be either GRE or UDP/GUE. 3509 * 3510 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3511 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3512 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3513 * **BPF_PROG_TYPE_LWT_XMIT**. 3514 * 3515 * A call to this helper is susceptible to change the underlying 3516 * packet buffer. Therefore, at load time, all checks on pointers 3517 * previously done by the verifier are invalidated and must be 3518 * performed again, if the helper is used in combination with 3519 * direct packet access. 3520 * Return 3521 * 0 on success, or a negative error in case of failure. 3522 * 3523 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3524 * Description 3525 * Store *len* bytes from address *from* into the packet 3526 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3527 * inside the outermost IPv6 Segment Routing Header can be 3528 * modified through this helper. 3529 * 3530 * A call to this helper is susceptible to change the underlying 3531 * packet buffer. Therefore, at load time, all checks on pointers 3532 * previously done by the verifier are invalidated and must be 3533 * performed again, if the helper is used in combination with 3534 * direct packet access. 3535 * Return 3536 * 0 on success, or a negative error in case of failure. 3537 * 3538 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3539 * Description 3540 * Adjust the size allocated to TLVs in the outermost IPv6 3541 * Segment Routing Header contained in the packet associated to 3542 * *skb*, at position *offset* by *delta* bytes. Only offsets 3543 * after the segments are accepted. *delta* can be as well 3544 * positive (growing) as negative (shrinking). 3545 * 3546 * A call to this helper is susceptible to change the underlying 3547 * packet buffer. Therefore, at load time, all checks on pointers 3548 * previously done by the verifier are invalidated and must be 3549 * performed again, if the helper is used in combination with 3550 * direct packet access. 3551 * Return 3552 * 0 on success, or a negative error in case of failure. 3553 * 3554 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3555 * Description 3556 * Apply an IPv6 Segment Routing action of type *action* to the 3557 * packet associated to *skb*. Each action takes a parameter 3558 * contained at address *param*, and of length *param_len* bytes. 3559 * *action* can be one of: 3560 * 3561 * **SEG6_LOCAL_ACTION_END_X** 3562 * End.X action: Endpoint with Layer-3 cross-connect. 3563 * Type of *param*: **struct in6_addr**. 3564 * **SEG6_LOCAL_ACTION_END_T** 3565 * End.T action: Endpoint with specific IPv6 table lookup. 3566 * Type of *param*: **int**. 3567 * **SEG6_LOCAL_ACTION_END_B6** 3568 * End.B6 action: Endpoint bound to an SRv6 policy. 3569 * Type of *param*: **struct ipv6_sr_hdr**. 3570 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3571 * End.B6.Encap action: Endpoint bound to an SRv6 3572 * encapsulation policy. 3573 * Type of *param*: **struct ipv6_sr_hdr**. 3574 * 3575 * A call to this helper is susceptible to change the underlying 3576 * packet buffer. Therefore, at load time, all checks on pointers 3577 * previously done by the verifier are invalidated and must be 3578 * performed again, if the helper is used in combination with 3579 * direct packet access. 3580 * Return 3581 * 0 on success, or a negative error in case of failure. 3582 * 3583 * long bpf_rc_repeat(void *ctx) 3584 * Description 3585 * This helper is used in programs implementing IR decoding, to 3586 * report a successfully decoded repeat key message. This delays 3587 * the generation of a key up event for previously generated 3588 * key down event. 3589 * 3590 * Some IR protocols like NEC have a special IR message for 3591 * repeating last button, for when a button is held down. 3592 * 3593 * The *ctx* should point to the lirc sample as passed into 3594 * the program. 3595 * 3596 * This helper is only available is the kernel was compiled with 3597 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3598 * "**y**". 3599 * Return 3600 * 0 3601 * 3602 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3603 * Description 3604 * This helper is used in programs implementing IR decoding, to 3605 * report a successfully decoded key press with *scancode*, 3606 * *toggle* value in the given *protocol*. The scancode will be 3607 * translated to a keycode using the rc keymap, and reported as 3608 * an input key down event. After a period a key up event is 3609 * generated. This period can be extended by calling either 3610 * **bpf_rc_keydown**\ () again with the same values, or calling 3611 * **bpf_rc_repeat**\ (). 3612 * 3613 * Some protocols include a toggle bit, in case the button was 3614 * released and pressed again between consecutive scancodes. 3615 * 3616 * The *ctx* should point to the lirc sample as passed into 3617 * the program. 3618 * 3619 * The *protocol* is the decoded protocol number (see 3620 * **enum rc_proto** for some predefined values). 3621 * 3622 * This helper is only available is the kernel was compiled with 3623 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3624 * "**y**". 3625 * Return 3626 * 0 3627 * 3628 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3629 * Description 3630 * Return the cgroup v2 id of the socket associated with the *skb*. 3631 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3632 * helper for cgroup v1 by providing a tag resp. identifier that 3633 * can be matched on or used for map lookups e.g. to implement 3634 * policy. The cgroup v2 id of a given path in the hierarchy is 3635 * exposed in user space through the f_handle API in order to get 3636 * to the same 64-bit id. 3637 * 3638 * This helper can be used on TC egress path, but not on ingress, 3639 * and is available only if the kernel was compiled with the 3640 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3641 * Return 3642 * The id is returned or 0 in case the id could not be retrieved. 3643 * 3644 * u64 bpf_get_current_cgroup_id(void) 3645 * Description 3646 * Get the current cgroup id based on the cgroup within which 3647 * the current task is running. 3648 * Return 3649 * A 64-bit integer containing the current cgroup id based 3650 * on the cgroup within which the current task is running. 3651 * 3652 * void *bpf_get_local_storage(void *map, u64 flags) 3653 * Description 3654 * Get the pointer to the local storage area. 3655 * The type and the size of the local storage is defined 3656 * by the *map* argument. 3657 * The *flags* meaning is specific for each map type, 3658 * and has to be 0 for cgroup local storage. 3659 * 3660 * Depending on the BPF program type, a local storage area 3661 * can be shared between multiple instances of the BPF program, 3662 * running simultaneously. 3663 * 3664 * A user should care about the synchronization by himself. 3665 * For example, by using the **BPF_ATOMIC** instructions to alter 3666 * the shared data. 3667 * Return 3668 * A pointer to the local storage area. 3669 * 3670 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3671 * Description 3672 * Select a **SO_REUSEPORT** socket from a 3673 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3674 * It checks the selected socket is matching the incoming 3675 * request in the socket buffer. 3676 * Return 3677 * 0 on success, or a negative error in case of failure. 3678 * 3679 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3680 * Description 3681 * Return id of cgroup v2 that is ancestor of cgroup associated 3682 * with the *skb* at the *ancestor_level*. The root cgroup is at 3683 * *ancestor_level* zero and each step down the hierarchy 3684 * increments the level. If *ancestor_level* == level of cgroup 3685 * associated with *skb*, then return value will be same as that 3686 * of **bpf_skb_cgroup_id**\ (). 3687 * 3688 * The helper is useful to implement policies based on cgroups 3689 * that are upper in hierarchy than immediate cgroup associated 3690 * with *skb*. 3691 * 3692 * The format of returned id and helper limitations are same as in 3693 * **bpf_skb_cgroup_id**\ (). 3694 * Return 3695 * The id is returned or 0 in case the id could not be retrieved. 3696 * 3697 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3698 * Description 3699 * Look for TCP socket matching *tuple*, optionally in a child 3700 * network namespace *netns*. The return value must be checked, 3701 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3702 * 3703 * The *ctx* should point to the context of the program, such as 3704 * the skb or socket (depending on the hook in use). This is used 3705 * to determine the base network namespace for the lookup. 3706 * 3707 * *tuple_size* must be one of: 3708 * 3709 * **sizeof**\ (*tuple*\ **->ipv4**) 3710 * Look for an IPv4 socket. 3711 * **sizeof**\ (*tuple*\ **->ipv6**) 3712 * Look for an IPv6 socket. 3713 * 3714 * If the *netns* is a negative signed 32-bit integer, then the 3715 * socket lookup table in the netns associated with the *ctx* 3716 * will be used. For the TC hooks, this is the netns of the device 3717 * in the skb. For socket hooks, this is the netns of the socket. 3718 * If *netns* is any other signed 32-bit value greater than or 3719 * equal to zero then it specifies the ID of the netns relative to 3720 * the netns associated with the *ctx*. *netns* values beyond the 3721 * range of 32-bit integers are reserved for future use. 3722 * 3723 * All values for *flags* are reserved for future usage, and must 3724 * be left at zero. 3725 * 3726 * This helper is available only if the kernel was compiled with 3727 * **CONFIG_NET** configuration option. 3728 * Return 3729 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3730 * For sockets with reuseport option, the **struct bpf_sock** 3731 * result is from *reuse*\ **->socks**\ [] using the hash of the 3732 * tuple. 3733 * 3734 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3735 * Description 3736 * Look for UDP socket matching *tuple*, optionally in a child 3737 * network namespace *netns*. The return value must be checked, 3738 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3739 * 3740 * The *ctx* should point to the context of the program, such as 3741 * the skb or socket (depending on the hook in use). This is used 3742 * to determine the base network namespace for the lookup. 3743 * 3744 * *tuple_size* must be one of: 3745 * 3746 * **sizeof**\ (*tuple*\ **->ipv4**) 3747 * Look for an IPv4 socket. 3748 * **sizeof**\ (*tuple*\ **->ipv6**) 3749 * Look for an IPv6 socket. 3750 * 3751 * If the *netns* is a negative signed 32-bit integer, then the 3752 * socket lookup table in the netns associated with the *ctx* 3753 * will be used. For the TC hooks, this is the netns of the device 3754 * in the skb. For socket hooks, this is the netns of the socket. 3755 * If *netns* is any other signed 32-bit value greater than or 3756 * equal to zero then it specifies the ID of the netns relative to 3757 * the netns associated with the *ctx*. *netns* values beyond the 3758 * range of 32-bit integers are reserved for future use. 3759 * 3760 * All values for *flags* are reserved for future usage, and must 3761 * be left at zero. 3762 * 3763 * This helper is available only if the kernel was compiled with 3764 * **CONFIG_NET** configuration option. 3765 * Return 3766 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3767 * For sockets with reuseport option, the **struct bpf_sock** 3768 * result is from *reuse*\ **->socks**\ [] using the hash of the 3769 * tuple. 3770 * 3771 * long bpf_sk_release(void *sock) 3772 * Description 3773 * Release the reference held by *sock*. *sock* must be a 3774 * non-**NULL** pointer that was returned from 3775 * **bpf_sk_lookup_xxx**\ (). 3776 * Return 3777 * 0 on success, or a negative error in case of failure. 3778 * 3779 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3780 * Description 3781 * Push an element *value* in *map*. *flags* is one of: 3782 * 3783 * **BPF_EXIST** 3784 * If the queue/stack is full, the oldest element is 3785 * removed to make room for this. 3786 * Return 3787 * 0 on success, or a negative error in case of failure. 3788 * 3789 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3790 * Description 3791 * Pop an element from *map*. 3792 * Return 3793 * 0 on success, or a negative error in case of failure. 3794 * 3795 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3796 * Description 3797 * Get an element from *map* without removing it. 3798 * Return 3799 * 0 on success, or a negative error in case of failure. 3800 * 3801 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3802 * Description 3803 * For socket policies, insert *len* bytes into *msg* at offset 3804 * *start*. 3805 * 3806 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3807 * *msg* it may want to insert metadata or options into the *msg*. 3808 * This can later be read and used by any of the lower layer BPF 3809 * hooks. 3810 * 3811 * This helper may fail if under memory pressure (a malloc 3812 * fails) in these cases BPF programs will get an appropriate 3813 * error and BPF programs will need to handle them. 3814 * Return 3815 * 0 on success, or a negative error in case of failure. 3816 * 3817 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3818 * Description 3819 * Will remove *len* bytes from a *msg* starting at byte *start*. 3820 * This may result in **ENOMEM** errors under certain situations if 3821 * an allocation and copy are required due to a full ring buffer. 3822 * However, the helper will try to avoid doing the allocation 3823 * if possible. Other errors can occur if input parameters are 3824 * invalid either due to *start* byte not being valid part of *msg* 3825 * payload and/or *pop* value being to large. 3826 * Return 3827 * 0 on success, or a negative error in case of failure. 3828 * 3829 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3830 * Description 3831 * This helper is used in programs implementing IR decoding, to 3832 * report a successfully decoded pointer movement. 3833 * 3834 * The *ctx* should point to the lirc sample as passed into 3835 * the program. 3836 * 3837 * This helper is only available is the kernel was compiled with 3838 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3839 * "**y**". 3840 * Return 3841 * 0 3842 * 3843 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3844 * Description 3845 * Acquire a spinlock represented by the pointer *lock*, which is 3846 * stored as part of a value of a map. Taking the lock allows to 3847 * safely update the rest of the fields in that value. The 3848 * spinlock can (and must) later be released with a call to 3849 * **bpf_spin_unlock**\ (\ *lock*\ ). 3850 * 3851 * Spinlocks in BPF programs come with a number of restrictions 3852 * and constraints: 3853 * 3854 * * **bpf_spin_lock** objects are only allowed inside maps of 3855 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3856 * list could be extended in the future). 3857 * * BTF description of the map is mandatory. 3858 * * The BPF program can take ONE lock at a time, since taking two 3859 * or more could cause dead locks. 3860 * * Only one **struct bpf_spin_lock** is allowed per map element. 3861 * * When the lock is taken, calls (either BPF to BPF or helpers) 3862 * are not allowed. 3863 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3864 * allowed inside a spinlock-ed region. 3865 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3866 * the lock, on all execution paths, before it returns. 3867 * * The BPF program can access **struct bpf_spin_lock** only via 3868 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3869 * helpers. Loading or storing data into the **struct 3870 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3871 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3872 * of the map value must be a struct and have **struct 3873 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3874 * Nested lock inside another struct is not allowed. 3875 * * The **struct bpf_spin_lock** *lock* field in a map value must 3876 * be aligned on a multiple of 4 bytes in that value. 3877 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3878 * the **bpf_spin_lock** field to user space. 3879 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3880 * a BPF program, do not update the **bpf_spin_lock** field. 3881 * * **bpf_spin_lock** cannot be on the stack or inside a 3882 * networking packet (it can only be inside of a map values). 3883 * * **bpf_spin_lock** is available to root only. 3884 * * Tracing programs and socket filter programs cannot use 3885 * **bpf_spin_lock**\ () due to insufficient preemption checks 3886 * (but this may change in the future). 3887 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3888 * Return 3889 * 0 3890 * 3891 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3892 * Description 3893 * Release the *lock* previously locked by a call to 3894 * **bpf_spin_lock**\ (\ *lock*\ ). 3895 * Return 3896 * 0 3897 * 3898 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3899 * Description 3900 * This helper gets a **struct bpf_sock** pointer such 3901 * that all the fields in this **bpf_sock** can be accessed. 3902 * Return 3903 * A **struct bpf_sock** pointer on success, or **NULL** in 3904 * case of failure. 3905 * 3906 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3907 * Description 3908 * This helper gets a **struct bpf_tcp_sock** pointer from a 3909 * **struct bpf_sock** pointer. 3910 * Return 3911 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3912 * case of failure. 3913 * 3914 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3915 * Description 3916 * Set ECN (Explicit Congestion Notification) field of IP header 3917 * to **CE** (Congestion Encountered) if current value is **ECT** 3918 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3919 * and IPv4. 3920 * Return 3921 * 1 if the **CE** flag is set (either by the current helper call 3922 * or because it was already present), 0 if it is not set. 3923 * 3924 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3925 * Description 3926 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3927 * **bpf_sk_release**\ () is unnecessary and not allowed. 3928 * Return 3929 * A **struct bpf_sock** pointer on success, or **NULL** in 3930 * case of failure. 3931 * 3932 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3933 * Description 3934 * Look for TCP socket matching *tuple*, optionally in a child 3935 * network namespace *netns*. The return value must be checked, 3936 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3937 * 3938 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3939 * that it also returns timewait or request sockets. Use 3940 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3941 * full structure. 3942 * 3943 * This helper is available only if the kernel was compiled with 3944 * **CONFIG_NET** configuration option. 3945 * Return 3946 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3947 * For sockets with reuseport option, the **struct bpf_sock** 3948 * result is from *reuse*\ **->socks**\ [] using the hash of the 3949 * tuple. 3950 * 3951 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3952 * Description 3953 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3954 * the listening socket in *sk*. 3955 * 3956 * *iph* points to the start of the IPv4 or IPv6 header, while 3957 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3958 * **sizeof**\ (**struct ipv6hdr**). 3959 * 3960 * *th* points to the start of the TCP header, while *th_len* 3961 * contains the length of the TCP header (at least 3962 * **sizeof**\ (**struct tcphdr**)). 3963 * Return 3964 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3965 * error otherwise. 3966 * 3967 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3968 * Description 3969 * Get name of sysctl in /proc/sys/ and copy it into provided by 3970 * program buffer *buf* of size *buf_len*. 3971 * 3972 * The buffer is always NUL terminated, unless it's zero-sized. 3973 * 3974 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3975 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3976 * only (e.g. "tcp_mem"). 3977 * Return 3978 * Number of character copied (not including the trailing NUL). 3979 * 3980 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3981 * truncated name in this case). 3982 * 3983 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3984 * Description 3985 * Get current value of sysctl as it is presented in /proc/sys 3986 * (incl. newline, etc), and copy it as a string into provided 3987 * by program buffer *buf* of size *buf_len*. 3988 * 3989 * The whole value is copied, no matter what file position user 3990 * space issued e.g. sys_read at. 3991 * 3992 * The buffer is always NUL terminated, unless it's zero-sized. 3993 * Return 3994 * Number of character copied (not including the trailing NUL). 3995 * 3996 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3997 * truncated name in this case). 3998 * 3999 * **-EINVAL** if current value was unavailable, e.g. because 4000 * sysctl is uninitialized and read returns -EIO for it. 4001 * 4002 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 4003 * Description 4004 * Get new value being written by user space to sysctl (before 4005 * the actual write happens) and copy it as a string into 4006 * provided by program buffer *buf* of size *buf_len*. 4007 * 4008 * User space may write new value at file position > 0. 4009 * 4010 * The buffer is always NUL terminated, unless it's zero-sized. 4011 * Return 4012 * Number of character copied (not including the trailing NUL). 4013 * 4014 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 4015 * truncated name in this case). 4016 * 4017 * **-EINVAL** if sysctl is being read. 4018 * 4019 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 4020 * Description 4021 * Override new value being written by user space to sysctl with 4022 * value provided by program in buffer *buf* of size *buf_len*. 4023 * 4024 * *buf* should contain a string in same form as provided by user 4025 * space on sysctl write. 4026 * 4027 * User space may write new value at file position > 0. To override 4028 * the whole sysctl value file position should be set to zero. 4029 * Return 4030 * 0 on success. 4031 * 4032 * **-E2BIG** if the *buf_len* is too big. 4033 * 4034 * **-EINVAL** if sysctl is being read. 4035 * 4036 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 4037 * Description 4038 * Convert the initial part of the string from buffer *buf* of 4039 * size *buf_len* to a long integer according to the given base 4040 * and save the result in *res*. 4041 * 4042 * The string may begin with an arbitrary amount of white space 4043 * (as determined by **isspace**\ (3)) followed by a single 4044 * optional '**-**' sign. 4045 * 4046 * Five least significant bits of *flags* encode base, other bits 4047 * are currently unused. 4048 * 4049 * Base must be either 8, 10, 16 or 0 to detect it automatically 4050 * similar to user space **strtol**\ (3). 4051 * Return 4052 * Number of characters consumed on success. Must be positive but 4053 * no more than *buf_len*. 4054 * 4055 * **-EINVAL** if no valid digits were found or unsupported base 4056 * was provided. 4057 * 4058 * **-ERANGE** if resulting value was out of range. 4059 * 4060 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 4061 * Description 4062 * Convert the initial part of the string from buffer *buf* of 4063 * size *buf_len* to an unsigned long integer according to the 4064 * given base and save the result in *res*. 4065 * 4066 * The string may begin with an arbitrary amount of white space 4067 * (as determined by **isspace**\ (3)). 4068 * 4069 * Five least significant bits of *flags* encode base, other bits 4070 * are currently unused. 4071 * 4072 * Base must be either 8, 10, 16 or 0 to detect it automatically 4073 * similar to user space **strtoul**\ (3). 4074 * Return 4075 * Number of characters consumed on success. Must be positive but 4076 * no more than *buf_len*. 4077 * 4078 * **-EINVAL** if no valid digits were found or unsupported base 4079 * was provided. 4080 * 4081 * **-ERANGE** if resulting value was out of range. 4082 * 4083 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4084 * Description 4085 * Get a bpf-local-storage from a *sk*. 4086 * 4087 * Logically, it could be thought of getting the value from 4088 * a *map* with *sk* as the **key**. From this 4089 * perspective, the usage is not much different from 4090 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4091 * helper enforces the key must be a full socket and the map must 4092 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4093 * 4094 * Underneath, the value is stored locally at *sk* instead of 4095 * the *map*. The *map* is used as the bpf-local-storage 4096 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4097 * searched against all bpf-local-storages residing at *sk*. 4098 * 4099 * *sk* is a kernel **struct sock** pointer for LSM program. 4100 * *sk* is a **struct bpf_sock** pointer for other program types. 4101 * 4102 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4103 * used such that a new bpf-local-storage will be 4104 * created if one does not exist. *value* can be used 4105 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4106 * the initial value of a bpf-local-storage. If *value* is 4107 * **NULL**, the new bpf-local-storage will be zero initialized. 4108 * Return 4109 * A bpf-local-storage pointer is returned on success. 4110 * 4111 * **NULL** if not found or there was an error in adding 4112 * a new bpf-local-storage. 4113 * 4114 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4115 * Description 4116 * Delete a bpf-local-storage from a *sk*. 4117 * Return 4118 * 0 on success. 4119 * 4120 * **-ENOENT** if the bpf-local-storage cannot be found. 4121 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4122 * 4123 * long bpf_send_signal(u32 sig) 4124 * Description 4125 * Send signal *sig* to the process of the current task. 4126 * The signal may be delivered to any of this process's threads. 4127 * Return 4128 * 0 on success or successfully queued. 4129 * 4130 * **-EBUSY** if work queue under nmi is full. 4131 * 4132 * **-EINVAL** if *sig* is invalid. 4133 * 4134 * **-EPERM** if no permission to send the *sig*. 4135 * 4136 * **-EAGAIN** if bpf program can try again. 4137 * 4138 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4139 * Description 4140 * Try to issue a SYN cookie for the packet with corresponding 4141 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4142 * 4143 * *iph* points to the start of the IPv4 or IPv6 header, while 4144 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4145 * **sizeof**\ (**struct ipv6hdr**). 4146 * 4147 * *th* points to the start of the TCP header, while *th_len* 4148 * contains the length of the TCP header with options (at least 4149 * **sizeof**\ (**struct tcphdr**)). 4150 * Return 4151 * On success, lower 32 bits hold the generated SYN cookie in 4152 * followed by 16 bits which hold the MSS value for that cookie, 4153 * and the top 16 bits are unused. 4154 * 4155 * On failure, the returned value is one of the following: 4156 * 4157 * **-EINVAL** SYN cookie cannot be issued due to error 4158 * 4159 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4160 * 4161 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4162 * 4163 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4164 * 4165 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4166 * Description 4167 * Write raw *data* blob into a special BPF perf event held by 4168 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4169 * event must have the following attributes: **PERF_SAMPLE_RAW** 4170 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4171 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4172 * 4173 * The *flags* are used to indicate the index in *map* for which 4174 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4175 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4176 * to indicate that the index of the current CPU core should be 4177 * used. 4178 * 4179 * The value to write, of *size*, is passed through eBPF stack and 4180 * pointed by *data*. 4181 * 4182 * *ctx* is a pointer to in-kernel struct sk_buff. 4183 * 4184 * This helper is similar to **bpf_perf_event_output**\ () but 4185 * restricted to raw_tracepoint bpf programs. 4186 * Return 4187 * 0 on success, or a negative error in case of failure. 4188 * 4189 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4190 * Description 4191 * Safely attempt to read *size* bytes from user 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_kernel(void *dst, u32 size, const void *unsafe_ptr) 4197 * Description 4198 * Safely attempt to read *size* bytes from kernel space address 4199 * *unsafe_ptr* and store the data in *dst*. 4200 * Return 4201 * 0 on success, or a negative error in case of failure. 4202 * 4203 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4204 * Description 4205 * Copy a NUL terminated string from an unsafe user address 4206 * *unsafe_ptr* to *dst*. The *size* should include the 4207 * terminating NUL byte. In case the string length is smaller than 4208 * *size*, the target is not padded with further NUL bytes. If the 4209 * string length is larger than *size*, just *size*-1 bytes are 4210 * copied and the last byte is set to NUL. 4211 * 4212 * On success, returns the number of bytes that were written, 4213 * including the terminal NUL. This makes this helper useful in 4214 * tracing programs for reading strings, and more importantly to 4215 * get its length at runtime. See the following snippet: 4216 * 4217 * :: 4218 * 4219 * SEC("kprobe/sys_open") 4220 * void bpf_sys_open(struct pt_regs *ctx) 4221 * { 4222 * char buf[PATHLEN]; // PATHLEN is defined to 256 4223 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4224 * ctx->di); 4225 * 4226 * // Consume buf, for example push it to 4227 * // userspace via bpf_perf_event_output(); we 4228 * // can use res (the string length) as event 4229 * // size, after checking its boundaries. 4230 * } 4231 * 4232 * In comparison, using **bpf_probe_read_user**\ () helper here 4233 * instead to read the string would require to estimate the length 4234 * at compile time, and would often result in copying more memory 4235 * than necessary. 4236 * 4237 * Another useful use case is when parsing individual process 4238 * arguments or individual environment variables navigating 4239 * *current*\ **->mm->arg_start** and *current*\ 4240 * **->mm->env_start**: using this helper and the return value, 4241 * one can quickly iterate at the right offset of the memory area. 4242 * Return 4243 * On success, the strictly positive length of the output string, 4244 * including the trailing NUL character. On error, a negative 4245 * value. 4246 * 4247 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4248 * Description 4249 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4250 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4251 * Return 4252 * On success, the strictly positive length of the string, including 4253 * the trailing NUL character. On error, a negative value. 4254 * 4255 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4256 * Description 4257 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4258 * *rcv_nxt* is the ack_seq to be sent out. 4259 * Return 4260 * 0 on success, or a negative error in case of failure. 4261 * 4262 * long bpf_send_signal_thread(u32 sig) 4263 * Description 4264 * Send signal *sig* to the thread corresponding to the current task. 4265 * Return 4266 * 0 on success or successfully queued. 4267 * 4268 * **-EBUSY** if work queue under nmi is full. 4269 * 4270 * **-EINVAL** if *sig* is invalid. 4271 * 4272 * **-EPERM** if no permission to send the *sig*. 4273 * 4274 * **-EAGAIN** if bpf program can try again. 4275 * 4276 * u64 bpf_jiffies64(void) 4277 * Description 4278 * Obtain the 64bit jiffies 4279 * Return 4280 * The 64 bit jiffies 4281 * 4282 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4283 * Description 4284 * For an eBPF program attached to a perf event, retrieve the 4285 * branch records (**struct perf_branch_entry**) associated to *ctx* 4286 * and store it in the buffer pointed by *buf* up to size 4287 * *size* bytes. 4288 * Return 4289 * On success, number of bytes written to *buf*. On error, a 4290 * negative value. 4291 * 4292 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4293 * instead return the number of bytes required to store all the 4294 * branch entries. If this flag is set, *buf* may be NULL. 4295 * 4296 * **-EINVAL** if arguments invalid or **size** not a multiple 4297 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4298 * 4299 * **-ENOENT** if architecture does not support branch records. 4300 * 4301 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4302 * Description 4303 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4304 * *namespace* will be returned in *nsdata*. 4305 * Return 4306 * 0 on success, or one of the following in case of failure: 4307 * 4308 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4309 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4310 * 4311 * **-ENOENT** if pidns does not exists for the current task. 4312 * 4313 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4314 * Description 4315 * Write raw *data* blob into a special BPF perf event held by 4316 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4317 * event must have the following attributes: **PERF_SAMPLE_RAW** 4318 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4319 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4320 * 4321 * The *flags* are used to indicate the index in *map* for which 4322 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4323 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4324 * to indicate that the index of the current CPU core should be 4325 * used. 4326 * 4327 * The value to write, of *size*, is passed through eBPF stack and 4328 * pointed by *data*. 4329 * 4330 * *ctx* is a pointer to in-kernel struct xdp_buff. 4331 * 4332 * This helper is similar to **bpf_perf_eventoutput**\ () but 4333 * restricted to raw_tracepoint bpf programs. 4334 * Return 4335 * 0 on success, or a negative error in case of failure. 4336 * 4337 * u64 bpf_get_netns_cookie(void *ctx) 4338 * Description 4339 * Retrieve the cookie (generated by the kernel) of the network 4340 * namespace the input *ctx* is associated with. The network 4341 * namespace cookie remains stable for its lifetime and provides 4342 * a global identifier that can be assumed unique. If *ctx* is 4343 * NULL, then the helper returns the cookie for the initial 4344 * network namespace. The cookie itself is very similar to that 4345 * of **bpf_get_socket_cookie**\ () helper, but for network 4346 * namespaces instead of sockets. 4347 * Return 4348 * A 8-byte long opaque number. 4349 * 4350 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4351 * Description 4352 * Return id of cgroup v2 that is ancestor of the cgroup associated 4353 * with the current task at the *ancestor_level*. The root cgroup 4354 * is at *ancestor_level* zero and each step down the hierarchy 4355 * increments the level. If *ancestor_level* == level of cgroup 4356 * associated with the current task, then return value will be the 4357 * same as that of **bpf_get_current_cgroup_id**\ (). 4358 * 4359 * The helper is useful to implement policies based on cgroups 4360 * that are upper in hierarchy than immediate cgroup associated 4361 * with the current task. 4362 * 4363 * The format of returned id and helper limitations are same as in 4364 * **bpf_get_current_cgroup_id**\ (). 4365 * Return 4366 * The id is returned or 0 in case the id could not be retrieved. 4367 * 4368 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4369 * Description 4370 * Helper is overloaded depending on BPF program type. This 4371 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4372 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4373 * 4374 * Assign the *sk* to the *skb*. When combined with appropriate 4375 * routing configuration to receive the packet towards the socket, 4376 * will cause *skb* to be delivered to the specified socket. 4377 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4378 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4379 * interfere with successful delivery to the socket. 4380 * 4381 * This operation is only valid from TC ingress path. 4382 * 4383 * The *flags* argument must be zero. 4384 * Return 4385 * 0 on success, or a negative error in case of failure: 4386 * 4387 * **-EINVAL** if specified *flags* are not supported. 4388 * 4389 * **-ENOENT** if the socket is unavailable for assignment. 4390 * 4391 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4392 * 4393 * **-EOPNOTSUPP** if the operation is not supported, for example 4394 * a call from outside of TC ingress. 4395 * 4396 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4397 * Description 4398 * Helper is overloaded depending on BPF program type. This 4399 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4400 * 4401 * Select the *sk* as a result of a socket lookup. 4402 * 4403 * For the operation to succeed passed socket must be compatible 4404 * with the packet description provided by the *ctx* object. 4405 * 4406 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4407 * be an exact match. While IP family (**AF_INET** or 4408 * **AF_INET6**) must be compatible, that is IPv6 sockets 4409 * that are not v6-only can be selected for IPv4 packets. 4410 * 4411 * Only TCP listeners and UDP unconnected sockets can be 4412 * selected. *sk* can also be NULL to reset any previous 4413 * selection. 4414 * 4415 * *flags* argument can combination of following values: 4416 * 4417 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4418 * socket selection, potentially done by a BPF program 4419 * that ran before us. 4420 * 4421 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4422 * load-balancing within reuseport group for the socket 4423 * being selected. 4424 * 4425 * On success *ctx->sk* will point to the selected socket. 4426 * 4427 * Return 4428 * 0 on success, or a negative errno in case of failure. 4429 * 4430 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4431 * not compatible with packet family (*ctx->family*). 4432 * 4433 * * **-EEXIST** if socket has been already selected, 4434 * potentially by another program, and 4435 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4436 * 4437 * * **-EINVAL** if unsupported flags were specified. 4438 * 4439 * * **-EPROTOTYPE** if socket L4 protocol 4440 * (*sk->protocol*) doesn't match packet protocol 4441 * (*ctx->protocol*). 4442 * 4443 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4444 * state (TCP listening or UDP unconnected). 4445 * 4446 * u64 bpf_ktime_get_boot_ns(void) 4447 * Description 4448 * Return the time elapsed since system boot, in nanoseconds. 4449 * Does include the time the system was suspended. 4450 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4451 * Return 4452 * Current *ktime*. 4453 * 4454 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4455 * Description 4456 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4457 * out the format string. 4458 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4459 * the format string itself. The *data* and *data_len* are format string 4460 * arguments. The *data* are a **u64** array and corresponding format string 4461 * values are stored in the array. For strings and pointers where pointees 4462 * are accessed, only the pointer values are stored in the *data* array. 4463 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4464 * 4465 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4466 * Reading kernel memory may fail due to either invalid address or 4467 * valid address but requiring a major memory fault. If reading kernel memory 4468 * fails, the string for **%s** will be an empty string, and the ip 4469 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4470 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4471 * Return 4472 * 0 on success, or a negative error in case of failure: 4473 * 4474 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4475 * by returning 1 from bpf program. 4476 * 4477 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4478 * 4479 * **-E2BIG** if *fmt* contains too many format specifiers. 4480 * 4481 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4482 * 4483 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4484 * Description 4485 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4486 * The *m* represents the seq_file. The *data* and *len* represent the 4487 * data to write in bytes. 4488 * Return 4489 * 0 on success, or a negative error in case of failure: 4490 * 4491 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4492 * 4493 * u64 bpf_sk_cgroup_id(void *sk) 4494 * Description 4495 * Return the cgroup v2 id of the socket *sk*. 4496 * 4497 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4498 * returned from **bpf_sk_lookup_xxx**\ (), 4499 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4500 * same as in **bpf_skb_cgroup_id**\ (). 4501 * 4502 * This helper is available only if the kernel was compiled with 4503 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4504 * Return 4505 * The id is returned or 0 in case the id could not be retrieved. 4506 * 4507 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4508 * Description 4509 * Return id of cgroup v2 that is ancestor of cgroup associated 4510 * with the *sk* at the *ancestor_level*. The root cgroup is at 4511 * *ancestor_level* zero and each step down the hierarchy 4512 * increments the level. If *ancestor_level* == level of cgroup 4513 * associated with *sk*, then return value will be same as that 4514 * of **bpf_sk_cgroup_id**\ (). 4515 * 4516 * The helper is useful to implement policies based on cgroups 4517 * that are upper in hierarchy than immediate cgroup associated 4518 * with *sk*. 4519 * 4520 * The format of returned id and helper limitations are same as in 4521 * **bpf_sk_cgroup_id**\ (). 4522 * Return 4523 * The id is returned or 0 in case the id could not be retrieved. 4524 * 4525 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4526 * Description 4527 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4528 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4529 * of new data availability is sent. 4530 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4531 * of new data availability is sent unconditionally. 4532 * If **0** is specified in *flags*, an adaptive notification 4533 * of new data availability is sent. 4534 * 4535 * An adaptive notification is a notification sent whenever the user-space 4536 * process has caught up and consumed all available payloads. In case the user-space 4537 * process is still processing a previous payload, then no notification is needed 4538 * as it will process the newly added payload automatically. 4539 * Return 4540 * 0 on success, or a negative error in case of failure. 4541 * 4542 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4543 * Description 4544 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4545 * *flags* must be 0. 4546 * Return 4547 * Valid pointer with *size* bytes of memory available; NULL, 4548 * otherwise. 4549 * 4550 * void bpf_ringbuf_submit(void *data, u64 flags) 4551 * Description 4552 * Submit reserved ring buffer sample, pointed to by *data*. 4553 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4554 * of new data availability is sent. 4555 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4556 * of new data availability is sent unconditionally. 4557 * If **0** is specified in *flags*, an adaptive notification 4558 * of new data availability is sent. 4559 * 4560 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4561 * Return 4562 * Nothing. Always succeeds. 4563 * 4564 * void bpf_ringbuf_discard(void *data, u64 flags) 4565 * Description 4566 * Discard reserved ring buffer sample, pointed to by *data*. 4567 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4568 * of new data availability is sent. 4569 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4570 * of new data availability is sent unconditionally. 4571 * If **0** is specified in *flags*, an adaptive notification 4572 * of new data availability is sent. 4573 * 4574 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4575 * Return 4576 * Nothing. Always succeeds. 4577 * 4578 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4579 * Description 4580 * Query various characteristics of provided ring buffer. What 4581 * exactly is queries is determined by *flags*: 4582 * 4583 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4584 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4585 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4586 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4587 * 4588 * Data returned is just a momentary snapshot of actual values 4589 * and could be inaccurate, so this facility should be used to 4590 * power heuristics and for reporting, not to make 100% correct 4591 * calculation. 4592 * Return 4593 * Requested value, or 0, if *flags* are not recognized. 4594 * 4595 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4596 * Description 4597 * Change the skbs checksum level by one layer up or down, or 4598 * reset it entirely to none in order to have the stack perform 4599 * checksum validation. The level is applicable to the following 4600 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4601 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4602 * through **bpf_skb_adjust_room**\ () helper with passing in 4603 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4604 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4605 * the UDP header is removed. Similarly, an encap of the latter 4606 * into the former could be accompanied by a helper call to 4607 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4608 * skb is still intended to be processed in higher layers of the 4609 * stack instead of just egressing at tc. 4610 * 4611 * There are three supported level settings at this time: 4612 * 4613 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4614 * with CHECKSUM_UNNECESSARY. 4615 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4616 * with CHECKSUM_UNNECESSARY. 4617 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4618 * sets CHECKSUM_NONE to force checksum validation by the stack. 4619 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4620 * skb->csum_level. 4621 * Return 4622 * 0 on success, or a negative error in case of failure. In the 4623 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4624 * is returned or the error code -EACCES in case the skb is not 4625 * subject to CHECKSUM_UNNECESSARY. 4626 * 4627 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4628 * Description 4629 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4630 * Return 4631 * *sk* if casting is valid, or **NULL** otherwise. 4632 * 4633 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4634 * Description 4635 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4636 * Return 4637 * *sk* if casting is valid, or **NULL** otherwise. 4638 * 4639 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4640 * Description 4641 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4642 * Return 4643 * *sk* if casting is valid, or **NULL** otherwise. 4644 * 4645 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4646 * Description 4647 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4648 * Return 4649 * *sk* if casting is valid, or **NULL** otherwise. 4650 * 4651 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4652 * Description 4653 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4654 * Return 4655 * *sk* if casting is valid, or **NULL** otherwise. 4656 * 4657 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4658 * Description 4659 * Return a user or a kernel stack in bpf program provided buffer. 4660 * Note: the user stack will only be populated if the *task* is 4661 * the current task; all other tasks will return -EOPNOTSUPP. 4662 * To achieve this, the helper needs *task*, which is a valid 4663 * pointer to **struct task_struct**. To store the stacktrace, the 4664 * bpf program provides *buf* with a nonnegative *size*. 4665 * 4666 * The last argument, *flags*, holds the number of stack frames to 4667 * skip (from 0 to 255), masked with 4668 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4669 * the following flags: 4670 * 4671 * **BPF_F_USER_STACK** 4672 * Collect a user space stack instead of a kernel stack. 4673 * The *task* must be the current task. 4674 * **BPF_F_USER_BUILD_ID** 4675 * Collect buildid+offset instead of ips for user stack, 4676 * only valid if **BPF_F_USER_STACK** is also specified. 4677 * 4678 * **bpf_get_task_stack**\ () can collect up to 4679 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4680 * to sufficient large buffer size. Note that 4681 * this limit can be controlled with the **sysctl** program, and 4682 * that it should be manually increased in order to profile long 4683 * user stacks (such as stacks for Java programs). To do so, use: 4684 * 4685 * :: 4686 * 4687 * # sysctl kernel.perf_event_max_stack=<new value> 4688 * Return 4689 * The non-negative copied *buf* length equal to or less than 4690 * *size* on success, or a negative error in case of failure. 4691 * 4692 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4693 * Description 4694 * Load header option. Support reading a particular TCP header 4695 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4696 * 4697 * If *flags* is 0, it will search the option from the 4698 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4699 * has details on what skb_data contains under different 4700 * *skops*\ **->op**. 4701 * 4702 * The first byte of the *searchby_res* specifies the 4703 * kind that it wants to search. 4704 * 4705 * If the searching kind is an experimental kind 4706 * (i.e. 253 or 254 according to RFC6994). It also 4707 * needs to specify the "magic" which is either 4708 * 2 bytes or 4 bytes. It then also needs to 4709 * specify the size of the magic by using 4710 * the 2nd byte which is "kind-length" of a TCP 4711 * header option and the "kind-length" also 4712 * includes the first 2 bytes "kind" and "kind-length" 4713 * itself as a normal TCP header option also does. 4714 * 4715 * For example, to search experimental kind 254 with 4716 * 2 byte magic 0xeB9F, the searchby_res should be 4717 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4718 * 4719 * To search for the standard window scale option (3), 4720 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4721 * Note, kind-length must be 0 for regular option. 4722 * 4723 * Searching for No-Op (0) and End-of-Option-List (1) are 4724 * not supported. 4725 * 4726 * *len* must be at least 2 bytes which is the minimal size 4727 * of a header option. 4728 * 4729 * Supported flags: 4730 * 4731 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4732 * saved_syn packet or the just-received syn packet. 4733 * 4734 * Return 4735 * > 0 when found, the header option is copied to *searchby_res*. 4736 * The return value is the total length copied. On failure, a 4737 * negative error code is returned: 4738 * 4739 * **-EINVAL** if a parameter is invalid. 4740 * 4741 * **-ENOMSG** if the option is not found. 4742 * 4743 * **-ENOENT** if no syn packet is available when 4744 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4745 * 4746 * **-ENOSPC** if there is not enough space. Only *len* number of 4747 * bytes are copied. 4748 * 4749 * **-EFAULT** on failure to parse the header options in the 4750 * packet. 4751 * 4752 * **-EPERM** if the helper cannot be used under the current 4753 * *skops*\ **->op**. 4754 * 4755 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4756 * Description 4757 * Store header option. The data will be copied 4758 * from buffer *from* with length *len* to the TCP header. 4759 * 4760 * The buffer *from* should have the whole option that 4761 * includes the kind, kind-length, and the actual 4762 * option data. The *len* must be at least kind-length 4763 * long. The kind-length does not have to be 4 byte 4764 * aligned. The kernel will take care of the padding 4765 * and setting the 4 bytes aligned value to th->doff. 4766 * 4767 * This helper will check for duplicated option 4768 * by searching the same option in the outgoing skb. 4769 * 4770 * This helper can only be called during 4771 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4772 * 4773 * Return 4774 * 0 on success, or negative error in case of failure: 4775 * 4776 * **-EINVAL** If param is invalid. 4777 * 4778 * **-ENOSPC** if there is not enough space in the header. 4779 * Nothing has been written 4780 * 4781 * **-EEXIST** if the option already exists. 4782 * 4783 * **-EFAULT** on failure to parse the existing header options. 4784 * 4785 * **-EPERM** if the helper cannot be used under the current 4786 * *skops*\ **->op**. 4787 * 4788 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4789 * Description 4790 * Reserve *len* bytes for the bpf header option. The 4791 * space will be used by **bpf_store_hdr_opt**\ () later in 4792 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4793 * 4794 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4795 * the total number of bytes will be reserved. 4796 * 4797 * This helper can only be called during 4798 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4799 * 4800 * Return 4801 * 0 on success, or negative error in case of failure: 4802 * 4803 * **-EINVAL** if a parameter is invalid. 4804 * 4805 * **-ENOSPC** if there is not enough space in the header. 4806 * 4807 * **-EPERM** if the helper cannot be used under the current 4808 * *skops*\ **->op**. 4809 * 4810 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4811 * Description 4812 * Get a bpf_local_storage from an *inode*. 4813 * 4814 * Logically, it could be thought of as getting the value from 4815 * a *map* with *inode* as the **key**. From this 4816 * perspective, the usage is not much different from 4817 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4818 * helper enforces the key must be an inode and the map must also 4819 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4820 * 4821 * Underneath, the value is stored locally at *inode* instead of 4822 * the *map*. The *map* is used as the bpf-local-storage 4823 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4824 * searched against all bpf_local_storage residing at *inode*. 4825 * 4826 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4827 * used such that a new bpf_local_storage will be 4828 * created if one does not exist. *value* can be used 4829 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4830 * the initial value of a bpf_local_storage. If *value* is 4831 * **NULL**, the new bpf_local_storage will be zero initialized. 4832 * Return 4833 * A bpf_local_storage pointer is returned on success. 4834 * 4835 * **NULL** if not found or there was an error in adding 4836 * a new bpf_local_storage. 4837 * 4838 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4839 * Description 4840 * Delete a bpf_local_storage from an *inode*. 4841 * Return 4842 * 0 on success. 4843 * 4844 * **-ENOENT** if the bpf_local_storage cannot be found. 4845 * 4846 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4847 * Description 4848 * Return full path for given **struct path** object, which 4849 * needs to be the kernel BTF *path* object. The path is 4850 * returned in the provided buffer *buf* of size *sz* and 4851 * is zero terminated. 4852 * 4853 * Return 4854 * On success, the strictly positive length of the string, 4855 * including the trailing NUL character. On error, a negative 4856 * value. 4857 * 4858 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4859 * Description 4860 * Read *size* bytes from user space address *user_ptr* and store 4861 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4862 * Return 4863 * 0 on success, or a negative error in case of failure. 4864 * 4865 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4866 * Description 4867 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4868 * using *ptr*->type_id. This value should specify the type 4869 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4870 * can be used to look up vmlinux BTF type ids. Traversing the 4871 * data structure using BTF, the type information and values are 4872 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4873 * the pointer data is carried out to avoid kernel crashes during 4874 * operation. Smaller types can use string space on the stack; 4875 * larger programs can use map data to store the string 4876 * representation. 4877 * 4878 * The string can be subsequently shared with userspace via 4879 * bpf_perf_event_output() or ring buffer interfaces. 4880 * bpf_trace_printk() is to be avoided as it places too small 4881 * a limit on string size to be useful. 4882 * 4883 * *flags* is a combination of 4884 * 4885 * **BTF_F_COMPACT** 4886 * no formatting around type information 4887 * **BTF_F_NONAME** 4888 * no struct/union member names/types 4889 * **BTF_F_PTR_RAW** 4890 * show raw (unobfuscated) pointer values; 4891 * equivalent to printk specifier %px. 4892 * **BTF_F_ZERO** 4893 * show zero-valued struct/union members; they 4894 * are not displayed by default 4895 * 4896 * Return 4897 * The number of bytes that were written (or would have been 4898 * written if output had to be truncated due to string size), 4899 * or a negative error in cases of failure. 4900 * 4901 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4902 * Description 4903 * Use BTF to write to seq_write a string representation of 4904 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4905 * *flags* are identical to those used for bpf_snprintf_btf. 4906 * Return 4907 * 0 on success or a negative error in case of failure. 4908 * 4909 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4910 * Description 4911 * See **bpf_get_cgroup_classid**\ () for the main description. 4912 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4913 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4914 * associated socket instead of the current process. 4915 * Return 4916 * The id is returned or 0 in case the id could not be retrieved. 4917 * 4918 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4919 * Description 4920 * Redirect the packet to another net device of index *ifindex* 4921 * and fill in L2 addresses from neighboring subsystem. This helper 4922 * is somewhat similar to **bpf_redirect**\ (), except that it 4923 * populates L2 addresses as well, meaning, internally, the helper 4924 * relies on the neighbor lookup for the L2 address of the nexthop. 4925 * 4926 * The helper will perform a FIB lookup based on the skb's 4927 * networking header to get the address of the next hop, unless 4928 * this is supplied by the caller in the *params* argument. The 4929 * *plen* argument indicates the len of *params* and should be set 4930 * to 0 if *params* is NULL. 4931 * 4932 * The *flags* argument is reserved and must be 0. The helper is 4933 * currently only supported for tc BPF program types, and enabled 4934 * for IPv4 and IPv6 protocols. 4935 * Return 4936 * The helper returns **TC_ACT_REDIRECT** on success or 4937 * **TC_ACT_SHOT** on error. 4938 * 4939 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4940 * Description 4941 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4942 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4943 * extern variable decorated with '__ksym'. For ksym, there is a 4944 * global var (either static or global) defined of the same name 4945 * in the kernel. The ksym is percpu if the global var is percpu. 4946 * The returned pointer points to the global percpu var on *cpu*. 4947 * 4948 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4949 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4950 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4951 * bpf_per_cpu_ptr() must check the returned value. 4952 * Return 4953 * A pointer pointing to the kernel percpu variable on *cpu*, or 4954 * NULL, if *cpu* is invalid. 4955 * 4956 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4957 * Description 4958 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4959 * pointer to the percpu kernel variable on this cpu. See the 4960 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4961 * 4962 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4963 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4964 * never return NULL. 4965 * Return 4966 * A pointer pointing to the kernel percpu variable on this cpu. 4967 * 4968 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4969 * Description 4970 * Redirect the packet to another net device of index *ifindex*. 4971 * This helper is somewhat similar to **bpf_redirect**\ (), except 4972 * that the redirection happens to the *ifindex*' peer device and 4973 * the netns switch takes place from ingress to ingress without 4974 * going through the CPU's backlog queue. 4975 * 4976 * *skb*\ **->mark** and *skb*\ **->tstamp** are not cleared during 4977 * the netns switch. 4978 * 4979 * The *flags* argument is reserved and must be 0. The helper is 4980 * currently only supported for tc BPF program types at the 4981 * ingress hook and for veth and netkit target device types. The 4982 * peer device must reside in a different network namespace. 4983 * Return 4984 * The helper returns **TC_ACT_REDIRECT** on success or 4985 * **TC_ACT_SHOT** on error. 4986 * 4987 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4988 * Description 4989 * Get a bpf_local_storage from the *task*. 4990 * 4991 * Logically, it could be thought of as getting the value from 4992 * a *map* with *task* as the **key**. From this 4993 * perspective, the usage is not much different from 4994 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4995 * helper enforces the key must be a task_struct and the map must also 4996 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4997 * 4998 * Underneath, the value is stored locally at *task* instead of 4999 * the *map*. The *map* is used as the bpf-local-storage 5000 * "type". The bpf-local-storage "type" (i.e. the *map*) is 5001 * searched against all bpf_local_storage residing at *task*. 5002 * 5003 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5004 * used such that a new bpf_local_storage will be 5005 * created if one does not exist. *value* can be used 5006 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5007 * the initial value of a bpf_local_storage. If *value* is 5008 * **NULL**, the new bpf_local_storage will be zero initialized. 5009 * Return 5010 * A bpf_local_storage pointer is returned on success. 5011 * 5012 * **NULL** if not found or there was an error in adding 5013 * a new bpf_local_storage. 5014 * 5015 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 5016 * Description 5017 * Delete a bpf_local_storage from a *task*. 5018 * Return 5019 * 0 on success. 5020 * 5021 * **-ENOENT** if the bpf_local_storage cannot be found. 5022 * 5023 * struct task_struct *bpf_get_current_task_btf(void) 5024 * Description 5025 * Return a BTF pointer to the "current" task. 5026 * This pointer can also be used in helpers that accept an 5027 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 5028 * Return 5029 * Pointer to the current task. 5030 * 5031 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 5032 * Description 5033 * Set or clear certain options on *bprm*: 5034 * 5035 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 5036 * which sets the **AT_SECURE** auxv for glibc. The bit 5037 * is cleared if the flag is not specified. 5038 * Return 5039 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 5040 * 5041 * u64 bpf_ktime_get_coarse_ns(void) 5042 * Description 5043 * Return a coarse-grained version of the time elapsed since 5044 * system boot, in nanoseconds. Does not include time the system 5045 * was suspended. 5046 * 5047 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 5048 * Return 5049 * Current *ktime*. 5050 * 5051 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 5052 * Description 5053 * Returns the stored IMA hash of the *inode* (if it's available). 5054 * If the hash is larger than *size*, then only *size* 5055 * bytes will be copied to *dst* 5056 * Return 5057 * The **hash_algo** is returned on success, 5058 * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if 5059 * invalid arguments are passed. 5060 * 5061 * struct socket *bpf_sock_from_file(struct file *file) 5062 * Description 5063 * If the given file represents a socket, returns the associated 5064 * socket. 5065 * Return 5066 * A pointer to a struct socket on success or NULL if the file is 5067 * not a socket. 5068 * 5069 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 5070 * Description 5071 * Check packet size against exceeding MTU of net device (based 5072 * on *ifindex*). This helper will likely be used in combination 5073 * with helpers that adjust/change the packet size. 5074 * 5075 * The argument *len_diff* can be used for querying with a planned 5076 * size change. This allows to check MTU prior to changing packet 5077 * ctx. Providing a *len_diff* adjustment that is larger than the 5078 * actual packet size (resulting in negative packet size) will in 5079 * principle not exceed the MTU, which is why it is not considered 5080 * a failure. Other BPF helpers are needed for performing the 5081 * planned size change; therefore the responsibility for catching 5082 * a negative packet size belongs in those helpers. 5083 * 5084 * Specifying *ifindex* zero means the MTU check is performed 5085 * against the current net device. This is practical if this isn't 5086 * used prior to redirect. 5087 * 5088 * On input *mtu_len* must be a valid pointer, else verifier will 5089 * reject BPF program. If the value *mtu_len* is initialized to 5090 * zero then the ctx packet size is use. When value *mtu_len* is 5091 * provided as input this specify the L3 length that the MTU check 5092 * is done against. Remember XDP and TC length operate at L2, but 5093 * this value is L3 as this correlate to MTU and IP-header tot_len 5094 * values which are L3 (similar behavior as bpf_fib_lookup). 5095 * 5096 * The Linux kernel route table can configure MTUs on a more 5097 * specific per route level, which is not provided by this helper. 5098 * For route level MTU checks use the **bpf_fib_lookup**\ () 5099 * helper. 5100 * 5101 * *ctx* is either **struct xdp_md** for XDP programs or 5102 * **struct sk_buff** for tc cls_act programs. 5103 * 5104 * The *flags* argument can be a combination of one or more of the 5105 * following values: 5106 * 5107 * **BPF_MTU_CHK_SEGS** 5108 * This flag will only works for *ctx* **struct sk_buff**. 5109 * If packet context contains extra packet segment buffers 5110 * (often knows as GSO skb), then MTU check is harder to 5111 * check at this point, because in transmit path it is 5112 * possible for the skb packet to get re-segmented 5113 * (depending on net device features). This could still be 5114 * a MTU violation, so this flag enables performing MTU 5115 * check against segments, with a different violation 5116 * return code to tell it apart. Check cannot use len_diff. 5117 * 5118 * On return *mtu_len* pointer contains the MTU value of the net 5119 * device. Remember the net device configured MTU is the L3 size, 5120 * which is returned here and XDP and TC length operate at L2. 5121 * Helper take this into account for you, but remember when using 5122 * MTU value in your BPF-code. 5123 * 5124 * Return 5125 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5126 * 5127 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5128 * 5129 * MTU violations return positive values, but also populate MTU 5130 * value in *mtu_len* pointer, as this can be needed for 5131 * implementing PMTU handing: 5132 * 5133 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5134 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5135 * 5136 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5137 * Description 5138 * For each element in **map**, call **callback_fn** function with 5139 * **map**, **callback_ctx** and other map-specific parameters. 5140 * The **callback_fn** should be a static function and 5141 * the **callback_ctx** should be a pointer to the stack. 5142 * The **flags** is used to control certain aspects of the helper. 5143 * Currently, the **flags** must be 0. 5144 * 5145 * The following are a list of supported map types and their 5146 * respective expected callback signatures: 5147 * 5148 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5149 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5150 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5151 * 5152 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5153 * 5154 * For per_cpu maps, the map_value is the value on the cpu where the 5155 * bpf_prog is running. 5156 * 5157 * If **callback_fn** return 0, the helper will continue to the next 5158 * element. If return value is 1, the helper will skip the rest of 5159 * elements and return. Other return values are not used now. 5160 * 5161 * Return 5162 * The number of traversed map elements for success, **-EINVAL** for 5163 * invalid **flags**. 5164 * 5165 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5166 * Description 5167 * Outputs a string into the **str** buffer of size **str_size** 5168 * based on a format string stored in a read-only map pointed by 5169 * **fmt**. 5170 * 5171 * Each format specifier in **fmt** corresponds to one u64 element 5172 * in the **data** array. For strings and pointers where pointees 5173 * are accessed, only the pointer values are stored in the *data* 5174 * array. The *data_len* is the size of *data* in bytes - must be 5175 * a multiple of 8. 5176 * 5177 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5178 * memory. Reading kernel memory may fail due to either invalid 5179 * address or valid address but requiring a major memory fault. If 5180 * reading kernel memory fails, the string for **%s** will be an 5181 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5182 * Not returning error to bpf program is consistent with what 5183 * **bpf_trace_printk**\ () does for now. 5184 * 5185 * Return 5186 * The strictly positive length of the formatted string, including 5187 * the trailing zero character. If the return value is greater than 5188 * **str_size**, **str** contains a truncated string, guaranteed to 5189 * be zero-terminated except when **str_size** is 0. 5190 * 5191 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5192 * 5193 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5194 * Description 5195 * Execute bpf syscall with given arguments. 5196 * Return 5197 * A syscall result. 5198 * 5199 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5200 * Description 5201 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5202 * Return 5203 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5204 * 5205 * long bpf_sys_close(u32 fd) 5206 * Description 5207 * Execute close syscall for given FD. 5208 * Return 5209 * A syscall result. 5210 * 5211 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5212 * Description 5213 * Initialize the timer. 5214 * First 4 bits of *flags* specify clockid. 5215 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5216 * All other bits of *flags* are reserved. 5217 * The verifier will reject the program if *timer* is not from 5218 * the same *map*. 5219 * Return 5220 * 0 on success. 5221 * **-EBUSY** if *timer* is already initialized. 5222 * **-EINVAL** if invalid *flags* are passed. 5223 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5224 * The user space should either hold a file descriptor to a map with timers 5225 * or pin such map in bpffs. When map is unpinned or file descriptor is 5226 * closed all timers in the map will be cancelled and freed. 5227 * 5228 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5229 * Description 5230 * Configure the timer to call *callback_fn* static function. 5231 * Return 5232 * 0 on success. 5233 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5234 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5235 * The user space should either hold a file descriptor to a map with timers 5236 * or pin such map in bpffs. When map is unpinned or file descriptor is 5237 * closed all timers in the map will be cancelled and freed. 5238 * 5239 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5240 * Description 5241 * Set timer expiration N nanoseconds from the current time. The 5242 * configured callback will be invoked in soft irq context on some cpu 5243 * and will not repeat unless another bpf_timer_start() is made. 5244 * In such case the next invocation can migrate to a different cpu. 5245 * Since struct bpf_timer is a field inside map element the map 5246 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5247 * of BPF program to make sure that callback_fn code stays valid. 5248 * When user space reference to a map reaches zero all timers 5249 * in a map are cancelled and corresponding program's refcnts are 5250 * decremented. This is done to make sure that Ctrl-C of a user 5251 * process doesn't leave any timers running. If map is pinned in 5252 * bpffs the callback_fn can re-arm itself indefinitely. 5253 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5254 * cancel and free the timer in the given map element. 5255 * The map can contain timers that invoke callback_fn-s from different 5256 * programs. The same callback_fn can serve different timers from 5257 * different maps if key/value layout matches across maps. 5258 * Every bpf_timer_set_callback() can have different callback_fn. 5259 * 5260 * *flags* can be one of: 5261 * 5262 * **BPF_F_TIMER_ABS** 5263 * Start the timer in absolute expire value instead of the 5264 * default relative one. 5265 * **BPF_F_TIMER_CPU_PIN** 5266 * Timer will be pinned to the CPU of the caller. 5267 * 5268 * Return 5269 * 0 on success. 5270 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5271 * or invalid *flags* are passed. 5272 * 5273 * long bpf_timer_cancel(struct bpf_timer *timer) 5274 * Description 5275 * Cancel the timer and wait for callback_fn to finish if it was running. 5276 * Return 5277 * 0 if the timer was not active. 5278 * 1 if the timer was active. 5279 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5280 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5281 * own timer which would have led to a deadlock otherwise. 5282 * 5283 * u64 bpf_get_func_ip(void *ctx) 5284 * Description 5285 * Get address of the traced function (for tracing and kprobe programs). 5286 * 5287 * When called for kprobe program attached as uprobe it returns 5288 * probe address for both entry and return uprobe. 5289 * 5290 * Return 5291 * Address of the traced function for kprobe. 5292 * 0 for kprobes placed within the function (not at the entry). 5293 * Address of the probe for uprobe and return uprobe. 5294 * 5295 * u64 bpf_get_attach_cookie(void *ctx) 5296 * Description 5297 * Get bpf_cookie value provided (optionally) during the program 5298 * attachment. It might be different for each individual 5299 * attachment, even if BPF program itself is the same. 5300 * Expects BPF program context *ctx* as a first argument. 5301 * 5302 * Supported for the following program types: 5303 * - kprobe/uprobe; 5304 * - tracepoint; 5305 * - perf_event. 5306 * Return 5307 * Value specified by user at BPF link creation/attachment time 5308 * or 0, if it was not specified. 5309 * 5310 * long bpf_task_pt_regs(struct task_struct *task) 5311 * Description 5312 * Get the struct pt_regs associated with **task**. 5313 * Return 5314 * A pointer to struct pt_regs. 5315 * 5316 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5317 * Description 5318 * Get branch trace from hardware engines like Intel LBR. The 5319 * hardware engine is stopped shortly after the helper is 5320 * called. Therefore, the user need to filter branch entries 5321 * based on the actual use case. To capture branch trace 5322 * before the trigger point of the BPF program, the helper 5323 * should be called at the beginning of the BPF program. 5324 * 5325 * The data is stored as struct perf_branch_entry into output 5326 * buffer *entries*. *size* is the size of *entries* in bytes. 5327 * *flags* is reserved for now and must be zero. 5328 * 5329 * Return 5330 * On success, number of bytes written to *buf*. On error, a 5331 * negative value. 5332 * 5333 * **-EINVAL** if *flags* is not zero. 5334 * 5335 * **-ENOENT** if architecture does not support branch records. 5336 * 5337 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5338 * Description 5339 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5340 * to format and can handle more format args as a result. 5341 * 5342 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5343 * Return 5344 * The number of bytes written to the buffer, or a negative error 5345 * in case of failure. 5346 * 5347 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5348 * Description 5349 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5350 * Return 5351 * *sk* if casting is valid, or **NULL** otherwise. 5352 * 5353 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5354 * Description 5355 * Get the address of a kernel symbol, returned in *res*. *res* is 5356 * set to 0 if the symbol is not found. 5357 * Return 5358 * On success, zero. On error, a negative value. 5359 * 5360 * **-EINVAL** if *flags* is not zero. 5361 * 5362 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5363 * 5364 * **-ENOENT** if symbol is not found. 5365 * 5366 * **-EPERM** if caller does not have permission to obtain kernel address. 5367 * 5368 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5369 * Description 5370 * Find vma of *task* that contains *addr*, call *callback_fn* 5371 * function with *task*, *vma*, and *callback_ctx*. 5372 * The *callback_fn* should be a static function and 5373 * the *callback_ctx* should be a pointer to the stack. 5374 * The *flags* is used to control certain aspects of the helper. 5375 * Currently, the *flags* must be 0. 5376 * 5377 * The expected callback signature is 5378 * 5379 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5380 * 5381 * Return 5382 * 0 on success. 5383 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5384 * **-EBUSY** if failed to try lock mmap_lock. 5385 * **-EINVAL** for invalid **flags**. 5386 * 5387 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5388 * Description 5389 * For **nr_loops**, call **callback_fn** function 5390 * with **callback_ctx** as the context parameter. 5391 * The **callback_fn** should be a static function and 5392 * the **callback_ctx** should be a pointer to the stack. 5393 * The **flags** is used to control certain aspects of the helper. 5394 * Currently, the **flags** must be 0. Currently, nr_loops is 5395 * limited to 1 << 23 (~8 million) loops. 5396 * 5397 * long (\*callback_fn)(u64 index, void \*ctx); 5398 * 5399 * where **index** is the current index in the loop. The index 5400 * is zero-indexed. 5401 * 5402 * If **callback_fn** returns 0, the helper will continue to the next 5403 * loop. If return value is 1, the helper will skip the rest of 5404 * the loops and return. Other return values are not used now, 5405 * and will be rejected by the verifier. 5406 * 5407 * Return 5408 * The number of loops performed, **-EINVAL** for invalid **flags**, 5409 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5410 * 5411 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5412 * Description 5413 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5414 * to be null-terminated and **s1_sz** is the maximum storage 5415 * size of **s1**. **s2** must be a read-only string. 5416 * Return 5417 * An integer less than, equal to, or greater than zero 5418 * if the first **s1_sz** bytes of **s1** is found to be 5419 * less than, to match, or be greater than **s2**. 5420 * 5421 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5422 * Description 5423 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5424 * returned in **value**. 5425 * 5426 * Return 5427 * 0 on success. 5428 * **-EINVAL** if n >= argument register count of traced function. 5429 * 5430 * long bpf_get_func_ret(void *ctx, u64 *value) 5431 * Description 5432 * Get return value of the traced function (for tracing programs) 5433 * in **value**. 5434 * 5435 * Return 5436 * 0 on success. 5437 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5438 * 5439 * long bpf_get_func_arg_cnt(void *ctx) 5440 * Description 5441 * Get number of registers of the traced function (for tracing programs) where 5442 * function arguments are stored in these registers. 5443 * 5444 * Return 5445 * The number of argument registers of the traced function. 5446 * 5447 * int bpf_get_retval(void) 5448 * Description 5449 * Get the BPF program's return value that will be returned to the upper layers. 5450 * 5451 * This helper is currently supported by cgroup programs and only by the hooks 5452 * where BPF program's return value is returned to the userspace via errno. 5453 * Return 5454 * The BPF program's return value. 5455 * 5456 * int bpf_set_retval(int retval) 5457 * Description 5458 * Set the BPF program's return value that will be returned to the upper layers. 5459 * 5460 * This helper is currently supported by cgroup programs and only by the hooks 5461 * where BPF program's return value is returned to the userspace via errno. 5462 * 5463 * Note that there is the following corner case where the program exports an error 5464 * via bpf_set_retval but signals success via 'return 1': 5465 * 5466 * bpf_set_retval(-EPERM); 5467 * return 1; 5468 * 5469 * In this case, the BPF program's return value will use helper's -EPERM. This 5470 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5471 * 5472 * Return 5473 * 0 on success, or a negative error in case of failure. 5474 * 5475 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5476 * Description 5477 * Get the total size of a given xdp buff (linear and paged area) 5478 * Return 5479 * The total size of a given xdp buffer. 5480 * 5481 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5482 * Description 5483 * This helper is provided as an easy way to load data from a 5484 * xdp buffer. It can be used to load *len* bytes from *offset* from 5485 * the frame associated to *xdp_md*, into the buffer pointed by 5486 * *buf*. 5487 * Return 5488 * 0 on success, or a negative error in case of failure. 5489 * 5490 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5491 * Description 5492 * Store *len* bytes from buffer *buf* into the frame 5493 * associated to *xdp_md*, at *offset*. 5494 * Return 5495 * 0 on success, or a negative error in case of failure. 5496 * 5497 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5498 * Description 5499 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5500 * address space, and stores the data in *dst*. *flags* is not 5501 * used yet and is provided for future extensibility. This helper 5502 * can only be used by sleepable programs. 5503 * Return 5504 * 0 on success, or a negative error in case of failure. On error 5505 * *dst* buffer is zeroed out. 5506 * 5507 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5508 * Description 5509 * Change the __sk_buff->tstamp_type to *tstamp_type* 5510 * and set *tstamp* to the __sk_buff->tstamp together. 5511 * 5512 * If there is no need to change the __sk_buff->tstamp_type, 5513 * the tstamp value can be directly written to __sk_buff->tstamp 5514 * instead. 5515 * 5516 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5517 * will be kept during bpf_redirect_*(). A non zero 5518 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5519 * *tstamp_type*. 5520 * 5521 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5522 * with a zero *tstamp*. 5523 * 5524 * Only IPv4 and IPv6 skb->protocol are supported. 5525 * 5526 * This function is most useful when it needs to set a 5527 * mono delivery time to __sk_buff->tstamp and then 5528 * bpf_redirect_*() to the egress of an iface. For example, 5529 * changing the (rcv) timestamp in __sk_buff->tstamp at 5530 * ingress to a mono delivery time and then bpf_redirect_*() 5531 * to sch_fq@phy-dev. 5532 * Return 5533 * 0 on success. 5534 * **-EINVAL** for invalid input 5535 * **-EOPNOTSUPP** for unsupported protocol 5536 * 5537 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5538 * Description 5539 * Returns a calculated IMA hash of the *file*. 5540 * If the hash is larger than *size*, then only *size* 5541 * bytes will be copied to *dst* 5542 * Return 5543 * The **hash_algo** is returned on success, 5544 * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if 5545 * invalid arguments are passed. 5546 * 5547 * void *bpf_kptr_xchg(void *dst, void *ptr) 5548 * Description 5549 * Exchange kptr at pointer *dst* with *ptr*, and return the old value. 5550 * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise 5551 * it must be a referenced pointer which will be released when this helper 5552 * is called. 5553 * Return 5554 * The old value of kptr (which can be NULL). The returned pointer 5555 * if not NULL, is a reference which must be released using its 5556 * corresponding release function, or moved into a BPF map before 5557 * program exit. 5558 * 5559 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5560 * Description 5561 * Perform a lookup in *percpu map* for an entry associated to 5562 * *key* on *cpu*. 5563 * Return 5564 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5565 * was found or *cpu* is invalid. 5566 * 5567 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5568 * Description 5569 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5570 * Return 5571 * *sk* if casting is valid, or **NULL** otherwise. 5572 * 5573 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5574 * Description 5575 * Get a dynptr to local memory *data*. 5576 * 5577 * *data* must be a ptr to a map value. 5578 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5579 * *flags* is currently unused. 5580 * Return 5581 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5582 * -EINVAL if flags is not 0. 5583 * 5584 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5585 * Description 5586 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5587 * through the dynptr interface. *flags* must be 0. 5588 * 5589 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5590 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5591 * reservation fails. This is enforced by the verifier. 5592 * Return 5593 * 0 on success, or a negative error in case of failure. 5594 * 5595 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5596 * Description 5597 * Submit reserved ring buffer sample, pointed to by *data*, 5598 * through the dynptr interface. This is a no-op if the dynptr is 5599 * invalid/null. 5600 * 5601 * For more information on *flags*, please see 5602 * 'bpf_ringbuf_submit'. 5603 * Return 5604 * Nothing. Always succeeds. 5605 * 5606 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5607 * Description 5608 * Discard reserved ring buffer sample through the dynptr 5609 * interface. This is a no-op if the dynptr is invalid/null. 5610 * 5611 * For more information on *flags*, please see 5612 * 'bpf_ringbuf_discard'. 5613 * Return 5614 * Nothing. Always succeeds. 5615 * 5616 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5617 * Description 5618 * Read *len* bytes from *src* into *dst*, starting from *offset* 5619 * into *src*. 5620 * *flags* is currently unused. 5621 * Return 5622 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5623 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5624 * *flags* is not 0. 5625 * 5626 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5627 * Description 5628 * Write *len* bytes from *src* into *dst*, starting from *offset* 5629 * into *dst*. 5630 * 5631 * *flags* must be 0 except for skb-type dynptrs. 5632 * 5633 * For skb-type dynptrs: 5634 * * All data slices of the dynptr are automatically 5635 * invalidated after **bpf_dynptr_write**\ (). This is 5636 * because writing may pull the skb and change the 5637 * underlying packet buffer. 5638 * 5639 * * For *flags*, please see the flags accepted by 5640 * **bpf_skb_store_bytes**\ (). 5641 * Return 5642 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5643 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5644 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5645 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5646 * 5647 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5648 * Description 5649 * Get a pointer to the underlying dynptr data. 5650 * 5651 * *len* must be a statically known value. The returned data slice 5652 * is invalidated whenever the dynptr is invalidated. 5653 * 5654 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5655 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5656 * Return 5657 * Pointer to the underlying dynptr data, NULL if the dynptr is 5658 * read-only, if the dynptr is invalid, or if the offset and length 5659 * is out of bounds. 5660 * 5661 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5662 * Description 5663 * Try to issue a SYN cookie for the packet with corresponding 5664 * IPv4/TCP headers, *iph* and *th*, without depending on a 5665 * listening socket. 5666 * 5667 * *iph* points to the IPv4 header. 5668 * 5669 * *th* points to the start of the TCP header, while *th_len* 5670 * contains the length of the TCP header (at least 5671 * **sizeof**\ (**struct tcphdr**)). 5672 * Return 5673 * On success, lower 32 bits hold the generated SYN cookie in 5674 * followed by 16 bits which hold the MSS value for that cookie, 5675 * and the top 16 bits are unused. 5676 * 5677 * On failure, the returned value is one of the following: 5678 * 5679 * **-EINVAL** if *th_len* is invalid. 5680 * 5681 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5682 * Description 5683 * Try to issue a SYN cookie for the packet with corresponding 5684 * IPv6/TCP headers, *iph* and *th*, without depending on a 5685 * listening socket. 5686 * 5687 * *iph* points to the IPv6 header. 5688 * 5689 * *th* points to the start of the TCP header, while *th_len* 5690 * contains the length of the TCP header (at least 5691 * **sizeof**\ (**struct tcphdr**)). 5692 * Return 5693 * On success, lower 32 bits hold the generated SYN cookie in 5694 * followed by 16 bits which hold the MSS value for that cookie, 5695 * and the top 16 bits are unused. 5696 * 5697 * On failure, the returned value is one of the following: 5698 * 5699 * **-EINVAL** if *th_len* is invalid. 5700 * 5701 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5702 * 5703 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5704 * Description 5705 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5706 * without depending on a listening socket. 5707 * 5708 * *iph* points to the IPv4 header. 5709 * 5710 * *th* points to the TCP header. 5711 * Return 5712 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5713 * 5714 * On failure, the returned value is one of the following: 5715 * 5716 * **-EACCES** if the SYN cookie is not valid. 5717 * 5718 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5719 * Description 5720 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5721 * without depending on a listening socket. 5722 * 5723 * *iph* points to the IPv6 header. 5724 * 5725 * *th* points to the TCP header. 5726 * Return 5727 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5728 * 5729 * On failure, the returned value is one of the following: 5730 * 5731 * **-EACCES** if the SYN cookie is not valid. 5732 * 5733 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5734 * 5735 * u64 bpf_ktime_get_tai_ns(void) 5736 * Description 5737 * A nonsettable system-wide clock derived from wall-clock time but 5738 * ignoring leap seconds. This clock does not experience 5739 * discontinuities and backwards jumps caused by NTP inserting leap 5740 * seconds as CLOCK_REALTIME does. 5741 * 5742 * See: **clock_gettime**\ (**CLOCK_TAI**) 5743 * Return 5744 * Current *ktime*. 5745 * 5746 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5747 * Description 5748 * Drain samples from the specified user ring buffer, and invoke 5749 * the provided callback for each such sample: 5750 * 5751 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5752 * 5753 * If **callback_fn** returns 0, the helper will continue to try 5754 * and drain the next sample, up to a maximum of 5755 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5756 * the helper will skip the rest of the samples and return. Other 5757 * return values are not used now, and will be rejected by the 5758 * verifier. 5759 * Return 5760 * The number of drained samples if no error was encountered while 5761 * draining samples, or 0 if no samples were present in the ring 5762 * buffer. If a user-space producer was epoll-waiting on this map, 5763 * and at least one sample was drained, they will receive an event 5764 * notification notifying them of available space in the ring 5765 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5766 * function, no wakeup notification will be sent. If the 5767 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5768 * be sent even if no sample was drained. 5769 * 5770 * On failure, the returned value is one of the following: 5771 * 5772 * **-EBUSY** if the ring buffer is contended, and another calling 5773 * context was concurrently draining the ring buffer. 5774 * 5775 * **-EINVAL** if user-space is not properly tracking the ring 5776 * buffer due to the producer position not being aligned to 8 5777 * bytes, a sample not being aligned to 8 bytes, or the producer 5778 * position not matching the advertised length of a sample. 5779 * 5780 * **-E2BIG** if user-space has tried to publish a sample which is 5781 * larger than the size of the ring buffer, or which cannot fit 5782 * within a struct bpf_dynptr. 5783 * 5784 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5785 * Description 5786 * Get a bpf_local_storage from the *cgroup*. 5787 * 5788 * Logically, it could be thought of as getting the value from 5789 * a *map* with *cgroup* as the **key**. From this 5790 * perspective, the usage is not much different from 5791 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5792 * helper enforces the key must be a cgroup struct and the map must also 5793 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5794 * 5795 * In reality, the local-storage value is embedded directly inside of the 5796 * *cgroup* object itself, rather than being located in the 5797 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5798 * queried for some *map* on a *cgroup* object, the kernel will perform an 5799 * O(n) iteration over all of the live local-storage values for that 5800 * *cgroup* object until the local-storage value for the *map* is found. 5801 * 5802 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5803 * used such that a new bpf_local_storage will be 5804 * created if one does not exist. *value* can be used 5805 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5806 * the initial value of a bpf_local_storage. If *value* is 5807 * **NULL**, the new bpf_local_storage will be zero initialized. 5808 * Return 5809 * A bpf_local_storage pointer is returned on success. 5810 * 5811 * **NULL** if not found or there was an error in adding 5812 * a new bpf_local_storage. 5813 * 5814 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5815 * Description 5816 * Delete a bpf_local_storage from a *cgroup*. 5817 * Return 5818 * 0 on success. 5819 * 5820 * **-ENOENT** if the bpf_local_storage cannot be found. 5821 */ 5822 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5823 FN(unspec, 0, ##ctx) \ 5824 FN(map_lookup_elem, 1, ##ctx) \ 5825 FN(map_update_elem, 2, ##ctx) \ 5826 FN(map_delete_elem, 3, ##ctx) \ 5827 FN(probe_read, 4, ##ctx) \ 5828 FN(ktime_get_ns, 5, ##ctx) \ 5829 FN(trace_printk, 6, ##ctx) \ 5830 FN(get_prandom_u32, 7, ##ctx) \ 5831 FN(get_smp_processor_id, 8, ##ctx) \ 5832 FN(skb_store_bytes, 9, ##ctx) \ 5833 FN(l3_csum_replace, 10, ##ctx) \ 5834 FN(l4_csum_replace, 11, ##ctx) \ 5835 FN(tail_call, 12, ##ctx) \ 5836 FN(clone_redirect, 13, ##ctx) \ 5837 FN(get_current_pid_tgid, 14, ##ctx) \ 5838 FN(get_current_uid_gid, 15, ##ctx) \ 5839 FN(get_current_comm, 16, ##ctx) \ 5840 FN(get_cgroup_classid, 17, ##ctx) \ 5841 FN(skb_vlan_push, 18, ##ctx) \ 5842 FN(skb_vlan_pop, 19, ##ctx) \ 5843 FN(skb_get_tunnel_key, 20, ##ctx) \ 5844 FN(skb_set_tunnel_key, 21, ##ctx) \ 5845 FN(perf_event_read, 22, ##ctx) \ 5846 FN(redirect, 23, ##ctx) \ 5847 FN(get_route_realm, 24, ##ctx) \ 5848 FN(perf_event_output, 25, ##ctx) \ 5849 FN(skb_load_bytes, 26, ##ctx) \ 5850 FN(get_stackid, 27, ##ctx) \ 5851 FN(csum_diff, 28, ##ctx) \ 5852 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5853 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5854 FN(skb_change_proto, 31, ##ctx) \ 5855 FN(skb_change_type, 32, ##ctx) \ 5856 FN(skb_under_cgroup, 33, ##ctx) \ 5857 FN(get_hash_recalc, 34, ##ctx) \ 5858 FN(get_current_task, 35, ##ctx) \ 5859 FN(probe_write_user, 36, ##ctx) \ 5860 FN(current_task_under_cgroup, 37, ##ctx) \ 5861 FN(skb_change_tail, 38, ##ctx) \ 5862 FN(skb_pull_data, 39, ##ctx) \ 5863 FN(csum_update, 40, ##ctx) \ 5864 FN(set_hash_invalid, 41, ##ctx) \ 5865 FN(get_numa_node_id, 42, ##ctx) \ 5866 FN(skb_change_head, 43, ##ctx) \ 5867 FN(xdp_adjust_head, 44, ##ctx) \ 5868 FN(probe_read_str, 45, ##ctx) \ 5869 FN(get_socket_cookie, 46, ##ctx) \ 5870 FN(get_socket_uid, 47, ##ctx) \ 5871 FN(set_hash, 48, ##ctx) \ 5872 FN(setsockopt, 49, ##ctx) \ 5873 FN(skb_adjust_room, 50, ##ctx) \ 5874 FN(redirect_map, 51, ##ctx) \ 5875 FN(sk_redirect_map, 52, ##ctx) \ 5876 FN(sock_map_update, 53, ##ctx) \ 5877 FN(xdp_adjust_meta, 54, ##ctx) \ 5878 FN(perf_event_read_value, 55, ##ctx) \ 5879 FN(perf_prog_read_value, 56, ##ctx) \ 5880 FN(getsockopt, 57, ##ctx) \ 5881 FN(override_return, 58, ##ctx) \ 5882 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5883 FN(msg_redirect_map, 60, ##ctx) \ 5884 FN(msg_apply_bytes, 61, ##ctx) \ 5885 FN(msg_cork_bytes, 62, ##ctx) \ 5886 FN(msg_pull_data, 63, ##ctx) \ 5887 FN(bind, 64, ##ctx) \ 5888 FN(xdp_adjust_tail, 65, ##ctx) \ 5889 FN(skb_get_xfrm_state, 66, ##ctx) \ 5890 FN(get_stack, 67, ##ctx) \ 5891 FN(skb_load_bytes_relative, 68, ##ctx) \ 5892 FN(fib_lookup, 69, ##ctx) \ 5893 FN(sock_hash_update, 70, ##ctx) \ 5894 FN(msg_redirect_hash, 71, ##ctx) \ 5895 FN(sk_redirect_hash, 72, ##ctx) \ 5896 FN(lwt_push_encap, 73, ##ctx) \ 5897 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5898 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5899 FN(lwt_seg6_action, 76, ##ctx) \ 5900 FN(rc_repeat, 77, ##ctx) \ 5901 FN(rc_keydown, 78, ##ctx) \ 5902 FN(skb_cgroup_id, 79, ##ctx) \ 5903 FN(get_current_cgroup_id, 80, ##ctx) \ 5904 FN(get_local_storage, 81, ##ctx) \ 5905 FN(sk_select_reuseport, 82, ##ctx) \ 5906 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5907 FN(sk_lookup_tcp, 84, ##ctx) \ 5908 FN(sk_lookup_udp, 85, ##ctx) \ 5909 FN(sk_release, 86, ##ctx) \ 5910 FN(map_push_elem, 87, ##ctx) \ 5911 FN(map_pop_elem, 88, ##ctx) \ 5912 FN(map_peek_elem, 89, ##ctx) \ 5913 FN(msg_push_data, 90, ##ctx) \ 5914 FN(msg_pop_data, 91, ##ctx) \ 5915 FN(rc_pointer_rel, 92, ##ctx) \ 5916 FN(spin_lock, 93, ##ctx) \ 5917 FN(spin_unlock, 94, ##ctx) \ 5918 FN(sk_fullsock, 95, ##ctx) \ 5919 FN(tcp_sock, 96, ##ctx) \ 5920 FN(skb_ecn_set_ce, 97, ##ctx) \ 5921 FN(get_listener_sock, 98, ##ctx) \ 5922 FN(skc_lookup_tcp, 99, ##ctx) \ 5923 FN(tcp_check_syncookie, 100, ##ctx) \ 5924 FN(sysctl_get_name, 101, ##ctx) \ 5925 FN(sysctl_get_current_value, 102, ##ctx) \ 5926 FN(sysctl_get_new_value, 103, ##ctx) \ 5927 FN(sysctl_set_new_value, 104, ##ctx) \ 5928 FN(strtol, 105, ##ctx) \ 5929 FN(strtoul, 106, ##ctx) \ 5930 FN(sk_storage_get, 107, ##ctx) \ 5931 FN(sk_storage_delete, 108, ##ctx) \ 5932 FN(send_signal, 109, ##ctx) \ 5933 FN(tcp_gen_syncookie, 110, ##ctx) \ 5934 FN(skb_output, 111, ##ctx) \ 5935 FN(probe_read_user, 112, ##ctx) \ 5936 FN(probe_read_kernel, 113, ##ctx) \ 5937 FN(probe_read_user_str, 114, ##ctx) \ 5938 FN(probe_read_kernel_str, 115, ##ctx) \ 5939 FN(tcp_send_ack, 116, ##ctx) \ 5940 FN(send_signal_thread, 117, ##ctx) \ 5941 FN(jiffies64, 118, ##ctx) \ 5942 FN(read_branch_records, 119, ##ctx) \ 5943 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5944 FN(xdp_output, 121, ##ctx) \ 5945 FN(get_netns_cookie, 122, ##ctx) \ 5946 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5947 FN(sk_assign, 124, ##ctx) \ 5948 FN(ktime_get_boot_ns, 125, ##ctx) \ 5949 FN(seq_printf, 126, ##ctx) \ 5950 FN(seq_write, 127, ##ctx) \ 5951 FN(sk_cgroup_id, 128, ##ctx) \ 5952 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5953 FN(ringbuf_output, 130, ##ctx) \ 5954 FN(ringbuf_reserve, 131, ##ctx) \ 5955 FN(ringbuf_submit, 132, ##ctx) \ 5956 FN(ringbuf_discard, 133, ##ctx) \ 5957 FN(ringbuf_query, 134, ##ctx) \ 5958 FN(csum_level, 135, ##ctx) \ 5959 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5960 FN(skc_to_tcp_sock, 137, ##ctx) \ 5961 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5962 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5963 FN(skc_to_udp6_sock, 140, ##ctx) \ 5964 FN(get_task_stack, 141, ##ctx) \ 5965 FN(load_hdr_opt, 142, ##ctx) \ 5966 FN(store_hdr_opt, 143, ##ctx) \ 5967 FN(reserve_hdr_opt, 144, ##ctx) \ 5968 FN(inode_storage_get, 145, ##ctx) \ 5969 FN(inode_storage_delete, 146, ##ctx) \ 5970 FN(d_path, 147, ##ctx) \ 5971 FN(copy_from_user, 148, ##ctx) \ 5972 FN(snprintf_btf, 149, ##ctx) \ 5973 FN(seq_printf_btf, 150, ##ctx) \ 5974 FN(skb_cgroup_classid, 151, ##ctx) \ 5975 FN(redirect_neigh, 152, ##ctx) \ 5976 FN(per_cpu_ptr, 153, ##ctx) \ 5977 FN(this_cpu_ptr, 154, ##ctx) \ 5978 FN(redirect_peer, 155, ##ctx) \ 5979 FN(task_storage_get, 156, ##ctx) \ 5980 FN(task_storage_delete, 157, ##ctx) \ 5981 FN(get_current_task_btf, 158, ##ctx) \ 5982 FN(bprm_opts_set, 159, ##ctx) \ 5983 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5984 FN(ima_inode_hash, 161, ##ctx) \ 5985 FN(sock_from_file, 162, ##ctx) \ 5986 FN(check_mtu, 163, ##ctx) \ 5987 FN(for_each_map_elem, 164, ##ctx) \ 5988 FN(snprintf, 165, ##ctx) \ 5989 FN(sys_bpf, 166, ##ctx) \ 5990 FN(btf_find_by_name_kind, 167, ##ctx) \ 5991 FN(sys_close, 168, ##ctx) \ 5992 FN(timer_init, 169, ##ctx) \ 5993 FN(timer_set_callback, 170, ##ctx) \ 5994 FN(timer_start, 171, ##ctx) \ 5995 FN(timer_cancel, 172, ##ctx) \ 5996 FN(get_func_ip, 173, ##ctx) \ 5997 FN(get_attach_cookie, 174, ##ctx) \ 5998 FN(task_pt_regs, 175, ##ctx) \ 5999 FN(get_branch_snapshot, 176, ##ctx) \ 6000 FN(trace_vprintk, 177, ##ctx) \ 6001 FN(skc_to_unix_sock, 178, ##ctx) \ 6002 FN(kallsyms_lookup_name, 179, ##ctx) \ 6003 FN(find_vma, 180, ##ctx) \ 6004 FN(loop, 181, ##ctx) \ 6005 FN(strncmp, 182, ##ctx) \ 6006 FN(get_func_arg, 183, ##ctx) \ 6007 FN(get_func_ret, 184, ##ctx) \ 6008 FN(get_func_arg_cnt, 185, ##ctx) \ 6009 FN(get_retval, 186, ##ctx) \ 6010 FN(set_retval, 187, ##ctx) \ 6011 FN(xdp_get_buff_len, 188, ##ctx) \ 6012 FN(xdp_load_bytes, 189, ##ctx) \ 6013 FN(xdp_store_bytes, 190, ##ctx) \ 6014 FN(copy_from_user_task, 191, ##ctx) \ 6015 FN(skb_set_tstamp, 192, ##ctx) \ 6016 FN(ima_file_hash, 193, ##ctx) \ 6017 FN(kptr_xchg, 194, ##ctx) \ 6018 FN(map_lookup_percpu_elem, 195, ##ctx) \ 6019 FN(skc_to_mptcp_sock, 196, ##ctx) \ 6020 FN(dynptr_from_mem, 197, ##ctx) \ 6021 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 6022 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 6023 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 6024 FN(dynptr_read, 201, ##ctx) \ 6025 FN(dynptr_write, 202, ##ctx) \ 6026 FN(dynptr_data, 203, ##ctx) \ 6027 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 6028 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 6029 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 6030 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 6031 FN(ktime_get_tai_ns, 208, ##ctx) \ 6032 FN(user_ringbuf_drain, 209, ##ctx) \ 6033 FN(cgrp_storage_get, 210, ##ctx) \ 6034 FN(cgrp_storage_delete, 211, ##ctx) \ 6035 /* This helper list is effectively frozen. If you are trying to \ 6036 * add a new helper, you should add a kfunc instead which has \ 6037 * less stability guarantees. See Documentation/bpf/kfuncs.rst \ 6038 */ 6039 6040 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 6041 * know or care about integer value that is now passed as second argument 6042 */ 6043 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 6044 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 6045 6046 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 6047 * function eBPF program intends to call 6048 */ 6049 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 6050 enum bpf_func_id { 6051 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 6052 __BPF_FUNC_MAX_ID, 6053 }; 6054 #undef __BPF_ENUM_FN 6055 6056 /* All flags used by eBPF helper functions, placed here. */ 6057 6058 /* BPF_FUNC_skb_store_bytes flags. */ 6059 enum { 6060 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 6061 BPF_F_INVALIDATE_HASH = (1ULL << 1), 6062 }; 6063 6064 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 6065 * First 4 bits are for passing the header field size. 6066 */ 6067 enum { 6068 BPF_F_HDR_FIELD_MASK = 0xfULL, 6069 }; 6070 6071 /* BPF_FUNC_l4_csum_replace flags. */ 6072 enum { 6073 BPF_F_PSEUDO_HDR = (1ULL << 4), 6074 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 6075 BPF_F_MARK_ENFORCE = (1ULL << 6), 6076 BPF_F_IPV6 = (1ULL << 7), 6077 }; 6078 6079 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6080 enum { 6081 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6082 }; 6083 6084 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6085 enum { 6086 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6087 BPF_F_USER_STACK = (1ULL << 8), 6088 /* flags used by BPF_FUNC_get_stackid only. */ 6089 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6090 BPF_F_REUSE_STACKID = (1ULL << 10), 6091 /* flags used by BPF_FUNC_get_stack only. */ 6092 BPF_F_USER_BUILD_ID = (1ULL << 11), 6093 }; 6094 6095 /* BPF_FUNC_skb_set_tunnel_key flags. */ 6096 enum { 6097 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6098 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6099 BPF_F_SEQ_NUMBER = (1ULL << 3), 6100 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6101 }; 6102 6103 /* BPF_FUNC_skb_get_tunnel_key flags. */ 6104 enum { 6105 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6106 }; 6107 6108 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6109 * BPF_FUNC_perf_event_read_value flags. 6110 */ 6111 enum { 6112 BPF_F_INDEX_MASK = 0xffffffffULL, 6113 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6114 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 6115 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6116 }; 6117 6118 /* Current network namespace */ 6119 enum { 6120 BPF_F_CURRENT_NETNS = (-1L), 6121 }; 6122 6123 /* BPF_FUNC_csum_level level values. */ 6124 enum { 6125 BPF_CSUM_LEVEL_QUERY, 6126 BPF_CSUM_LEVEL_INC, 6127 BPF_CSUM_LEVEL_DEC, 6128 BPF_CSUM_LEVEL_RESET, 6129 }; 6130 6131 /* BPF_FUNC_skb_adjust_room flags. */ 6132 enum { 6133 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6134 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6135 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6136 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6137 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6138 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6139 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6140 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6141 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6142 }; 6143 6144 enum { 6145 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6146 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6147 }; 6148 6149 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6150 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6151 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6152 6153 /* BPF_FUNC_sysctl_get_name flags. */ 6154 enum { 6155 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6156 }; 6157 6158 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 6159 enum { 6160 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6161 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6162 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6163 */ 6164 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6165 }; 6166 6167 /* BPF_FUNC_read_branch_records flags. */ 6168 enum { 6169 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6170 }; 6171 6172 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6173 * BPF_FUNC_bpf_ringbuf_output flags. 6174 */ 6175 enum { 6176 BPF_RB_NO_WAKEUP = (1ULL << 0), 6177 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6178 }; 6179 6180 /* BPF_FUNC_bpf_ringbuf_query flags */ 6181 enum { 6182 BPF_RB_AVAIL_DATA = 0, 6183 BPF_RB_RING_SIZE = 1, 6184 BPF_RB_CONS_POS = 2, 6185 BPF_RB_PROD_POS = 3, 6186 }; 6187 6188 /* BPF ring buffer constants */ 6189 enum { 6190 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6191 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6192 BPF_RINGBUF_HDR_SZ = 8, 6193 }; 6194 6195 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6196 enum { 6197 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6198 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6199 }; 6200 6201 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6202 enum bpf_adj_room_mode { 6203 BPF_ADJ_ROOM_NET, 6204 BPF_ADJ_ROOM_MAC, 6205 }; 6206 6207 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6208 enum bpf_hdr_start_off { 6209 BPF_HDR_START_MAC, 6210 BPF_HDR_START_NET, 6211 }; 6212 6213 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6214 enum bpf_lwt_encap_mode { 6215 BPF_LWT_ENCAP_SEG6, 6216 BPF_LWT_ENCAP_SEG6_INLINE, 6217 BPF_LWT_ENCAP_IP, 6218 }; 6219 6220 /* Flags for bpf_bprm_opts_set helper */ 6221 enum { 6222 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6223 }; 6224 6225 /* Flags for bpf_redirect and bpf_redirect_map helpers */ 6226 enum { 6227 BPF_F_INGRESS = (1ULL << 0), /* used for skb path */ 6228 BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */ 6229 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */ 6230 #define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS) 6231 }; 6232 6233 #define __bpf_md_ptr(type, name) \ 6234 union { \ 6235 type name; \ 6236 __u64 :64; \ 6237 } __attribute__((aligned(8))) 6238 6239 /* The enum used in skb->tstamp_type. It specifies the clock type 6240 * of the time stored in the skb->tstamp. 6241 */ 6242 enum { 6243 BPF_SKB_TSTAMP_UNSPEC = 0, /* DEPRECATED */ 6244 BPF_SKB_TSTAMP_DELIVERY_MONO = 1, /* DEPRECATED */ 6245 BPF_SKB_CLOCK_REALTIME = 0, 6246 BPF_SKB_CLOCK_MONOTONIC = 1, 6247 BPF_SKB_CLOCK_TAI = 2, 6248 /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle, 6249 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid. 6250 */ 6251 }; 6252 6253 /* user accessible mirror of in-kernel sk_buff. 6254 * new fields can only be added to the end of this structure 6255 */ 6256 struct __sk_buff { 6257 __u32 len; 6258 __u32 pkt_type; 6259 __u32 mark; 6260 __u32 queue_mapping; 6261 __u32 protocol; 6262 __u32 vlan_present; 6263 __u32 vlan_tci; 6264 __u32 vlan_proto; 6265 __u32 priority; 6266 __u32 ingress_ifindex; 6267 __u32 ifindex; 6268 __u32 tc_index; 6269 __u32 cb[5]; 6270 __u32 hash; 6271 __u32 tc_classid; 6272 __u32 data; 6273 __u32 data_end; 6274 __u32 napi_id; 6275 6276 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6277 __u32 family; 6278 __u32 remote_ip4; /* Stored in network byte order */ 6279 __u32 local_ip4; /* Stored in network byte order */ 6280 __u32 remote_ip6[4]; /* Stored in network byte order */ 6281 __u32 local_ip6[4]; /* Stored in network byte order */ 6282 __u32 remote_port; /* Stored in network byte order */ 6283 __u32 local_port; /* stored in host byte order */ 6284 /* ... here. */ 6285 6286 __u32 data_meta; 6287 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6288 __u64 tstamp; 6289 __u32 wire_len; 6290 __u32 gso_segs; 6291 __bpf_md_ptr(struct bpf_sock *, sk); 6292 __u32 gso_size; 6293 __u8 tstamp_type; 6294 __u32 :24; /* Padding, future use. */ 6295 __u64 hwtstamp; 6296 }; 6297 6298 struct bpf_tunnel_key { 6299 __u32 tunnel_id; 6300 union { 6301 __u32 remote_ipv4; 6302 __u32 remote_ipv6[4]; 6303 }; 6304 __u8 tunnel_tos; 6305 __u8 tunnel_ttl; 6306 union { 6307 __u16 tunnel_ext; /* compat */ 6308 __be16 tunnel_flags; 6309 }; 6310 __u32 tunnel_label; 6311 union { 6312 __u32 local_ipv4; 6313 __u32 local_ipv6[4]; 6314 }; 6315 }; 6316 6317 /* user accessible mirror of in-kernel xfrm_state. 6318 * new fields can only be added to the end of this structure 6319 */ 6320 struct bpf_xfrm_state { 6321 __u32 reqid; 6322 __u32 spi; /* Stored in network byte order */ 6323 __u16 family; 6324 __u16 ext; /* Padding, future use. */ 6325 union { 6326 __u32 remote_ipv4; /* Stored in network byte order */ 6327 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6328 }; 6329 }; 6330 6331 /* Generic BPF return codes which all BPF program types may support. 6332 * The values are binary compatible with their TC_ACT_* counter-part to 6333 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6334 * programs. 6335 * 6336 * XDP is handled seprately, see XDP_*. 6337 */ 6338 enum bpf_ret_code { 6339 BPF_OK = 0, 6340 /* 1 reserved */ 6341 BPF_DROP = 2, 6342 /* 3-6 reserved */ 6343 BPF_REDIRECT = 7, 6344 /* >127 are reserved for prog type specific return codes. 6345 * 6346 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6347 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6348 * changed and should be routed based on its new L3 header. 6349 * (This is an L3 redirect, as opposed to L2 redirect 6350 * represented by BPF_REDIRECT above). 6351 */ 6352 BPF_LWT_REROUTE = 128, 6353 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6354 * to indicate that no custom dissection was performed, and 6355 * fallback to standard dissector is requested. 6356 */ 6357 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6358 }; 6359 6360 struct bpf_sock { 6361 __u32 bound_dev_if; 6362 __u32 family; 6363 __u32 type; 6364 __u32 protocol; 6365 __u32 mark; 6366 __u32 priority; 6367 /* IP address also allows 1 and 2 bytes access */ 6368 __u32 src_ip4; 6369 __u32 src_ip6[4]; 6370 __u32 src_port; /* host byte order */ 6371 __be16 dst_port; /* network byte order */ 6372 __u16 :16; /* zero padding */ 6373 __u32 dst_ip4; 6374 __u32 dst_ip6[4]; 6375 __u32 state; 6376 __s32 rx_queue_mapping; 6377 }; 6378 6379 struct bpf_tcp_sock { 6380 __u32 snd_cwnd; /* Sending congestion window */ 6381 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6382 __u32 rtt_min; 6383 __u32 snd_ssthresh; /* Slow start size threshold */ 6384 __u32 rcv_nxt; /* What we want to receive next */ 6385 __u32 snd_nxt; /* Next sequence we send */ 6386 __u32 snd_una; /* First byte we want an ack for */ 6387 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6388 __u32 ecn_flags; /* ECN status bits. */ 6389 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6390 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6391 __u32 packets_out; /* Packets which are "in flight" */ 6392 __u32 retrans_out; /* Retransmitted packets out */ 6393 __u32 total_retrans; /* Total retransmits for entire connection */ 6394 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6395 * total number of segments in. 6396 */ 6397 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6398 * total number of data segments in. 6399 */ 6400 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6401 * The total number of segments sent. 6402 */ 6403 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6404 * total number of data segments sent. 6405 */ 6406 __u32 lost_out; /* Lost packets */ 6407 __u32 sacked_out; /* SACK'd packets */ 6408 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6409 * sum(delta(rcv_nxt)), or how many bytes 6410 * were acked. 6411 */ 6412 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6413 * sum(delta(snd_una)), or how many bytes 6414 * were acked. 6415 */ 6416 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6417 * total number of DSACK blocks received 6418 */ 6419 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6420 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6421 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6422 }; 6423 6424 struct bpf_sock_tuple { 6425 union { 6426 struct { 6427 __be32 saddr; 6428 __be32 daddr; 6429 __be16 sport; 6430 __be16 dport; 6431 } ipv4; 6432 struct { 6433 __be32 saddr[4]; 6434 __be32 daddr[4]; 6435 __be16 sport; 6436 __be16 dport; 6437 } ipv6; 6438 }; 6439 }; 6440 6441 /* (Simplified) user return codes for tcx prog type. 6442 * A valid tcx program must return one of these defined values. All other 6443 * return codes are reserved for future use. Must remain compatible with 6444 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6445 * return codes are mapped to TCX_NEXT. 6446 */ 6447 enum tcx_action_base { 6448 TCX_NEXT = -1, 6449 TCX_PASS = 0, 6450 TCX_DROP = 2, 6451 TCX_REDIRECT = 7, 6452 }; 6453 6454 struct bpf_xdp_sock { 6455 __u32 queue_id; 6456 }; 6457 6458 #define XDP_PACKET_HEADROOM 256 6459 6460 /* User return codes for XDP prog type. 6461 * A valid XDP program must return one of these defined values. All other 6462 * return codes are reserved for future use. Unknown return codes will 6463 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6464 */ 6465 enum xdp_action { 6466 XDP_ABORTED = 0, 6467 XDP_DROP, 6468 XDP_PASS, 6469 XDP_TX, 6470 XDP_REDIRECT, 6471 }; 6472 6473 /* user accessible metadata for XDP packet hook 6474 * new fields must be added to the end of this structure 6475 */ 6476 struct xdp_md { 6477 __u32 data; 6478 __u32 data_end; 6479 __u32 data_meta; 6480 /* Below access go through struct xdp_rxq_info */ 6481 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6482 __u32 rx_queue_index; /* rxq->queue_index */ 6483 6484 __u32 egress_ifindex; /* txq->dev->ifindex */ 6485 }; 6486 6487 /* DEVMAP 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_devmap_val { 6493 __u32 ifindex; /* device index */ 6494 union { 6495 int fd; /* prog fd on map write */ 6496 __u32 id; /* prog id on map read */ 6497 } bpf_prog; 6498 }; 6499 6500 /* CPUMAP map-value layout 6501 * 6502 * The struct data-layout of map-value is a configuration interface. 6503 * New members can only be added to the end of this structure. 6504 */ 6505 struct bpf_cpumap_val { 6506 __u32 qsize; /* queue size to remote target CPU */ 6507 union { 6508 int fd; /* prog fd on map write */ 6509 __u32 id; /* prog id on map read */ 6510 } bpf_prog; 6511 }; 6512 6513 enum sk_action { 6514 SK_DROP = 0, 6515 SK_PASS, 6516 }; 6517 6518 /* user accessible metadata for SK_MSG packet hook, new fields must 6519 * be added to the end of this structure 6520 */ 6521 struct sk_msg_md { 6522 __bpf_md_ptr(void *, data); 6523 __bpf_md_ptr(void *, data_end); 6524 6525 __u32 family; 6526 __u32 remote_ip4; /* Stored in network byte order */ 6527 __u32 local_ip4; /* Stored in network byte order */ 6528 __u32 remote_ip6[4]; /* Stored in network byte order */ 6529 __u32 local_ip6[4]; /* Stored in network byte order */ 6530 __u32 remote_port; /* Stored in network byte order */ 6531 __u32 local_port; /* stored in host byte order */ 6532 __u32 size; /* Total size of sk_msg */ 6533 6534 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6535 }; 6536 6537 struct sk_reuseport_md { 6538 /* 6539 * Start of directly accessible data. It begins from 6540 * the tcp/udp header. 6541 */ 6542 __bpf_md_ptr(void *, data); 6543 /* End of directly accessible data */ 6544 __bpf_md_ptr(void *, data_end); 6545 /* 6546 * Total length of packet (starting from the tcp/udp header). 6547 * Note that the directly accessible bytes (data_end - data) 6548 * could be less than this "len". Those bytes could be 6549 * indirectly read by a helper "bpf_skb_load_bytes()". 6550 */ 6551 __u32 len; 6552 /* 6553 * Eth protocol in the mac header (network byte order). e.g. 6554 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6555 */ 6556 __u32 eth_protocol; 6557 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6558 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6559 __u32 hash; /* A hash of the packet 4 tuples */ 6560 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6561 * new incoming connection request (e.g. selecting a listen sk for 6562 * the received SYN in the TCP case). reuse->sk is one of the sk 6563 * in the reuseport group. The bpf prog can use reuse->sk to learn 6564 * the local listening ip/port without looking into the skb. 6565 * 6566 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6567 * reuse->migrating_sk is the socket that needs to be migrated 6568 * to another listening socket. migrating_sk could be a fullsock 6569 * sk that is fully established or a reqsk that is in-the-middle 6570 * of 3-way handshake. 6571 */ 6572 __bpf_md_ptr(struct bpf_sock *, sk); 6573 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6574 }; 6575 6576 #define BPF_TAG_SIZE 8 6577 6578 struct bpf_prog_info { 6579 __u32 type; 6580 __u32 id; 6581 __u8 tag[BPF_TAG_SIZE]; 6582 __u32 jited_prog_len; 6583 __u32 xlated_prog_len; 6584 __aligned_u64 jited_prog_insns; 6585 __aligned_u64 xlated_prog_insns; 6586 __u64 load_time; /* ns since boottime */ 6587 __u32 created_by_uid; 6588 __u32 nr_map_ids; 6589 __aligned_u64 map_ids; 6590 char name[BPF_OBJ_NAME_LEN]; 6591 __u32 ifindex; 6592 __u32 gpl_compatible:1; 6593 __u32 :31; /* alignment pad */ 6594 __u64 netns_dev; 6595 __u64 netns_ino; 6596 __u32 nr_jited_ksyms; 6597 __u32 nr_jited_func_lens; 6598 __aligned_u64 jited_ksyms; 6599 __aligned_u64 jited_func_lens; 6600 __u32 btf_id; 6601 __u32 func_info_rec_size; 6602 __aligned_u64 func_info; 6603 __u32 nr_func_info; 6604 __u32 nr_line_info; 6605 __aligned_u64 line_info; 6606 __aligned_u64 jited_line_info; 6607 __u32 nr_jited_line_info; 6608 __u32 line_info_rec_size; 6609 __u32 jited_line_info_rec_size; 6610 __u32 nr_prog_tags; 6611 __aligned_u64 prog_tags; 6612 __u64 run_time_ns; 6613 __u64 run_cnt; 6614 __u64 recursion_misses; 6615 __u32 verified_insns; 6616 __u32 attach_btf_obj_id; 6617 __u32 attach_btf_id; 6618 } __attribute__((aligned(8))); 6619 6620 struct bpf_map_info { 6621 __u32 type; 6622 __u32 id; 6623 __u32 key_size; 6624 __u32 value_size; 6625 __u32 max_entries; 6626 __u32 map_flags; 6627 char name[BPF_OBJ_NAME_LEN]; 6628 __u32 ifindex; 6629 __u32 btf_vmlinux_value_type_id; 6630 __u64 netns_dev; 6631 __u64 netns_ino; 6632 __u32 btf_id; 6633 __u32 btf_key_type_id; 6634 __u32 btf_value_type_id; 6635 __u32 btf_vmlinux_id; 6636 __u64 map_extra; 6637 } __attribute__((aligned(8))); 6638 6639 struct bpf_btf_info { 6640 __aligned_u64 btf; 6641 __u32 btf_size; 6642 __u32 id; 6643 __aligned_u64 name; 6644 __u32 name_len; 6645 __u32 kernel_btf; 6646 } __attribute__((aligned(8))); 6647 6648 struct bpf_link_info { 6649 __u32 type; 6650 __u32 id; 6651 __u32 prog_id; 6652 union { 6653 struct { 6654 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6655 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6656 } raw_tracepoint; 6657 struct { 6658 __u32 attach_type; 6659 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6660 __u32 target_btf_id; /* BTF type id inside the object */ 6661 } tracing; 6662 struct { 6663 __u64 cgroup_id; 6664 __u32 attach_type; 6665 } cgroup; 6666 struct { 6667 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6668 __u32 target_name_len; /* in/out: target_name buffer len */ 6669 6670 /* If the iter specific field is 32 bits, it can be put 6671 * in the first or second union. Otherwise it should be 6672 * put in the second union. 6673 */ 6674 union { 6675 struct { 6676 __u32 map_id; 6677 } map; 6678 }; 6679 union { 6680 struct { 6681 __u64 cgroup_id; 6682 __u32 order; 6683 } cgroup; 6684 struct { 6685 __u32 tid; 6686 __u32 pid; 6687 } task; 6688 }; 6689 } iter; 6690 struct { 6691 __u32 netns_ino; 6692 __u32 attach_type; 6693 } netns; 6694 struct { 6695 __u32 ifindex; 6696 } xdp; 6697 struct { 6698 __u32 map_id; 6699 } struct_ops; 6700 struct { 6701 __u32 pf; 6702 __u32 hooknum; 6703 __s32 priority; 6704 __u32 flags; 6705 } netfilter; 6706 struct { 6707 __aligned_u64 addrs; 6708 __u32 count; /* in/out: kprobe_multi function count */ 6709 __u32 flags; 6710 __u64 missed; 6711 __aligned_u64 cookies; 6712 } kprobe_multi; 6713 struct { 6714 __aligned_u64 path; 6715 __aligned_u64 offsets; 6716 __aligned_u64 ref_ctr_offsets; 6717 __aligned_u64 cookies; 6718 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6719 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6720 __u32 flags; 6721 __u32 pid; 6722 } uprobe_multi; 6723 struct { 6724 __u32 type; /* enum bpf_perf_event_type */ 6725 __u32 :32; 6726 union { 6727 struct { 6728 __aligned_u64 file_name; /* in/out */ 6729 __u32 name_len; 6730 __u32 offset; /* offset from file_name */ 6731 __u64 cookie; 6732 __u64 ref_ctr_offset; 6733 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6734 struct { 6735 __aligned_u64 func_name; /* in/out */ 6736 __u32 name_len; 6737 __u32 offset; /* offset from func_name */ 6738 __u64 addr; 6739 __u64 missed; 6740 __u64 cookie; 6741 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6742 struct { 6743 __aligned_u64 tp_name; /* in/out */ 6744 __u32 name_len; 6745 __u32 :32; 6746 __u64 cookie; 6747 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6748 struct { 6749 __u64 config; 6750 __u32 type; 6751 __u32 :32; 6752 __u64 cookie; 6753 } event; /* BPF_PERF_EVENT_EVENT */ 6754 }; 6755 } perf_event; 6756 struct { 6757 __u32 ifindex; 6758 __u32 attach_type; 6759 } tcx; 6760 struct { 6761 __u32 ifindex; 6762 __u32 attach_type; 6763 } netkit; 6764 struct { 6765 __u32 map_id; 6766 __u32 attach_type; 6767 } sockmap; 6768 }; 6769 } __attribute__((aligned(8))); 6770 6771 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6772 * by user and intended to be used by socket (e.g. to bind to, depends on 6773 * attach type). 6774 */ 6775 struct bpf_sock_addr { 6776 __u32 user_family; /* Allows 4-byte read, but no write. */ 6777 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6778 * Stored in network byte order. 6779 */ 6780 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6781 * Stored in network byte order. 6782 */ 6783 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6784 * Stored in network byte order 6785 */ 6786 __u32 family; /* Allows 4-byte read, but no write */ 6787 __u32 type; /* Allows 4-byte read, but no write */ 6788 __u32 protocol; /* Allows 4-byte read, but no write */ 6789 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6790 * Stored in network byte order. 6791 */ 6792 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6793 * Stored in network byte order. 6794 */ 6795 __bpf_md_ptr(struct bpf_sock *, sk); 6796 }; 6797 6798 /* User bpf_sock_ops struct to access socket values and specify request ops 6799 * and their replies. 6800 * Some of this fields are in network (bigendian) byte order and may need 6801 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6802 * New fields can only be added at the end of this structure 6803 */ 6804 struct bpf_sock_ops { 6805 __u32 op; 6806 union { 6807 __u32 args[4]; /* Optionally passed to bpf program */ 6808 __u32 reply; /* Returned by bpf program */ 6809 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6810 }; 6811 __u32 family; 6812 __u32 remote_ip4; /* Stored in network byte order */ 6813 __u32 local_ip4; /* Stored in network byte order */ 6814 __u32 remote_ip6[4]; /* Stored in network byte order */ 6815 __u32 local_ip6[4]; /* Stored in network byte order */ 6816 __u32 remote_port; /* Stored in network byte order */ 6817 __u32 local_port; /* stored in host byte order */ 6818 __u32 is_fullsock; /* Some TCP fields are only valid if 6819 * there is a full socket. If not, the 6820 * fields read as zero. 6821 */ 6822 __u32 snd_cwnd; 6823 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6824 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6825 __u32 state; 6826 __u32 rtt_min; 6827 __u32 snd_ssthresh; 6828 __u32 rcv_nxt; 6829 __u32 snd_nxt; 6830 __u32 snd_una; 6831 __u32 mss_cache; 6832 __u32 ecn_flags; 6833 __u32 rate_delivered; 6834 __u32 rate_interval_us; 6835 __u32 packets_out; 6836 __u32 retrans_out; 6837 __u32 total_retrans; 6838 __u32 segs_in; 6839 __u32 data_segs_in; 6840 __u32 segs_out; 6841 __u32 data_segs_out; 6842 __u32 lost_out; 6843 __u32 sacked_out; 6844 __u32 sk_txhash; 6845 __u64 bytes_received; 6846 __u64 bytes_acked; 6847 __bpf_md_ptr(struct bpf_sock *, sk); 6848 /* [skb_data, skb_data_end) covers the whole TCP header. 6849 * 6850 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6851 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6852 * header has not been written. 6853 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6854 * been written so far. 6855 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6856 * the 3WHS. 6857 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6858 * the 3WHS. 6859 * 6860 * bpf_load_hdr_opt() can also be used to read a particular option. 6861 */ 6862 __bpf_md_ptr(void *, skb_data); 6863 __bpf_md_ptr(void *, skb_data_end); 6864 __u32 skb_len; /* The total length of a packet. 6865 * It includes the header, options, 6866 * and payload. 6867 */ 6868 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6869 * an easy way to check for tcp_flags 6870 * without parsing skb_data. 6871 * 6872 * In particular, the skb_tcp_flags 6873 * will still be available in 6874 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6875 * the outgoing header has not 6876 * been written yet. 6877 */ 6878 __u64 skb_hwtstamp; 6879 }; 6880 6881 /* Definitions for bpf_sock_ops_cb_flags */ 6882 enum { 6883 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6884 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6885 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6886 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6887 /* Call bpf for all received TCP headers. The bpf prog will be 6888 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6889 * 6890 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6891 * for the header option related helpers that will be useful 6892 * to the bpf programs. 6893 * 6894 * It could be used at the client/active side (i.e. connect() side) 6895 * when the server told it that the server was in syncookie 6896 * mode and required the active side to resend the bpf-written 6897 * options. The active side can keep writing the bpf-options until 6898 * it received a valid packet from the server side to confirm 6899 * the earlier packet (and options) has been received. The later 6900 * example patch is using it like this at the active side when the 6901 * server is in syncookie mode. 6902 * 6903 * The bpf prog will usually turn this off in the common cases. 6904 */ 6905 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6906 /* Call bpf when kernel has received a header option that 6907 * the kernel cannot handle. The bpf prog will be called under 6908 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6909 * 6910 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6911 * for the header option related helpers that will be useful 6912 * to the bpf programs. 6913 */ 6914 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6915 /* Call bpf when the kernel is writing header options for the 6916 * outgoing packet. The bpf prog will first be called 6917 * to reserve space in a skb under 6918 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6919 * the bpf prog will be called to write the header option(s) 6920 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6921 * 6922 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6923 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6924 * related helpers that will be useful to the bpf programs. 6925 * 6926 * The kernel gets its chance to reserve space and write 6927 * options first before the BPF program does. 6928 */ 6929 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6930 /* Mask of all currently supported cb flags */ 6931 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6932 }; 6933 6934 enum { 6935 SK_BPF_CB_TX_TIMESTAMPING = 1<<0, 6936 SK_BPF_CB_MASK = (SK_BPF_CB_TX_TIMESTAMPING - 1) | 6937 SK_BPF_CB_TX_TIMESTAMPING 6938 }; 6939 6940 /* List of known BPF sock_ops operators. 6941 * New entries can only be added at the end 6942 */ 6943 enum { 6944 BPF_SOCK_OPS_VOID, 6945 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6946 * -1 if default value should be used 6947 */ 6948 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6949 * window (in packets) or -1 if default 6950 * value should be used 6951 */ 6952 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6953 * active connection is initialized 6954 */ 6955 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6956 * active connection is 6957 * established 6958 */ 6959 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6960 * passive connection is 6961 * established 6962 */ 6963 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6964 * needs ECN 6965 */ 6966 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6967 * based on the path and may be 6968 * dependent on the congestion control 6969 * algorithm. In general it indicates 6970 * a congestion threshold. RTTs above 6971 * this indicate congestion 6972 */ 6973 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6974 * Arg1: value of icsk_retransmits 6975 * Arg2: value of icsk_rto 6976 * Arg3: whether RTO has expired 6977 */ 6978 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6979 * Arg1: sequence number of 1st byte 6980 * Arg2: # segments 6981 * Arg3: return value of 6982 * tcp_transmit_skb (0 => success) 6983 */ 6984 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6985 * Arg1: old_state 6986 * Arg2: new_state 6987 */ 6988 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6989 * socket transition to LISTEN state. 6990 */ 6991 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6992 * Arg1: measured RTT input (mrtt) 6993 * Arg2: updated srtt 6994 */ 6995 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6996 * It will be called to handle 6997 * the packets received at 6998 * an already established 6999 * connection. 7000 * 7001 * sock_ops->skb_data: 7002 * Referring to the received skb. 7003 * It covers the TCP header only. 7004 * 7005 * bpf_load_hdr_opt() can also 7006 * be used to search for a 7007 * particular option. 7008 */ 7009 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 7010 * header option later in 7011 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7012 * Arg1: bool want_cookie. (in 7013 * writing SYNACK only) 7014 * 7015 * sock_ops->skb_data: 7016 * Not available because no header has 7017 * been written yet. 7018 * 7019 * sock_ops->skb_tcp_flags: 7020 * The tcp_flags of the 7021 * outgoing skb. (e.g. SYN, ACK, FIN). 7022 * 7023 * bpf_reserve_hdr_opt() should 7024 * be used to reserve space. 7025 */ 7026 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 7027 * Arg1: bool want_cookie. (in 7028 * writing SYNACK only) 7029 * 7030 * sock_ops->skb_data: 7031 * Referring to the outgoing skb. 7032 * It covers the TCP header 7033 * that has already been written 7034 * by the kernel and the 7035 * earlier bpf-progs. 7036 * 7037 * sock_ops->skb_tcp_flags: 7038 * The tcp_flags of the outgoing 7039 * skb. (e.g. SYN, ACK, FIN). 7040 * 7041 * bpf_store_hdr_opt() should 7042 * be used to write the 7043 * option. 7044 * 7045 * bpf_load_hdr_opt() can also 7046 * be used to search for a 7047 * particular option that 7048 * has already been written 7049 * by the kernel or the 7050 * earlier bpf-progs. 7051 */ 7052 BPF_SOCK_OPS_TSTAMP_SCHED_CB, /* Called when skb is passing 7053 * through dev layer when 7054 * SK_BPF_CB_TX_TIMESTAMPING 7055 * feature is on. 7056 */ 7057 BPF_SOCK_OPS_TSTAMP_SND_SW_CB, /* Called when skb is about to send 7058 * to the nic when SK_BPF_CB_TX_TIMESTAMPING 7059 * feature is on. 7060 */ 7061 BPF_SOCK_OPS_TSTAMP_SND_HW_CB, /* Called in hardware phase when 7062 * SK_BPF_CB_TX_TIMESTAMPING feature 7063 * is on. 7064 */ 7065 BPF_SOCK_OPS_TSTAMP_ACK_CB, /* Called when all the skbs in the 7066 * same sendmsg call are acked 7067 * when SK_BPF_CB_TX_TIMESTAMPING 7068 * feature is on. 7069 */ 7070 BPF_SOCK_OPS_TSTAMP_SENDMSG_CB, /* Called when every sendmsg syscall 7071 * is triggered. It's used to correlate 7072 * sendmsg timestamp with corresponding 7073 * tskey. 7074 */ 7075 }; 7076 7077 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 7078 * changes between the TCP and BPF versions. Ideally this should never happen. 7079 * If it does, we need to add code to convert them before calling 7080 * the BPF sock_ops function. 7081 */ 7082 enum { 7083 BPF_TCP_ESTABLISHED = 1, 7084 BPF_TCP_SYN_SENT, 7085 BPF_TCP_SYN_RECV, 7086 BPF_TCP_FIN_WAIT1, 7087 BPF_TCP_FIN_WAIT2, 7088 BPF_TCP_TIME_WAIT, 7089 BPF_TCP_CLOSE, 7090 BPF_TCP_CLOSE_WAIT, 7091 BPF_TCP_LAST_ACK, 7092 BPF_TCP_LISTEN, 7093 BPF_TCP_CLOSING, /* Now a valid state */ 7094 BPF_TCP_NEW_SYN_RECV, 7095 BPF_TCP_BOUND_INACTIVE, 7096 7097 BPF_TCP_MAX_STATES /* Leave at the end! */ 7098 }; 7099 7100 enum { 7101 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 7102 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 7103 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 7104 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 7105 /* Copy the SYN pkt to optval 7106 * 7107 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 7108 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 7109 * to only getting from the saved_syn. It can either get the 7110 * syn packet from: 7111 * 7112 * 1. the just-received SYN packet (only available when writing the 7113 * SYNACK). It will be useful when it is not necessary to 7114 * save the SYN packet for latter use. It is also the only way 7115 * to get the SYN during syncookie mode because the syn 7116 * packet cannot be saved during syncookie. 7117 * 7118 * OR 7119 * 7120 * 2. the earlier saved syn which was done by 7121 * bpf_setsockopt(TCP_SAVE_SYN). 7122 * 7123 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7124 * SYN packet is obtained. 7125 * 7126 * If the bpf-prog does not need the IP[46] header, the 7127 * bpf-prog can avoid parsing the IP header by using 7128 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7129 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7130 * 7131 * >0: Total number of bytes copied 7132 * -ENOSPC: Not enough space in optval. Only optlen number of 7133 * bytes is copied. 7134 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7135 * is not saved by setsockopt(TCP_SAVE_SYN). 7136 */ 7137 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7138 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7139 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7140 TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */ 7141 SK_BPF_CB_FLAGS = 1009, /* Get or set sock ops flags in socket */ 7142 }; 7143 7144 enum { 7145 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7146 }; 7147 7148 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7149 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7150 */ 7151 enum { 7152 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7153 * total option spaces 7154 * required for an established 7155 * sk in order to calculate the 7156 * MSS. No skb is actually 7157 * sent. 7158 */ 7159 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7160 * when sending a SYN. 7161 */ 7162 }; 7163 7164 struct bpf_perf_event_value { 7165 __u64 counter; 7166 __u64 enabled; 7167 __u64 running; 7168 }; 7169 7170 enum { 7171 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7172 BPF_DEVCG_ACC_READ = (1ULL << 1), 7173 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7174 }; 7175 7176 enum { 7177 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7178 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7179 }; 7180 7181 struct bpf_cgroup_dev_ctx { 7182 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7183 __u32 access_type; 7184 __u32 major; 7185 __u32 minor; 7186 }; 7187 7188 struct bpf_raw_tracepoint_args { 7189 __u64 args[0]; 7190 }; 7191 7192 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7193 * OUTPUT: Do lookup from egress perspective; default is ingress 7194 */ 7195 enum { 7196 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7197 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7198 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7199 BPF_FIB_LOOKUP_TBID = (1U << 3), 7200 BPF_FIB_LOOKUP_SRC = (1U << 4), 7201 BPF_FIB_LOOKUP_MARK = (1U << 5), 7202 }; 7203 7204 enum { 7205 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7206 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7207 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7208 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7209 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7210 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7211 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7212 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7213 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7214 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7215 }; 7216 7217 struct bpf_fib_lookup { 7218 /* input: network family for lookup (AF_INET, AF_INET6) 7219 * output: network family of egress nexthop 7220 */ 7221 __u8 family; 7222 7223 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7224 __u8 l4_protocol; 7225 __be16 sport; 7226 __be16 dport; 7227 7228 union { /* used for MTU check */ 7229 /* input to lookup */ 7230 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7231 7232 /* output: MTU value */ 7233 __u16 mtu_result; 7234 } __attribute__((packed, aligned(2))); 7235 /* input: L3 device index for lookup 7236 * output: device index from FIB lookup 7237 */ 7238 __u32 ifindex; 7239 7240 union { 7241 /* inputs to lookup */ 7242 __u8 tos; /* AF_INET */ 7243 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7244 7245 /* output: metric of fib result (IPv4/IPv6 only) */ 7246 __u32 rt_metric; 7247 }; 7248 7249 /* input: source address to consider for lookup 7250 * output: source address result from lookup 7251 */ 7252 union { 7253 __be32 ipv4_src; 7254 __u32 ipv6_src[4]; /* in6_addr; network order */ 7255 }; 7256 7257 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7258 * network header. output: bpf_fib_lookup sets to gateway address 7259 * if FIB lookup returns gateway route 7260 */ 7261 union { 7262 __be32 ipv4_dst; 7263 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7264 }; 7265 7266 union { 7267 struct { 7268 /* output */ 7269 __be16 h_vlan_proto; 7270 __be16 h_vlan_TCI; 7271 }; 7272 /* input: when accompanied with the 7273 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7274 * specific routing table to use for the fib lookup. 7275 */ 7276 __u32 tbid; 7277 }; 7278 7279 union { 7280 /* input */ 7281 struct { 7282 __u32 mark; /* policy routing */ 7283 /* 2 4-byte holes for input */ 7284 }; 7285 7286 /* output: source and dest mac */ 7287 struct { 7288 __u8 smac[6]; /* ETH_ALEN */ 7289 __u8 dmac[6]; /* ETH_ALEN */ 7290 }; 7291 }; 7292 }; 7293 7294 struct bpf_redir_neigh { 7295 /* network family for lookup (AF_INET, AF_INET6) */ 7296 __u32 nh_family; 7297 /* network address of nexthop; skips fib lookup to find gateway */ 7298 union { 7299 __be32 ipv4_nh; 7300 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7301 }; 7302 }; 7303 7304 /* bpf_check_mtu flags*/ 7305 enum bpf_check_mtu_flags { 7306 BPF_MTU_CHK_SEGS = (1U << 0), 7307 }; 7308 7309 enum bpf_check_mtu_ret { 7310 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7311 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7312 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7313 }; 7314 7315 enum bpf_task_fd_type { 7316 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7317 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7318 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7319 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7320 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7321 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7322 }; 7323 7324 enum { 7325 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7326 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7327 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7328 }; 7329 7330 struct bpf_flow_keys { 7331 __u16 nhoff; 7332 __u16 thoff; 7333 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7334 __u8 is_frag; 7335 __u8 is_first_frag; 7336 __u8 is_encap; 7337 __u8 ip_proto; 7338 __be16 n_proto; 7339 __be16 sport; 7340 __be16 dport; 7341 union { 7342 struct { 7343 __be32 ipv4_src; 7344 __be32 ipv4_dst; 7345 }; 7346 struct { 7347 __u32 ipv6_src[4]; /* in6_addr; network order */ 7348 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7349 }; 7350 }; 7351 __u32 flags; 7352 __be32 flow_label; 7353 }; 7354 7355 struct bpf_func_info { 7356 __u32 insn_off; 7357 __u32 type_id; 7358 }; 7359 7360 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7361 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7362 7363 struct bpf_line_info { 7364 __u32 insn_off; 7365 __u32 file_name_off; 7366 __u32 line_off; 7367 __u32 line_col; 7368 }; 7369 7370 struct bpf_spin_lock { 7371 __u32 val; 7372 }; 7373 7374 struct bpf_timer { 7375 __u64 __opaque[2]; 7376 } __attribute__((aligned(8))); 7377 7378 struct bpf_wq { 7379 __u64 __opaque[2]; 7380 } __attribute__((aligned(8))); 7381 7382 struct bpf_dynptr { 7383 __u64 __opaque[2]; 7384 } __attribute__((aligned(8))); 7385 7386 struct bpf_list_head { 7387 __u64 __opaque[2]; 7388 } __attribute__((aligned(8))); 7389 7390 struct bpf_list_node { 7391 __u64 __opaque[3]; 7392 } __attribute__((aligned(8))); 7393 7394 struct bpf_rb_root { 7395 __u64 __opaque[2]; 7396 } __attribute__((aligned(8))); 7397 7398 struct bpf_rb_node { 7399 __u64 __opaque[4]; 7400 } __attribute__((aligned(8))); 7401 7402 struct bpf_refcount { 7403 __u32 __opaque[1]; 7404 } __attribute__((aligned(4))); 7405 7406 struct bpf_sysctl { 7407 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7408 * Allows 1,2,4-byte read, but no write. 7409 */ 7410 __u32 file_pos; /* Sysctl file position to read from, write to. 7411 * Allows 1,2,4-byte read an 4-byte write. 7412 */ 7413 }; 7414 7415 struct bpf_sockopt { 7416 __bpf_md_ptr(struct bpf_sock *, sk); 7417 __bpf_md_ptr(void *, optval); 7418 __bpf_md_ptr(void *, optval_end); 7419 7420 __s32 level; 7421 __s32 optname; 7422 __s32 optlen; 7423 __s32 retval; 7424 }; 7425 7426 struct bpf_pidns_info { 7427 __u32 pid; 7428 __u32 tgid; 7429 }; 7430 7431 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7432 struct bpf_sk_lookup { 7433 union { 7434 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7435 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7436 }; 7437 7438 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7439 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7440 __u32 remote_ip4; /* Network byte order */ 7441 __u32 remote_ip6[4]; /* Network byte order */ 7442 __be16 remote_port; /* Network byte order */ 7443 __u16 :16; /* Zero padding */ 7444 __u32 local_ip4; /* Network byte order */ 7445 __u32 local_ip6[4]; /* Network byte order */ 7446 __u32 local_port; /* Host byte order */ 7447 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7448 }; 7449 7450 /* 7451 * struct btf_ptr is used for typed pointer representation; the 7452 * type id is used to render the pointer data as the appropriate type 7453 * via the bpf_snprintf_btf() helper described above. A flags field - 7454 * potentially to specify additional details about the BTF pointer 7455 * (rather than its mode of display) - is included for future use. 7456 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7457 */ 7458 struct btf_ptr { 7459 void *ptr; 7460 __u32 type_id; 7461 __u32 flags; /* BTF ptr flags; unused at present. */ 7462 }; 7463 7464 /* 7465 * Flags to control bpf_snprintf_btf() behaviour. 7466 * - BTF_F_COMPACT: no formatting around type information 7467 * - BTF_F_NONAME: no struct/union member names/types 7468 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7469 * equivalent to %px. 7470 * - BTF_F_ZERO: show zero-valued struct/union members; they 7471 * are not displayed by default 7472 */ 7473 enum { 7474 BTF_F_COMPACT = (1ULL << 0), 7475 BTF_F_NONAME = (1ULL << 1), 7476 BTF_F_PTR_RAW = (1ULL << 2), 7477 BTF_F_ZERO = (1ULL << 3), 7478 }; 7479 7480 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7481 * has to be adjusted by relocations. It is emitted by llvm and passed to 7482 * libbpf and later to the kernel. 7483 */ 7484 enum bpf_core_relo_kind { 7485 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7486 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7487 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7488 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7489 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7490 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7491 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7492 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7493 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7494 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7495 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7496 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7497 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7498 }; 7499 7500 /* 7501 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7502 * and from libbpf to the kernel. 7503 * 7504 * CO-RE relocation captures the following data: 7505 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7506 * its insn->imm field to be relocated with actual field info; 7507 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7508 * type or field; 7509 * - access_str_off - offset into corresponding .BTF string section. String 7510 * interpretation depends on specific relocation kind: 7511 * - for field-based relocations, string encodes an accessed field using 7512 * a sequence of field and array indices, separated by colon (:). It's 7513 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7514 * arguments for identifying offset to a field. 7515 * - for type-based relocations, strings is expected to be just "0"; 7516 * - for enum value-based relocations, string contains an index of enum 7517 * value within its enum type; 7518 * - kind - one of enum bpf_core_relo_kind; 7519 * 7520 * Example: 7521 * struct sample { 7522 * int a; 7523 * struct { 7524 * int b[10]; 7525 * }; 7526 * }; 7527 * 7528 * struct sample *s = ...; 7529 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7530 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7531 * // b is field #0 inside anon struct, accessing elem #5) 7532 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7533 * 7534 * type_id for all relocs in this example will capture BTF type id of 7535 * `struct sample`. 7536 * 7537 * Such relocation is emitted when using __builtin_preserve_access_index() 7538 * Clang built-in, passing expression that captures field address, e.g.: 7539 * 7540 * bpf_probe_read(&dst, sizeof(dst), 7541 * __builtin_preserve_access_index(&src->a.b.c)); 7542 * 7543 * In this case Clang will emit field relocation recording necessary data to 7544 * be able to find offset of embedded `a.b.c` field within `src` struct. 7545 * 7546 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7547 */ 7548 struct bpf_core_relo { 7549 __u32 insn_off; 7550 __u32 type_id; 7551 __u32 access_str_off; 7552 enum bpf_core_relo_kind kind; 7553 }; 7554 7555 /* 7556 * Flags to control bpf_timer_start() behaviour. 7557 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7558 * relative to current time. 7559 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7560 */ 7561 enum { 7562 BPF_F_TIMER_ABS = (1ULL << 0), 7563 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7564 }; 7565 7566 /* BPF numbers iterator state */ 7567 struct bpf_iter_num { 7568 /* opaque iterator state; having __u64 here allows to preserve correct 7569 * alignment requirements in vmlinux.h, generated from BTF 7570 */ 7571 __u64 __opaque[1]; 7572 } __attribute__((aligned(8))); 7573 7574 /* 7575 * Flags to control BPF kfunc behaviour. 7576 * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective 7577 * helper documentation for details.) 7578 */ 7579 enum bpf_kfunc_flags { 7580 BPF_F_PAD_ZEROS = (1ULL << 0), 7581 }; 7582 7583 #endif /* _UAPI__LINUX_BPF_H__ */ 7584