1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_MEMSX 0x80 /* load with sign extension */ 23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 25 26 /* alu/jmp fields */ 27 #define BPF_MOV 0xb0 /* mov reg to reg */ 28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 29 30 /* change endianness of a register */ 31 #define BPF_END 0xd0 /* flags for endianness conversion: */ 32 #define BPF_TO_LE 0x00 /* convert to little-endian */ 33 #define BPF_TO_BE 0x08 /* convert to big-endian */ 34 #define BPF_FROM_LE BPF_TO_LE 35 #define BPF_FROM_BE BPF_TO_BE 36 37 /* jmp encodings */ 38 #define BPF_JNE 0x50 /* jump != */ 39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 45 #define BPF_CALL 0x80 /* function call */ 46 #define BPF_EXIT 0x90 /* function return */ 47 48 /* atomic op type fields (stored in immediate) */ 49 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 50 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 51 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 52 53 /* Register numbers */ 54 enum { 55 BPF_REG_0 = 0, 56 BPF_REG_1, 57 BPF_REG_2, 58 BPF_REG_3, 59 BPF_REG_4, 60 BPF_REG_5, 61 BPF_REG_6, 62 BPF_REG_7, 63 BPF_REG_8, 64 BPF_REG_9, 65 BPF_REG_10, 66 __MAX_BPF_REG, 67 }; 68 69 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 70 #define MAX_BPF_REG __MAX_BPF_REG 71 72 struct bpf_insn { 73 __u8 code; /* opcode */ 74 __u8 dst_reg:4; /* dest register */ 75 __u8 src_reg:4; /* source register */ 76 __s16 off; /* signed offset */ 77 __s32 imm; /* signed immediate constant */ 78 }; 79 80 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 81 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 82 * the trailing flexible array member) instead. 83 */ 84 struct bpf_lpm_trie_key { 85 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 86 __u8 data[0]; /* Arbitrary size */ 87 }; 88 89 /* Header for bpf_lpm_trie_key structs */ 90 struct bpf_lpm_trie_key_hdr { 91 __u32 prefixlen; 92 }; 93 94 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 95 struct bpf_lpm_trie_key_u8 { 96 union { 97 struct bpf_lpm_trie_key_hdr hdr; 98 __u32 prefixlen; 99 }; 100 __u8 data[]; /* Arbitrary size */ 101 }; 102 103 struct bpf_cgroup_storage_key { 104 __u64 cgroup_inode_id; /* cgroup inode id */ 105 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 106 }; 107 108 enum bpf_cgroup_iter_order { 109 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 110 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 111 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 112 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 113 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 114 }; 115 116 union bpf_iter_link_info { 117 struct { 118 __u32 map_fd; 119 } map; 120 struct { 121 enum bpf_cgroup_iter_order order; 122 123 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 124 * both are zero, the walk starts from the default cgroup v2 125 * root. For walking v1 hierarchy, one should always explicitly 126 * specify cgroup_fd. 127 */ 128 __u32 cgroup_fd; 129 __u64 cgroup_id; 130 } cgroup; 131 /* Parameters of task iterators. */ 132 struct { 133 __u32 tid; 134 __u32 pid; 135 __u32 pid_fd; 136 } task; 137 }; 138 139 /* BPF syscall commands, see bpf(2) man-page for more details. */ 140 /** 141 * DOC: eBPF Syscall Preamble 142 * 143 * The operation to be performed by the **bpf**\ () system call is determined 144 * by the *cmd* argument. Each operation takes an accompanying argument, 145 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 146 * below). The size argument is the size of the union pointed to by *attr*. 147 */ 148 /** 149 * DOC: eBPF Syscall Commands 150 * 151 * BPF_MAP_CREATE 152 * Description 153 * Create a map and return a file descriptor that refers to the 154 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 155 * is automatically enabled for the new file descriptor. 156 * 157 * Applying **close**\ (2) to the file descriptor returned by 158 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 159 * 160 * Return 161 * A new file descriptor (a nonnegative integer), or -1 if an 162 * error occurred (in which case, *errno* is set appropriately). 163 * 164 * BPF_MAP_LOOKUP_ELEM 165 * Description 166 * Look up an element with a given *key* in the map referred to 167 * by the file descriptor *map_fd*. 168 * 169 * The *flags* argument may be specified as one of the 170 * following: 171 * 172 * **BPF_F_LOCK** 173 * Look up the value of a spin-locked map without 174 * returning the lock. This must be specified if the 175 * elements contain a spinlock. 176 * 177 * Return 178 * Returns zero on success. On error, -1 is returned and *errno* 179 * is set appropriately. 180 * 181 * BPF_MAP_UPDATE_ELEM 182 * Description 183 * Create or update an element (key/value pair) in a specified map. 184 * 185 * The *flags* argument should be specified as one of the 186 * following: 187 * 188 * **BPF_ANY** 189 * Create a new element or update an existing element. 190 * **BPF_NOEXIST** 191 * Create a new element only if it did not exist. 192 * **BPF_EXIST** 193 * Update an existing element. 194 * **BPF_F_LOCK** 195 * Update a spin_lock-ed map element. 196 * 197 * Return 198 * Returns zero on success. On error, -1 is returned and *errno* 199 * is set appropriately. 200 * 201 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 202 * **E2BIG**, **EEXIST**, or **ENOENT**. 203 * 204 * **E2BIG** 205 * The number of elements in the map reached the 206 * *max_entries* limit specified at map creation time. 207 * **EEXIST** 208 * If *flags* specifies **BPF_NOEXIST** and the element 209 * with *key* already exists in the map. 210 * **ENOENT** 211 * If *flags* specifies **BPF_EXIST** and the element with 212 * *key* does not exist in the map. 213 * 214 * BPF_MAP_DELETE_ELEM 215 * Description 216 * Look up and delete an element by key in a specified map. 217 * 218 * Return 219 * Returns zero on success. On error, -1 is returned and *errno* 220 * is set appropriately. 221 * 222 * BPF_MAP_GET_NEXT_KEY 223 * Description 224 * Look up an element by key in a specified map and return the key 225 * of the next element. Can be used to iterate over all elements 226 * in the map. 227 * 228 * Return 229 * Returns zero on success. On error, -1 is returned and *errno* 230 * is set appropriately. 231 * 232 * The following cases can be used to iterate over all elements of 233 * the map: 234 * 235 * * If *key* is not found, the operation returns zero and sets 236 * the *next_key* pointer to the key of the first element. 237 * * If *key* is found, the operation returns zero and sets the 238 * *next_key* pointer to the key of the next element. 239 * * If *key* is the last element, returns -1 and *errno* is set 240 * to **ENOENT**. 241 * 242 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 243 * **EINVAL** on error. 244 * 245 * BPF_PROG_LOAD 246 * Description 247 * Verify and load an eBPF program, returning a new file 248 * descriptor associated with the program. 249 * 250 * Applying **close**\ (2) to the file descriptor returned by 251 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 252 * 253 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 254 * automatically enabled for the new file descriptor. 255 * 256 * Return 257 * A new file descriptor (a nonnegative integer), or -1 if an 258 * error occurred (in which case, *errno* is set appropriately). 259 * 260 * BPF_OBJ_PIN 261 * Description 262 * Pin an eBPF program or map referred by the specified *bpf_fd* 263 * to the provided *pathname* on the filesystem. 264 * 265 * The *pathname* argument must not contain a dot ("."). 266 * 267 * On success, *pathname* retains a reference to the eBPF object, 268 * preventing deallocation of the object when the original 269 * *bpf_fd* is closed. This allow the eBPF object to live beyond 270 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 271 * process. 272 * 273 * Applying **unlink**\ (2) or similar calls to the *pathname* 274 * unpins the object from the filesystem, removing the reference. 275 * If no other file descriptors or filesystem nodes refer to the 276 * same object, it will be deallocated (see NOTES). 277 * 278 * The filesystem type for the parent directory of *pathname* must 279 * be **BPF_FS_MAGIC**. 280 * 281 * Return 282 * Returns zero on success. On error, -1 is returned and *errno* 283 * is set appropriately. 284 * 285 * BPF_OBJ_GET 286 * Description 287 * Open a file descriptor for the eBPF object pinned to the 288 * specified *pathname*. 289 * 290 * Return 291 * A new file descriptor (a nonnegative integer), or -1 if an 292 * error occurred (in which case, *errno* is set appropriately). 293 * 294 * BPF_PROG_ATTACH 295 * Description 296 * Attach an eBPF program to a *target_fd* at the specified 297 * *attach_type* hook. 298 * 299 * The *attach_type* specifies the eBPF attachment point to 300 * attach the program to, and must be one of *bpf_attach_type* 301 * (see below). 302 * 303 * The *attach_bpf_fd* must be a valid file descriptor for a 304 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 305 * or sock_ops type corresponding to the specified *attach_type*. 306 * 307 * The *target_fd* must be a valid file descriptor for a kernel 308 * object which depends on the attach type of *attach_bpf_fd*: 309 * 310 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 311 * **BPF_PROG_TYPE_CGROUP_SKB**, 312 * **BPF_PROG_TYPE_CGROUP_SOCK**, 313 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 314 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 315 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 316 * **BPF_PROG_TYPE_SOCK_OPS** 317 * 318 * Control Group v2 hierarchy with the eBPF controller 319 * enabled. Requires the kernel to be compiled with 320 * **CONFIG_CGROUP_BPF**. 321 * 322 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 323 * 324 * Network namespace (eg /proc/self/ns/net). 325 * 326 * **BPF_PROG_TYPE_LIRC_MODE2** 327 * 328 * LIRC device path (eg /dev/lircN). Requires the kernel 329 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 330 * 331 * **BPF_PROG_TYPE_SK_SKB**, 332 * **BPF_PROG_TYPE_SK_MSG** 333 * 334 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 335 * 336 * Return 337 * Returns zero on success. On error, -1 is returned and *errno* 338 * is set appropriately. 339 * 340 * BPF_PROG_DETACH 341 * Description 342 * Detach the eBPF program associated with the *target_fd* at the 343 * hook specified by *attach_type*. The program must have been 344 * previously attached using **BPF_PROG_ATTACH**. 345 * 346 * Return 347 * Returns zero on success. On error, -1 is returned and *errno* 348 * is set appropriately. 349 * 350 * BPF_PROG_TEST_RUN 351 * Description 352 * Run the eBPF program associated with the *prog_fd* a *repeat* 353 * number of times against a provided program context *ctx_in* and 354 * data *data_in*, and return the modified program context 355 * *ctx_out*, *data_out* (for example, packet data), result of the 356 * execution *retval*, and *duration* of the test run. 357 * 358 * The sizes of the buffers provided as input and output 359 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 360 * be provided in the corresponding variables *ctx_size_in*, 361 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 362 * of these parameters are not provided (ie set to NULL), the 363 * corresponding size field must be zero. 364 * 365 * Some program types have particular requirements: 366 * 367 * **BPF_PROG_TYPE_SK_LOOKUP** 368 * *data_in* and *data_out* must be NULL. 369 * 370 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 371 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 372 * 373 * *ctx_out*, *data_in* and *data_out* must be NULL. 374 * *repeat* must be zero. 375 * 376 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 377 * 378 * Return 379 * Returns zero on success. On error, -1 is returned and *errno* 380 * is set appropriately. 381 * 382 * **ENOSPC** 383 * Either *data_size_out* or *ctx_size_out* is too small. 384 * **ENOTSUPP** 385 * This command is not supported by the program type of 386 * the program referred to by *prog_fd*. 387 * 388 * BPF_PROG_GET_NEXT_ID 389 * Description 390 * Fetch the next eBPF program currently loaded into the kernel. 391 * 392 * Looks for the eBPF program with an id greater than *start_id* 393 * and updates *next_id* on success. If no other eBPF programs 394 * remain with ids higher than *start_id*, returns -1 and sets 395 * *errno* to **ENOENT**. 396 * 397 * Return 398 * Returns zero on success. On error, or when no id remains, -1 399 * is returned and *errno* is set appropriately. 400 * 401 * BPF_MAP_GET_NEXT_ID 402 * Description 403 * Fetch the next eBPF map currently loaded into the kernel. 404 * 405 * Looks for the eBPF map with an id greater than *start_id* 406 * and updates *next_id* on success. If no other eBPF maps 407 * remain with ids higher than *start_id*, returns -1 and sets 408 * *errno* to **ENOENT**. 409 * 410 * Return 411 * Returns zero on success. On error, or when no id remains, -1 412 * is returned and *errno* is set appropriately. 413 * 414 * BPF_PROG_GET_FD_BY_ID 415 * Description 416 * Open a file descriptor for the eBPF program corresponding to 417 * *prog_id*. 418 * 419 * Return 420 * A new file descriptor (a nonnegative integer), or -1 if an 421 * error occurred (in which case, *errno* is set appropriately). 422 * 423 * BPF_MAP_GET_FD_BY_ID 424 * Description 425 * Open a file descriptor for the eBPF map corresponding to 426 * *map_id*. 427 * 428 * Return 429 * A new file descriptor (a nonnegative integer), or -1 if an 430 * error occurred (in which case, *errno* is set appropriately). 431 * 432 * BPF_OBJ_GET_INFO_BY_FD 433 * Description 434 * Obtain information about the eBPF object corresponding to 435 * *bpf_fd*. 436 * 437 * Populates up to *info_len* bytes of *info*, which will be in 438 * one of the following formats depending on the eBPF object type 439 * of *bpf_fd*: 440 * 441 * * **struct bpf_prog_info** 442 * * **struct bpf_map_info** 443 * * **struct bpf_btf_info** 444 * * **struct bpf_link_info** 445 * 446 * Return 447 * Returns zero on success. On error, -1 is returned and *errno* 448 * is set appropriately. 449 * 450 * BPF_PROG_QUERY 451 * Description 452 * Obtain information about eBPF programs associated with the 453 * specified *attach_type* hook. 454 * 455 * The *target_fd* must be a valid file descriptor for a kernel 456 * object which depends on the attach type of *attach_bpf_fd*: 457 * 458 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 459 * **BPF_PROG_TYPE_CGROUP_SKB**, 460 * **BPF_PROG_TYPE_CGROUP_SOCK**, 461 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 462 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 463 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 464 * **BPF_PROG_TYPE_SOCK_OPS** 465 * 466 * Control Group v2 hierarchy with the eBPF controller 467 * enabled. Requires the kernel to be compiled with 468 * **CONFIG_CGROUP_BPF**. 469 * 470 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 471 * 472 * Network namespace (eg /proc/self/ns/net). 473 * 474 * **BPF_PROG_TYPE_LIRC_MODE2** 475 * 476 * LIRC device path (eg /dev/lircN). Requires the kernel 477 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 478 * 479 * **BPF_PROG_QUERY** always fetches the number of programs 480 * attached and the *attach_flags* which were used to attach those 481 * programs. Additionally, if *prog_ids* is nonzero and the number 482 * of attached programs is less than *prog_cnt*, populates 483 * *prog_ids* with the eBPF program ids of the programs attached 484 * at *target_fd*. 485 * 486 * The following flags may alter the result: 487 * 488 * **BPF_F_QUERY_EFFECTIVE** 489 * Only return information regarding programs which are 490 * currently effective at the specified *target_fd*. 491 * 492 * Return 493 * Returns zero on success. On error, -1 is returned and *errno* 494 * is set appropriately. 495 * 496 * BPF_RAW_TRACEPOINT_OPEN 497 * Description 498 * Attach an eBPF program to a tracepoint *name* to access kernel 499 * internal arguments of the tracepoint in their raw form. 500 * 501 * The *prog_fd* must be a valid file descriptor associated with 502 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 503 * 504 * No ABI guarantees are made about the content of tracepoint 505 * arguments exposed to the corresponding eBPF program. 506 * 507 * Applying **close**\ (2) to the file descriptor returned by 508 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 509 * 510 * Return 511 * A new file descriptor (a nonnegative integer), or -1 if an 512 * error occurred (in which case, *errno* is set appropriately). 513 * 514 * BPF_BTF_LOAD 515 * Description 516 * Verify and load BPF Type Format (BTF) metadata into the kernel, 517 * returning a new file descriptor associated with the metadata. 518 * BTF is described in more detail at 519 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 520 * 521 * The *btf* parameter must point to valid memory providing 522 * *btf_size* bytes of BTF binary metadata. 523 * 524 * The returned file descriptor can be passed to other **bpf**\ () 525 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 526 * associate the BTF with those objects. 527 * 528 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 529 * parameters to specify a *btf_log_buf*, *btf_log_size* and 530 * *btf_log_level* which allow the kernel to return freeform log 531 * output regarding the BTF verification process. 532 * 533 * Return 534 * A new file descriptor (a nonnegative integer), or -1 if an 535 * error occurred (in which case, *errno* is set appropriately). 536 * 537 * BPF_BTF_GET_FD_BY_ID 538 * Description 539 * Open a file descriptor for the BPF Type Format (BTF) 540 * corresponding to *btf_id*. 541 * 542 * Return 543 * A new file descriptor (a nonnegative integer), or -1 if an 544 * error occurred (in which case, *errno* is set appropriately). 545 * 546 * BPF_TASK_FD_QUERY 547 * Description 548 * Obtain information about eBPF programs associated with the 549 * target process identified by *pid* and *fd*. 550 * 551 * If the *pid* and *fd* are associated with a tracepoint, kprobe 552 * or uprobe perf event, then the *prog_id* and *fd_type* will 553 * be populated with the eBPF program id and file descriptor type 554 * of type **bpf_task_fd_type**. If associated with a kprobe or 555 * uprobe, the *probe_offset* and *probe_addr* will also be 556 * populated. Optionally, if *buf* is provided, then up to 557 * *buf_len* bytes of *buf* will be populated with the name of 558 * the tracepoint, kprobe or uprobe. 559 * 560 * The resulting *prog_id* may be introspected in deeper detail 561 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 562 * 563 * Return 564 * Returns zero on success. On error, -1 is returned and *errno* 565 * is set appropriately. 566 * 567 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 568 * Description 569 * Look up an element with the given *key* in the map referred to 570 * by the file descriptor *fd*, and if found, delete the element. 571 * 572 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 573 * types, the *flags* argument needs to be set to 0, but for other 574 * map types, it may be specified as: 575 * 576 * **BPF_F_LOCK** 577 * Look up and delete the value of a spin-locked map 578 * without returning the lock. This must be specified if 579 * the elements contain a spinlock. 580 * 581 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 582 * implement this command as a "pop" operation, deleting the top 583 * element rather than one corresponding to *key*. 584 * The *key* and *key_len* parameters should be zeroed when 585 * issuing this operation for these map types. 586 * 587 * This command is only valid for the following map types: 588 * * **BPF_MAP_TYPE_QUEUE** 589 * * **BPF_MAP_TYPE_STACK** 590 * * **BPF_MAP_TYPE_HASH** 591 * * **BPF_MAP_TYPE_PERCPU_HASH** 592 * * **BPF_MAP_TYPE_LRU_HASH** 593 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 594 * 595 * Return 596 * Returns zero on success. On error, -1 is returned and *errno* 597 * is set appropriately. 598 * 599 * BPF_MAP_FREEZE 600 * Description 601 * Freeze the permissions of the specified map. 602 * 603 * Write permissions may be frozen by passing zero *flags*. 604 * Upon success, no future syscall invocations may alter the 605 * map state of *map_fd*. Write operations from eBPF programs 606 * are still possible for a frozen map. 607 * 608 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 609 * 610 * Return 611 * Returns zero on success. On error, -1 is returned and *errno* 612 * is set appropriately. 613 * 614 * BPF_BTF_GET_NEXT_ID 615 * Description 616 * Fetch the next BPF Type Format (BTF) object currently loaded 617 * into the kernel. 618 * 619 * Looks for the BTF object with an id greater than *start_id* 620 * and updates *next_id* on success. If no other BTF objects 621 * remain with ids higher than *start_id*, returns -1 and sets 622 * *errno* to **ENOENT**. 623 * 624 * Return 625 * Returns zero on success. On error, or when no id remains, -1 626 * is returned and *errno* is set appropriately. 627 * 628 * BPF_MAP_LOOKUP_BATCH 629 * Description 630 * Iterate and fetch multiple elements in a map. 631 * 632 * Two opaque values are used to manage batch operations, 633 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 634 * to NULL to begin the batched operation. After each subsequent 635 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 636 * *out_batch* as the *in_batch* for the next operation to 637 * continue iteration from the current point. Both *in_batch* and 638 * *out_batch* must point to memory large enough to hold a key, 639 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH, 640 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters 641 * must be at least 4 bytes wide regardless of key size. 642 * 643 * The *keys* and *values* are output parameters which must point 644 * to memory large enough to hold *count* items based on the key 645 * and value size of the map *map_fd*. The *keys* buffer must be 646 * of *key_size* * *count*. The *values* buffer must be of 647 * *value_size* * *count*. 648 * 649 * The *elem_flags* argument may be specified as one of the 650 * following: 651 * 652 * **BPF_F_LOCK** 653 * Look up the value of a spin-locked map without 654 * returning the lock. This must be specified if the 655 * elements contain a spinlock. 656 * 657 * On success, *count* elements from the map are copied into the 658 * user buffer, with the keys copied into *keys* and the values 659 * copied into the corresponding indices in *values*. 660 * 661 * If an error is returned and *errno* is not **EFAULT**, *count* 662 * is set to the number of successfully processed elements. 663 * 664 * Return 665 * Returns zero on success. On error, -1 is returned and *errno* 666 * is set appropriately. 667 * 668 * May set *errno* to **ENOSPC** to indicate that *keys* or 669 * *values* is too small to dump an entire bucket during 670 * iteration of a hash-based map type. 671 * 672 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 673 * Description 674 * Iterate and delete all elements in a map. 675 * 676 * This operation has the same behavior as 677 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 678 * 679 * * Every element that is successfully returned is also deleted 680 * from the map. This is at least *count* elements. Note that 681 * *count* is both an input and an output parameter. 682 * * Upon returning with *errno* set to **EFAULT**, up to 683 * *count* elements may be deleted without returning the keys 684 * and values of the deleted elements. 685 * 686 * Return 687 * Returns zero on success. On error, -1 is returned and *errno* 688 * is set appropriately. 689 * 690 * BPF_MAP_UPDATE_BATCH 691 * Description 692 * Update multiple elements in a map by *key*. 693 * 694 * The *keys* and *values* are input parameters which must point 695 * to memory large enough to hold *count* items based on the key 696 * and value size of the map *map_fd*. The *keys* buffer must be 697 * of *key_size* * *count*. The *values* buffer must be of 698 * *value_size* * *count*. 699 * 700 * Each element specified in *keys* is sequentially updated to the 701 * value in the corresponding index in *values*. The *in_batch* 702 * and *out_batch* parameters are ignored and should be zeroed. 703 * 704 * The *elem_flags* argument should be specified as one of the 705 * following: 706 * 707 * **BPF_ANY** 708 * Create new elements or update a existing elements. 709 * **BPF_NOEXIST** 710 * Create new elements only if they do not exist. 711 * **BPF_EXIST** 712 * Update existing elements. 713 * **BPF_F_LOCK** 714 * Update spin_lock-ed map elements. This must be 715 * specified if the map value contains a spinlock. 716 * 717 * On success, *count* elements from the map are updated. 718 * 719 * If an error is returned and *errno* is not **EFAULT**, *count* 720 * is set to the number of successfully processed elements. 721 * 722 * Return 723 * Returns zero on success. On error, -1 is returned and *errno* 724 * is set appropriately. 725 * 726 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 727 * **E2BIG**. **E2BIG** indicates that the number of elements in 728 * the map reached the *max_entries* limit specified at map 729 * creation time. 730 * 731 * May set *errno* to one of the following error codes under 732 * specific circumstances: 733 * 734 * **EEXIST** 735 * If *flags* specifies **BPF_NOEXIST** and the element 736 * with *key* already exists in the map. 737 * **ENOENT** 738 * If *flags* specifies **BPF_EXIST** and the element with 739 * *key* does not exist in the map. 740 * 741 * BPF_MAP_DELETE_BATCH 742 * Description 743 * Delete multiple elements in a map by *key*. 744 * 745 * The *keys* parameter is an input parameter which must point 746 * to memory large enough to hold *count* items based on the key 747 * size of the map *map_fd*, that is, *key_size* * *count*. 748 * 749 * Each element specified in *keys* is sequentially deleted. The 750 * *in_batch*, *out_batch*, and *values* parameters are ignored 751 * and should be zeroed. 752 * 753 * The *elem_flags* argument may be specified as one of the 754 * following: 755 * 756 * **BPF_F_LOCK** 757 * Look up the value of a spin-locked map without 758 * returning the lock. This must be specified if the 759 * elements contain a spinlock. 760 * 761 * On success, *count* elements from the map are updated. 762 * 763 * If an error is returned and *errno* is not **EFAULT**, *count* 764 * is set to the number of successfully processed elements. If 765 * *errno* is **EFAULT**, up to *count* elements may be been 766 * deleted. 767 * 768 * Return 769 * Returns zero on success. On error, -1 is returned and *errno* 770 * is set appropriately. 771 * 772 * BPF_LINK_CREATE 773 * Description 774 * Attach an eBPF program to a *target_fd* at the specified 775 * *attach_type* hook and return a file descriptor handle for 776 * managing the link. 777 * 778 * Return 779 * A new file descriptor (a nonnegative integer), or -1 if an 780 * error occurred (in which case, *errno* is set appropriately). 781 * 782 * BPF_LINK_UPDATE 783 * Description 784 * Update the eBPF program in the specified *link_fd* to 785 * *new_prog_fd*. 786 * 787 * Return 788 * Returns zero on success. On error, -1 is returned and *errno* 789 * is set appropriately. 790 * 791 * BPF_LINK_GET_FD_BY_ID 792 * Description 793 * Open a file descriptor for the eBPF Link corresponding to 794 * *link_id*. 795 * 796 * Return 797 * A new file descriptor (a nonnegative integer), or -1 if an 798 * error occurred (in which case, *errno* is set appropriately). 799 * 800 * BPF_LINK_GET_NEXT_ID 801 * Description 802 * Fetch the next eBPF link currently loaded into the kernel. 803 * 804 * Looks for the eBPF link with an id greater than *start_id* 805 * and updates *next_id* on success. If no other eBPF links 806 * remain with ids higher than *start_id*, returns -1 and sets 807 * *errno* to **ENOENT**. 808 * 809 * Return 810 * Returns zero on success. On error, or when no id remains, -1 811 * is returned and *errno* is set appropriately. 812 * 813 * BPF_ENABLE_STATS 814 * Description 815 * Enable eBPF runtime statistics gathering. 816 * 817 * Runtime statistics gathering for the eBPF runtime is disabled 818 * by default to minimize the corresponding performance overhead. 819 * This command enables statistics globally. 820 * 821 * Multiple programs may independently enable statistics. 822 * After gathering the desired statistics, eBPF runtime statistics 823 * may be disabled again by calling **close**\ (2) for the file 824 * descriptor returned by this function. Statistics will only be 825 * disabled system-wide when all outstanding file descriptors 826 * returned by prior calls for this subcommand are closed. 827 * 828 * Return 829 * A new file descriptor (a nonnegative integer), or -1 if an 830 * error occurred (in which case, *errno* is set appropriately). 831 * 832 * BPF_ITER_CREATE 833 * Description 834 * Create an iterator on top of the specified *link_fd* (as 835 * previously created using **BPF_LINK_CREATE**) and return a 836 * file descriptor that can be used to trigger the iteration. 837 * 838 * If the resulting file descriptor is pinned to the filesystem 839 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 840 * for that path will trigger the iterator to read kernel state 841 * using the eBPF program attached to *link_fd*. 842 * 843 * Return 844 * A new file descriptor (a nonnegative integer), or -1 if an 845 * error occurred (in which case, *errno* is set appropriately). 846 * 847 * BPF_LINK_DETACH 848 * Description 849 * Forcefully detach the specified *link_fd* from its 850 * corresponding attachment point. 851 * 852 * Return 853 * Returns zero on success. On error, -1 is returned and *errno* 854 * is set appropriately. 855 * 856 * BPF_PROG_BIND_MAP 857 * Description 858 * Bind a map to the lifetime of an eBPF program. 859 * 860 * The map identified by *map_fd* is bound to the program 861 * identified by *prog_fd* and only released when *prog_fd* is 862 * released. This may be used in cases where metadata should be 863 * associated with a program which otherwise does not contain any 864 * references to the map (for example, embedded in the eBPF 865 * program instructions). 866 * 867 * Return 868 * Returns zero on success. On error, -1 is returned and *errno* 869 * is set appropriately. 870 * 871 * BPF_TOKEN_CREATE 872 * Description 873 * Create BPF token with embedded information about what 874 * BPF-related functionality it allows: 875 * - a set of allowed bpf() syscall commands; 876 * - a set of allowed BPF map types to be created with 877 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 878 * - a set of allowed BPF program types and BPF program attach 879 * types to be loaded with BPF_PROG_LOAD command, if 880 * BPF_PROG_LOAD itself is allowed. 881 * 882 * BPF token is created (derived) from an instance of BPF FS, 883 * assuming it has necessary delegation mount options specified. 884 * This BPF token can be passed as an extra parameter to various 885 * bpf() syscall commands to grant BPF subsystem functionality to 886 * unprivileged processes. 887 * 888 * When created, BPF token is "associated" with the owning 889 * user namespace of BPF FS instance (super block) that it was 890 * derived from, and subsequent BPF operations performed with 891 * BPF token would be performing capabilities checks (i.e., 892 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 893 * that user namespace. Without BPF token, such capabilities 894 * have to be granted in init user namespace, making bpf() 895 * syscall incompatible with user namespace, for the most part. 896 * 897 * Return 898 * A new file descriptor (a nonnegative integer), or -1 if an 899 * error occurred (in which case, *errno* is set appropriately). 900 * 901 * NOTES 902 * eBPF objects (maps and programs) can be shared between processes. 903 * 904 * * After **fork**\ (2), the child inherits file descriptors 905 * referring to the same eBPF objects. 906 * * File descriptors referring to eBPF objects can be transferred over 907 * **unix**\ (7) domain sockets. 908 * * File descriptors referring to eBPF objects can be duplicated in the 909 * usual way, using **dup**\ (2) and similar calls. 910 * * File descriptors referring to eBPF objects can be pinned to the 911 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 912 * 913 * An eBPF object is deallocated only after all file descriptors referring 914 * to the object have been closed and no references remain pinned to the 915 * filesystem or attached (for example, bound to a program or device). 916 */ 917 enum bpf_cmd { 918 BPF_MAP_CREATE, 919 BPF_MAP_LOOKUP_ELEM, 920 BPF_MAP_UPDATE_ELEM, 921 BPF_MAP_DELETE_ELEM, 922 BPF_MAP_GET_NEXT_KEY, 923 BPF_PROG_LOAD, 924 BPF_OBJ_PIN, 925 BPF_OBJ_GET, 926 BPF_PROG_ATTACH, 927 BPF_PROG_DETACH, 928 BPF_PROG_TEST_RUN, 929 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 930 BPF_PROG_GET_NEXT_ID, 931 BPF_MAP_GET_NEXT_ID, 932 BPF_PROG_GET_FD_BY_ID, 933 BPF_MAP_GET_FD_BY_ID, 934 BPF_OBJ_GET_INFO_BY_FD, 935 BPF_PROG_QUERY, 936 BPF_RAW_TRACEPOINT_OPEN, 937 BPF_BTF_LOAD, 938 BPF_BTF_GET_FD_BY_ID, 939 BPF_TASK_FD_QUERY, 940 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 941 BPF_MAP_FREEZE, 942 BPF_BTF_GET_NEXT_ID, 943 BPF_MAP_LOOKUP_BATCH, 944 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 945 BPF_MAP_UPDATE_BATCH, 946 BPF_MAP_DELETE_BATCH, 947 BPF_LINK_CREATE, 948 BPF_LINK_UPDATE, 949 BPF_LINK_GET_FD_BY_ID, 950 BPF_LINK_GET_NEXT_ID, 951 BPF_ENABLE_STATS, 952 BPF_ITER_CREATE, 953 BPF_LINK_DETACH, 954 BPF_PROG_BIND_MAP, 955 BPF_TOKEN_CREATE, 956 __MAX_BPF_CMD, 957 }; 958 959 enum bpf_map_type { 960 BPF_MAP_TYPE_UNSPEC, 961 BPF_MAP_TYPE_HASH, 962 BPF_MAP_TYPE_ARRAY, 963 BPF_MAP_TYPE_PROG_ARRAY, 964 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 965 BPF_MAP_TYPE_PERCPU_HASH, 966 BPF_MAP_TYPE_PERCPU_ARRAY, 967 BPF_MAP_TYPE_STACK_TRACE, 968 BPF_MAP_TYPE_CGROUP_ARRAY, 969 BPF_MAP_TYPE_LRU_HASH, 970 BPF_MAP_TYPE_LRU_PERCPU_HASH, 971 BPF_MAP_TYPE_LPM_TRIE, 972 BPF_MAP_TYPE_ARRAY_OF_MAPS, 973 BPF_MAP_TYPE_HASH_OF_MAPS, 974 BPF_MAP_TYPE_DEVMAP, 975 BPF_MAP_TYPE_SOCKMAP, 976 BPF_MAP_TYPE_CPUMAP, 977 BPF_MAP_TYPE_XSKMAP, 978 BPF_MAP_TYPE_SOCKHASH, 979 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 980 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 981 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 982 * both cgroup-attached and other progs and supports all functionality 983 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 984 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 985 */ 986 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 987 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 988 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 989 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 990 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 991 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 992 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 993 * deprecated. 994 */ 995 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 996 BPF_MAP_TYPE_QUEUE, 997 BPF_MAP_TYPE_STACK, 998 BPF_MAP_TYPE_SK_STORAGE, 999 BPF_MAP_TYPE_DEVMAP_HASH, 1000 BPF_MAP_TYPE_STRUCT_OPS, 1001 BPF_MAP_TYPE_RINGBUF, 1002 BPF_MAP_TYPE_INODE_STORAGE, 1003 BPF_MAP_TYPE_TASK_STORAGE, 1004 BPF_MAP_TYPE_BLOOM_FILTER, 1005 BPF_MAP_TYPE_USER_RINGBUF, 1006 BPF_MAP_TYPE_CGRP_STORAGE, 1007 __MAX_BPF_MAP_TYPE 1008 }; 1009 1010 /* Note that tracing related programs such as 1011 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1012 * are not subject to a stable API since kernel internal data 1013 * structures can change from release to release and may 1014 * therefore break existing tracing BPF programs. Tracing BPF 1015 * programs correspond to /a/ specific kernel which is to be 1016 * analyzed, and not /a/ specific kernel /and/ all future ones. 1017 */ 1018 enum bpf_prog_type { 1019 BPF_PROG_TYPE_UNSPEC, 1020 BPF_PROG_TYPE_SOCKET_FILTER, 1021 BPF_PROG_TYPE_KPROBE, 1022 BPF_PROG_TYPE_SCHED_CLS, 1023 BPF_PROG_TYPE_SCHED_ACT, 1024 BPF_PROG_TYPE_TRACEPOINT, 1025 BPF_PROG_TYPE_XDP, 1026 BPF_PROG_TYPE_PERF_EVENT, 1027 BPF_PROG_TYPE_CGROUP_SKB, 1028 BPF_PROG_TYPE_CGROUP_SOCK, 1029 BPF_PROG_TYPE_LWT_IN, 1030 BPF_PROG_TYPE_LWT_OUT, 1031 BPF_PROG_TYPE_LWT_XMIT, 1032 BPF_PROG_TYPE_SOCK_OPS, 1033 BPF_PROG_TYPE_SK_SKB, 1034 BPF_PROG_TYPE_CGROUP_DEVICE, 1035 BPF_PROG_TYPE_SK_MSG, 1036 BPF_PROG_TYPE_RAW_TRACEPOINT, 1037 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1038 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1039 BPF_PROG_TYPE_LIRC_MODE2, 1040 BPF_PROG_TYPE_SK_REUSEPORT, 1041 BPF_PROG_TYPE_FLOW_DISSECTOR, 1042 BPF_PROG_TYPE_CGROUP_SYSCTL, 1043 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1044 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1045 BPF_PROG_TYPE_TRACING, 1046 BPF_PROG_TYPE_STRUCT_OPS, 1047 BPF_PROG_TYPE_EXT, 1048 BPF_PROG_TYPE_LSM, 1049 BPF_PROG_TYPE_SK_LOOKUP, 1050 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1051 BPF_PROG_TYPE_NETFILTER, 1052 __MAX_BPF_PROG_TYPE 1053 }; 1054 1055 enum bpf_attach_type { 1056 BPF_CGROUP_INET_INGRESS, 1057 BPF_CGROUP_INET_EGRESS, 1058 BPF_CGROUP_INET_SOCK_CREATE, 1059 BPF_CGROUP_SOCK_OPS, 1060 BPF_SK_SKB_STREAM_PARSER, 1061 BPF_SK_SKB_STREAM_VERDICT, 1062 BPF_CGROUP_DEVICE, 1063 BPF_SK_MSG_VERDICT, 1064 BPF_CGROUP_INET4_BIND, 1065 BPF_CGROUP_INET6_BIND, 1066 BPF_CGROUP_INET4_CONNECT, 1067 BPF_CGROUP_INET6_CONNECT, 1068 BPF_CGROUP_INET4_POST_BIND, 1069 BPF_CGROUP_INET6_POST_BIND, 1070 BPF_CGROUP_UDP4_SENDMSG, 1071 BPF_CGROUP_UDP6_SENDMSG, 1072 BPF_LIRC_MODE2, 1073 BPF_FLOW_DISSECTOR, 1074 BPF_CGROUP_SYSCTL, 1075 BPF_CGROUP_UDP4_RECVMSG, 1076 BPF_CGROUP_UDP6_RECVMSG, 1077 BPF_CGROUP_GETSOCKOPT, 1078 BPF_CGROUP_SETSOCKOPT, 1079 BPF_TRACE_RAW_TP, 1080 BPF_TRACE_FENTRY, 1081 BPF_TRACE_FEXIT, 1082 BPF_MODIFY_RETURN, 1083 BPF_LSM_MAC, 1084 BPF_TRACE_ITER, 1085 BPF_CGROUP_INET4_GETPEERNAME, 1086 BPF_CGROUP_INET6_GETPEERNAME, 1087 BPF_CGROUP_INET4_GETSOCKNAME, 1088 BPF_CGROUP_INET6_GETSOCKNAME, 1089 BPF_XDP_DEVMAP, 1090 BPF_CGROUP_INET_SOCK_RELEASE, 1091 BPF_XDP_CPUMAP, 1092 BPF_SK_LOOKUP, 1093 BPF_XDP, 1094 BPF_SK_SKB_VERDICT, 1095 BPF_SK_REUSEPORT_SELECT, 1096 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1097 BPF_PERF_EVENT, 1098 BPF_TRACE_KPROBE_MULTI, 1099 BPF_LSM_CGROUP, 1100 BPF_STRUCT_OPS, 1101 BPF_NETFILTER, 1102 BPF_TCX_INGRESS, 1103 BPF_TCX_EGRESS, 1104 BPF_TRACE_UPROBE_MULTI, 1105 BPF_CGROUP_UNIX_CONNECT, 1106 BPF_CGROUP_UNIX_SENDMSG, 1107 BPF_CGROUP_UNIX_RECVMSG, 1108 BPF_CGROUP_UNIX_GETPEERNAME, 1109 BPF_CGROUP_UNIX_GETSOCKNAME, 1110 BPF_NETKIT_PRIMARY, 1111 BPF_NETKIT_PEER, 1112 __MAX_BPF_ATTACH_TYPE 1113 }; 1114 1115 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1116 1117 enum bpf_link_type { 1118 BPF_LINK_TYPE_UNSPEC = 0, 1119 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1120 BPF_LINK_TYPE_TRACING = 2, 1121 BPF_LINK_TYPE_CGROUP = 3, 1122 BPF_LINK_TYPE_ITER = 4, 1123 BPF_LINK_TYPE_NETNS = 5, 1124 BPF_LINK_TYPE_XDP = 6, 1125 BPF_LINK_TYPE_PERF_EVENT = 7, 1126 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1127 BPF_LINK_TYPE_STRUCT_OPS = 9, 1128 BPF_LINK_TYPE_NETFILTER = 10, 1129 BPF_LINK_TYPE_TCX = 11, 1130 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1131 BPF_LINK_TYPE_NETKIT = 13, 1132 __MAX_BPF_LINK_TYPE, 1133 }; 1134 1135 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1136 1137 enum bpf_perf_event_type { 1138 BPF_PERF_EVENT_UNSPEC = 0, 1139 BPF_PERF_EVENT_UPROBE = 1, 1140 BPF_PERF_EVENT_URETPROBE = 2, 1141 BPF_PERF_EVENT_KPROBE = 3, 1142 BPF_PERF_EVENT_KRETPROBE = 4, 1143 BPF_PERF_EVENT_TRACEPOINT = 5, 1144 BPF_PERF_EVENT_EVENT = 6, 1145 }; 1146 1147 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1148 * 1149 * NONE(default): No further bpf programs allowed in the subtree. 1150 * 1151 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1152 * the program in this cgroup yields to sub-cgroup program. 1153 * 1154 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1155 * that cgroup program gets run in addition to the program in this cgroup. 1156 * 1157 * Only one program is allowed to be attached to a cgroup with 1158 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1159 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1160 * release old program and attach the new one. Attach flags has to match. 1161 * 1162 * Multiple programs are allowed to be attached to a cgroup with 1163 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1164 * (those that were attached first, run first) 1165 * The programs of sub-cgroup are executed first, then programs of 1166 * this cgroup and then programs of parent cgroup. 1167 * When children program makes decision (like picking TCP CA or sock bind) 1168 * parent program has a chance to override it. 1169 * 1170 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1171 * programs for a cgroup. Though it's possible to replace an old program at 1172 * any position by also specifying BPF_F_REPLACE flag and position itself in 1173 * replace_bpf_fd attribute. Old program at this position will be released. 1174 * 1175 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1176 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1177 * Ex1: 1178 * cgrp1 (MULTI progs A, B) -> 1179 * cgrp2 (OVERRIDE prog C) -> 1180 * cgrp3 (MULTI prog D) -> 1181 * cgrp4 (OVERRIDE prog E) -> 1182 * cgrp5 (NONE prog F) 1183 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1184 * if prog F is detached, the execution is E,D,A,B 1185 * if prog F and D are detached, the execution is E,A,B 1186 * if prog F, E and D are detached, the execution is C,A,B 1187 * 1188 * All eligible programs are executed regardless of return code from 1189 * earlier programs. 1190 */ 1191 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1192 #define BPF_F_ALLOW_MULTI (1U << 1) 1193 /* Generic attachment flags. */ 1194 #define BPF_F_REPLACE (1U << 2) 1195 #define BPF_F_BEFORE (1U << 3) 1196 #define BPF_F_AFTER (1U << 4) 1197 #define BPF_F_ID (1U << 5) 1198 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1199 1200 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1201 * verifier will perform strict alignment checking as if the kernel 1202 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1203 * and NET_IP_ALIGN defined to 2. 1204 */ 1205 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1206 1207 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1208 * verifier will allow any alignment whatsoever. On platforms 1209 * with strict alignment requirements for loads ands stores (such 1210 * as sparc and mips) the verifier validates that all loads and 1211 * stores provably follow this requirement. This flag turns that 1212 * checking and enforcement off. 1213 * 1214 * It is mostly used for testing when we want to validate the 1215 * context and memory access aspects of the verifier, but because 1216 * of an unaligned access the alignment check would trigger before 1217 * the one we are interested in. 1218 */ 1219 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1220 1221 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1222 * Verifier does sub-register def/use analysis and identifies instructions whose 1223 * def only matters for low 32-bit, high 32-bit is never referenced later 1224 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1225 * that it is safe to ignore clearing high 32-bit for these instructions. This 1226 * saves some back-ends a lot of code-gen. However such optimization is not 1227 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1228 * hence hasn't used verifier's analysis result. But, we really want to have a 1229 * way to be able to verify the correctness of the described optimization on 1230 * x86_64 on which testsuites are frequently exercised. 1231 * 1232 * So, this flag is introduced. Once it is set, verifier will randomize high 1233 * 32-bit for those instructions who has been identified as safe to ignore them. 1234 * Then, if verifier is not doing correct analysis, such randomization will 1235 * regress tests to expose bugs. 1236 */ 1237 #define BPF_F_TEST_RND_HI32 (1U << 2) 1238 1239 /* The verifier internal test flag. Behavior is undefined */ 1240 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1241 1242 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1243 * restrict map and helper usage for such programs. Sleepable BPF programs can 1244 * only be attached to hooks where kernel execution context allows sleeping. 1245 * Such programs are allowed to use helpers that may sleep like 1246 * bpf_copy_from_user(). 1247 */ 1248 #define BPF_F_SLEEPABLE (1U << 4) 1249 1250 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1251 * fully support xdp frags. 1252 */ 1253 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1254 1255 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1256 * program becomes device-bound but can access XDP metadata. 1257 */ 1258 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1259 1260 /* The verifier internal test flag. Behavior is undefined */ 1261 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1262 1263 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1264 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1265 */ 1266 enum { 1267 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1268 }; 1269 1270 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1271 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1272 */ 1273 enum { 1274 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1275 }; 1276 1277 /* link_create.netfilter.flags used in LINK_CREATE command for 1278 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1279 */ 1280 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1281 1282 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1283 * the following extensions: 1284 * 1285 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1286 * insn[0].imm: map fd or fd_idx 1287 * insn[1].imm: 0 1288 * insn[0].off: 0 1289 * insn[1].off: 0 1290 * ldimm64 rewrite: address of map 1291 * verifier type: CONST_PTR_TO_MAP 1292 */ 1293 #define BPF_PSEUDO_MAP_FD 1 1294 #define BPF_PSEUDO_MAP_IDX 5 1295 1296 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1297 * insn[0].imm: map fd or fd_idx 1298 * insn[1].imm: offset into value 1299 * insn[0].off: 0 1300 * insn[1].off: 0 1301 * ldimm64 rewrite: address of map[0]+offset 1302 * verifier type: PTR_TO_MAP_VALUE 1303 */ 1304 #define BPF_PSEUDO_MAP_VALUE 2 1305 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1306 1307 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1308 * insn[0].imm: kernel btd id of VAR 1309 * insn[1].imm: 0 1310 * insn[0].off: 0 1311 * insn[1].off: 0 1312 * ldimm64 rewrite: address of the kernel variable 1313 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1314 * is struct/union. 1315 */ 1316 #define BPF_PSEUDO_BTF_ID 3 1317 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1318 * insn[0].imm: insn offset to the func 1319 * insn[1].imm: 0 1320 * insn[0].off: 0 1321 * insn[1].off: 0 1322 * ldimm64 rewrite: address of the function 1323 * verifier type: PTR_TO_FUNC. 1324 */ 1325 #define BPF_PSEUDO_FUNC 4 1326 1327 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1328 * offset to another bpf function 1329 */ 1330 #define BPF_PSEUDO_CALL 1 1331 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1332 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1333 */ 1334 #define BPF_PSEUDO_KFUNC_CALL 2 1335 1336 /* flags for BPF_MAP_UPDATE_ELEM command */ 1337 enum { 1338 BPF_ANY = 0, /* create new element or update existing */ 1339 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1340 BPF_EXIST = 2, /* update existing element */ 1341 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1342 }; 1343 1344 /* flags for BPF_MAP_CREATE command */ 1345 enum { 1346 BPF_F_NO_PREALLOC = (1U << 0), 1347 /* Instead of having one common LRU list in the 1348 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1349 * which can scale and perform better. 1350 * Note, the LRU nodes (including free nodes) cannot be moved 1351 * across different LRU lists. 1352 */ 1353 BPF_F_NO_COMMON_LRU = (1U << 1), 1354 /* Specify numa node during map creation */ 1355 BPF_F_NUMA_NODE = (1U << 2), 1356 1357 /* Flags for accessing BPF object from syscall side. */ 1358 BPF_F_RDONLY = (1U << 3), 1359 BPF_F_WRONLY = (1U << 4), 1360 1361 /* Flag for stack_map, store build_id+offset instead of pointer */ 1362 BPF_F_STACK_BUILD_ID = (1U << 5), 1363 1364 /* Zero-initialize hash function seed. This should only be used for testing. */ 1365 BPF_F_ZERO_SEED = (1U << 6), 1366 1367 /* Flags for accessing BPF object from program side. */ 1368 BPF_F_RDONLY_PROG = (1U << 7), 1369 BPF_F_WRONLY_PROG = (1U << 8), 1370 1371 /* Clone map from listener for newly accepted socket */ 1372 BPF_F_CLONE = (1U << 9), 1373 1374 /* Enable memory-mapping BPF map */ 1375 BPF_F_MMAPABLE = (1U << 10), 1376 1377 /* Share perf_event among processes */ 1378 BPF_F_PRESERVE_ELEMS = (1U << 11), 1379 1380 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1381 BPF_F_INNER_MAP = (1U << 12), 1382 1383 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1384 BPF_F_LINK = (1U << 13), 1385 1386 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1387 BPF_F_PATH_FD = (1U << 14), 1388 1389 /* Flag for value_type_btf_obj_fd, the fd is available */ 1390 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1391 1392 /* BPF token FD is passed in a corresponding command's token_fd field */ 1393 BPF_F_TOKEN_FD = (1U << 16), 1394 }; 1395 1396 /* Flags for BPF_PROG_QUERY. */ 1397 1398 /* Query effective (directly attached + inherited from ancestor cgroups) 1399 * programs that will be executed for events within a cgroup. 1400 * attach_flags with this flag are always returned 0. 1401 */ 1402 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1403 1404 /* Flags for BPF_PROG_TEST_RUN */ 1405 1406 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1407 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1408 /* If set, XDP frames will be transmitted after processing */ 1409 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1410 1411 /* type for BPF_ENABLE_STATS */ 1412 enum bpf_stats_type { 1413 /* enabled run_time_ns and run_cnt */ 1414 BPF_STATS_RUN_TIME = 0, 1415 }; 1416 1417 enum bpf_stack_build_id_status { 1418 /* user space need an empty entry to identify end of a trace */ 1419 BPF_STACK_BUILD_ID_EMPTY = 0, 1420 /* with valid build_id and offset */ 1421 BPF_STACK_BUILD_ID_VALID = 1, 1422 /* couldn't get build_id, fallback to ip */ 1423 BPF_STACK_BUILD_ID_IP = 2, 1424 }; 1425 1426 #define BPF_BUILD_ID_SIZE 20 1427 struct bpf_stack_build_id { 1428 __s32 status; 1429 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1430 union { 1431 __u64 offset; 1432 __u64 ip; 1433 }; 1434 }; 1435 1436 #define BPF_OBJ_NAME_LEN 16U 1437 1438 union bpf_attr { 1439 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1440 __u32 map_type; /* one of enum bpf_map_type */ 1441 __u32 key_size; /* size of key in bytes */ 1442 __u32 value_size; /* size of value in bytes */ 1443 __u32 max_entries; /* max number of entries in a map */ 1444 __u32 map_flags; /* BPF_MAP_CREATE related 1445 * flags defined above. 1446 */ 1447 __u32 inner_map_fd; /* fd pointing to the inner map */ 1448 __u32 numa_node; /* numa node (effective only if 1449 * BPF_F_NUMA_NODE is set). 1450 */ 1451 char map_name[BPF_OBJ_NAME_LEN]; 1452 __u32 map_ifindex; /* ifindex of netdev to create on */ 1453 __u32 btf_fd; /* fd pointing to a BTF type data */ 1454 __u32 btf_key_type_id; /* BTF type_id of the key */ 1455 __u32 btf_value_type_id; /* BTF type_id of the value */ 1456 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1457 * struct stored as the 1458 * map value 1459 */ 1460 /* Any per-map-type extra fields 1461 * 1462 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1463 * number of hash functions (if 0, the bloom filter will default 1464 * to using 5 hash functions). 1465 */ 1466 __u64 map_extra; 1467 1468 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1469 * type data for 1470 * btf_vmlinux_value_type_id. 1471 */ 1472 /* BPF token FD to use with BPF_MAP_CREATE operation. 1473 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1474 */ 1475 __s32 map_token_fd; 1476 }; 1477 1478 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1479 __u32 map_fd; 1480 __aligned_u64 key; 1481 union { 1482 __aligned_u64 value; 1483 __aligned_u64 next_key; 1484 }; 1485 __u64 flags; 1486 }; 1487 1488 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1489 __aligned_u64 in_batch; /* start batch, 1490 * NULL to start from beginning 1491 */ 1492 __aligned_u64 out_batch; /* output: next start batch */ 1493 __aligned_u64 keys; 1494 __aligned_u64 values; 1495 __u32 count; /* input/output: 1496 * input: # of key/value 1497 * elements 1498 * output: # of filled elements 1499 */ 1500 __u32 map_fd; 1501 __u64 elem_flags; 1502 __u64 flags; 1503 } batch; 1504 1505 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1506 __u32 prog_type; /* one of enum bpf_prog_type */ 1507 __u32 insn_cnt; 1508 __aligned_u64 insns; 1509 __aligned_u64 license; 1510 __u32 log_level; /* verbosity level of verifier */ 1511 __u32 log_size; /* size of user buffer */ 1512 __aligned_u64 log_buf; /* user supplied buffer */ 1513 __u32 kern_version; /* not used */ 1514 __u32 prog_flags; 1515 char prog_name[BPF_OBJ_NAME_LEN]; 1516 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1517 /* For some prog types expected attach type must be known at 1518 * load time to verify attach type specific parts of prog 1519 * (context accesses, allowed helpers, etc). 1520 */ 1521 __u32 expected_attach_type; 1522 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1523 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1524 __aligned_u64 func_info; /* func info */ 1525 __u32 func_info_cnt; /* number of bpf_func_info records */ 1526 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1527 __aligned_u64 line_info; /* line info */ 1528 __u32 line_info_cnt; /* number of bpf_line_info records */ 1529 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1530 union { 1531 /* valid prog_fd to attach to bpf prog */ 1532 __u32 attach_prog_fd; 1533 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1534 __u32 attach_btf_obj_fd; 1535 }; 1536 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1537 __aligned_u64 fd_array; /* array of FDs */ 1538 __aligned_u64 core_relos; 1539 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1540 /* output: actual total log contents size (including termintaing zero). 1541 * It could be both larger than original log_size (if log was 1542 * truncated), or smaller (if log buffer wasn't filled completely). 1543 */ 1544 __u32 log_true_size; 1545 /* BPF token FD to use with BPF_PROG_LOAD operation. 1546 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1547 */ 1548 __s32 prog_token_fd; 1549 }; 1550 1551 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1552 __aligned_u64 pathname; 1553 __u32 bpf_fd; 1554 __u32 file_flags; 1555 /* Same as dirfd in openat() syscall; see openat(2) 1556 * manpage for details of path FD and pathname semantics; 1557 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1558 * file_flags field, otherwise it should be set to zero; 1559 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1560 */ 1561 __s32 path_fd; 1562 }; 1563 1564 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1565 union { 1566 __u32 target_fd; /* target object to attach to or ... */ 1567 __u32 target_ifindex; /* target ifindex */ 1568 }; 1569 __u32 attach_bpf_fd; 1570 __u32 attach_type; 1571 __u32 attach_flags; 1572 __u32 replace_bpf_fd; 1573 union { 1574 __u32 relative_fd; 1575 __u32 relative_id; 1576 }; 1577 __u64 expected_revision; 1578 }; 1579 1580 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1581 __u32 prog_fd; 1582 __u32 retval; 1583 __u32 data_size_in; /* input: len of data_in */ 1584 __u32 data_size_out; /* input/output: len of data_out 1585 * returns ENOSPC if data_out 1586 * is too small. 1587 */ 1588 __aligned_u64 data_in; 1589 __aligned_u64 data_out; 1590 __u32 repeat; 1591 __u32 duration; 1592 __u32 ctx_size_in; /* input: len of ctx_in */ 1593 __u32 ctx_size_out; /* input/output: len of ctx_out 1594 * returns ENOSPC if ctx_out 1595 * is too small. 1596 */ 1597 __aligned_u64 ctx_in; 1598 __aligned_u64 ctx_out; 1599 __u32 flags; 1600 __u32 cpu; 1601 __u32 batch_size; 1602 } test; 1603 1604 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1605 union { 1606 __u32 start_id; 1607 __u32 prog_id; 1608 __u32 map_id; 1609 __u32 btf_id; 1610 __u32 link_id; 1611 }; 1612 __u32 next_id; 1613 __u32 open_flags; 1614 }; 1615 1616 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1617 __u32 bpf_fd; 1618 __u32 info_len; 1619 __aligned_u64 info; 1620 } info; 1621 1622 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1623 union { 1624 __u32 target_fd; /* target object to query or ... */ 1625 __u32 target_ifindex; /* target ifindex */ 1626 }; 1627 __u32 attach_type; 1628 __u32 query_flags; 1629 __u32 attach_flags; 1630 __aligned_u64 prog_ids; 1631 union { 1632 __u32 prog_cnt; 1633 __u32 count; 1634 }; 1635 __u32 :32; 1636 /* output: per-program attach_flags. 1637 * not allowed to be set during effective query. 1638 */ 1639 __aligned_u64 prog_attach_flags; 1640 __aligned_u64 link_ids; 1641 __aligned_u64 link_attach_flags; 1642 __u64 revision; 1643 } query; 1644 1645 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1646 __u64 name; 1647 __u32 prog_fd; 1648 } raw_tracepoint; 1649 1650 struct { /* anonymous struct for BPF_BTF_LOAD */ 1651 __aligned_u64 btf; 1652 __aligned_u64 btf_log_buf; 1653 __u32 btf_size; 1654 __u32 btf_log_size; 1655 __u32 btf_log_level; 1656 /* output: actual total log contents size (including termintaing zero). 1657 * It could be both larger than original log_size (if log was 1658 * truncated), or smaller (if log buffer wasn't filled completely). 1659 */ 1660 __u32 btf_log_true_size; 1661 __u32 btf_flags; 1662 /* BPF token FD to use with BPF_BTF_LOAD operation. 1663 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1664 */ 1665 __s32 btf_token_fd; 1666 }; 1667 1668 struct { 1669 __u32 pid; /* input: pid */ 1670 __u32 fd; /* input: fd */ 1671 __u32 flags; /* input: flags */ 1672 __u32 buf_len; /* input/output: buf len */ 1673 __aligned_u64 buf; /* input/output: 1674 * tp_name for tracepoint 1675 * symbol for kprobe 1676 * filename for uprobe 1677 */ 1678 __u32 prog_id; /* output: prod_id */ 1679 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1680 __u64 probe_offset; /* output: probe_offset */ 1681 __u64 probe_addr; /* output: probe_addr */ 1682 } task_fd_query; 1683 1684 struct { /* struct used by BPF_LINK_CREATE command */ 1685 union { 1686 __u32 prog_fd; /* eBPF program to attach */ 1687 __u32 map_fd; /* struct_ops to attach */ 1688 }; 1689 union { 1690 __u32 target_fd; /* target object to attach to or ... */ 1691 __u32 target_ifindex; /* target ifindex */ 1692 }; 1693 __u32 attach_type; /* attach type */ 1694 __u32 flags; /* extra flags */ 1695 union { 1696 __u32 target_btf_id; /* btf_id of target to attach to */ 1697 struct { 1698 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1699 __u32 iter_info_len; /* iter_info length */ 1700 }; 1701 struct { 1702 /* black box user-provided value passed through 1703 * to BPF program at the execution time and 1704 * accessible through bpf_get_attach_cookie() BPF helper 1705 */ 1706 __u64 bpf_cookie; 1707 } perf_event; 1708 struct { 1709 __u32 flags; 1710 __u32 cnt; 1711 __aligned_u64 syms; 1712 __aligned_u64 addrs; 1713 __aligned_u64 cookies; 1714 } kprobe_multi; 1715 struct { 1716 /* this is overlaid with the target_btf_id above. */ 1717 __u32 target_btf_id; 1718 /* black box user-provided value passed through 1719 * to BPF program at the execution time and 1720 * accessible through bpf_get_attach_cookie() BPF helper 1721 */ 1722 __u64 cookie; 1723 } tracing; 1724 struct { 1725 __u32 pf; 1726 __u32 hooknum; 1727 __s32 priority; 1728 __u32 flags; 1729 } netfilter; 1730 struct { 1731 union { 1732 __u32 relative_fd; 1733 __u32 relative_id; 1734 }; 1735 __u64 expected_revision; 1736 } tcx; 1737 struct { 1738 __aligned_u64 path; 1739 __aligned_u64 offsets; 1740 __aligned_u64 ref_ctr_offsets; 1741 __aligned_u64 cookies; 1742 __u32 cnt; 1743 __u32 flags; 1744 __u32 pid; 1745 } uprobe_multi; 1746 struct { 1747 union { 1748 __u32 relative_fd; 1749 __u32 relative_id; 1750 }; 1751 __u64 expected_revision; 1752 } netkit; 1753 }; 1754 } link_create; 1755 1756 struct { /* struct used by BPF_LINK_UPDATE command */ 1757 __u32 link_fd; /* link fd */ 1758 union { 1759 /* new program fd to update link with */ 1760 __u32 new_prog_fd; 1761 /* new struct_ops map fd to update link with */ 1762 __u32 new_map_fd; 1763 }; 1764 __u32 flags; /* extra flags */ 1765 union { 1766 /* expected link's program fd; is specified only if 1767 * BPF_F_REPLACE flag is set in flags. 1768 */ 1769 __u32 old_prog_fd; 1770 /* expected link's map fd; is specified only 1771 * if BPF_F_REPLACE flag is set. 1772 */ 1773 __u32 old_map_fd; 1774 }; 1775 } link_update; 1776 1777 struct { 1778 __u32 link_fd; 1779 } link_detach; 1780 1781 struct { /* struct used by BPF_ENABLE_STATS command */ 1782 __u32 type; 1783 } enable_stats; 1784 1785 struct { /* struct used by BPF_ITER_CREATE command */ 1786 __u32 link_fd; 1787 __u32 flags; 1788 } iter_create; 1789 1790 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1791 __u32 prog_fd; 1792 __u32 map_fd; 1793 __u32 flags; /* extra flags */ 1794 } prog_bind_map; 1795 1796 struct { /* struct used by BPF_TOKEN_CREATE command */ 1797 __u32 flags; 1798 __u32 bpffs_fd; 1799 } token_create; 1800 1801 } __attribute__((aligned(8))); 1802 1803 /* The description below is an attempt at providing documentation to eBPF 1804 * developers about the multiple available eBPF helper functions. It can be 1805 * parsed and used to produce a manual page. The workflow is the following, 1806 * and requires the rst2man utility: 1807 * 1808 * $ ./scripts/bpf_doc.py \ 1809 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1810 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1811 * $ man /tmp/bpf-helpers.7 1812 * 1813 * Note that in order to produce this external documentation, some RST 1814 * formatting is used in the descriptions to get "bold" and "italics" in 1815 * manual pages. Also note that the few trailing white spaces are 1816 * intentional, removing them would break paragraphs for rst2man. 1817 * 1818 * Start of BPF helper function descriptions: 1819 * 1820 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1821 * Description 1822 * Perform a lookup in *map* for an entry associated to *key*. 1823 * Return 1824 * Map value associated to *key*, or **NULL** if no entry was 1825 * found. 1826 * 1827 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1828 * Description 1829 * Add or update the value of the entry associated to *key* in 1830 * *map* with *value*. *flags* is one of: 1831 * 1832 * **BPF_NOEXIST** 1833 * The entry for *key* must not exist in the map. 1834 * **BPF_EXIST** 1835 * The entry for *key* must already exist in the map. 1836 * **BPF_ANY** 1837 * No condition on the existence of the entry for *key*. 1838 * 1839 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1840 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1841 * elements always exist), the helper would return an error. 1842 * Return 1843 * 0 on success, or a negative error in case of failure. 1844 * 1845 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1846 * Description 1847 * Delete entry with *key* from *map*. 1848 * Return 1849 * 0 on success, or a negative error in case of failure. 1850 * 1851 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1852 * Description 1853 * For tracing programs, safely attempt to read *size* bytes from 1854 * kernel space address *unsafe_ptr* and store the data in *dst*. 1855 * 1856 * Generally, use **bpf_probe_read_user**\ () or 1857 * **bpf_probe_read_kernel**\ () instead. 1858 * Return 1859 * 0 on success, or a negative error in case of failure. 1860 * 1861 * u64 bpf_ktime_get_ns(void) 1862 * Description 1863 * Return the time elapsed since system boot, in nanoseconds. 1864 * Does not include time the system was suspended. 1865 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1866 * Return 1867 * Current *ktime*. 1868 * 1869 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1870 * Description 1871 * This helper is a "printk()-like" facility for debugging. It 1872 * prints a message defined by format *fmt* (of size *fmt_size*) 1873 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1874 * available. It can take up to three additional **u64** 1875 * arguments (as an eBPF helpers, the total number of arguments is 1876 * limited to five). 1877 * 1878 * Each time the helper is called, it appends a line to the trace. 1879 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1880 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1881 * The format of the trace is customizable, and the exact output 1882 * one will get depends on the options set in 1883 * *\/sys/kernel/tracing/trace_options* (see also the 1884 * *README* file under the same directory). However, it usually 1885 * defaults to something like: 1886 * 1887 * :: 1888 * 1889 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1890 * 1891 * In the above: 1892 * 1893 * * ``telnet`` is the name of the current task. 1894 * * ``470`` is the PID of the current task. 1895 * * ``001`` is the CPU number on which the task is 1896 * running. 1897 * * In ``.N..``, each character refers to a set of 1898 * options (whether irqs are enabled, scheduling 1899 * options, whether hard/softirqs are running, level of 1900 * preempt_disabled respectively). **N** means that 1901 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1902 * are set. 1903 * * ``419421.045894`` is a timestamp. 1904 * * ``0x00000001`` is a fake value used by BPF for the 1905 * instruction pointer register. 1906 * * ``<formatted msg>`` is the message formatted with 1907 * *fmt*. 1908 * 1909 * The conversion specifiers supported by *fmt* are similar, but 1910 * more limited than for printk(). They are **%d**, **%i**, 1911 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1912 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1913 * of field, padding with zeroes, etc.) is available, and the 1914 * helper will return **-EINVAL** (but print nothing) if it 1915 * encounters an unknown specifier. 1916 * 1917 * Also, note that **bpf_trace_printk**\ () is slow, and should 1918 * only be used for debugging purposes. For this reason, a notice 1919 * block (spanning several lines) is printed to kernel logs and 1920 * states that the helper should not be used "for production use" 1921 * the first time this helper is used (or more precisely, when 1922 * **trace_printk**\ () buffers are allocated). For passing values 1923 * to user space, perf events should be preferred. 1924 * Return 1925 * The number of bytes written to the buffer, or a negative error 1926 * in case of failure. 1927 * 1928 * u32 bpf_get_prandom_u32(void) 1929 * Description 1930 * Get a pseudo-random number. 1931 * 1932 * From a security point of view, this helper uses its own 1933 * pseudo-random internal state, and cannot be used to infer the 1934 * seed of other random functions in the kernel. However, it is 1935 * essential to note that the generator used by the helper is not 1936 * cryptographically secure. 1937 * Return 1938 * A random 32-bit unsigned value. 1939 * 1940 * u32 bpf_get_smp_processor_id(void) 1941 * Description 1942 * Get the SMP (symmetric multiprocessing) processor id. Note that 1943 * all programs run with migration disabled, which means that the 1944 * SMP processor id is stable during all the execution of the 1945 * program. 1946 * Return 1947 * The SMP id of the processor running the program. 1948 * 1949 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1950 * Description 1951 * Store *len* bytes from address *from* into the packet 1952 * associated to *skb*, at *offset*. *flags* are a combination of 1953 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1954 * checksum for the packet after storing the bytes) and 1955 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1956 * **->swhash** and *skb*\ **->l4hash** to 0). 1957 * 1958 * A call to this helper is susceptible to change the underlying 1959 * packet buffer. Therefore, at load time, all checks on pointers 1960 * previously done by the verifier are invalidated and must be 1961 * performed again, if the helper is used in combination with 1962 * direct packet access. 1963 * Return 1964 * 0 on success, or a negative error in case of failure. 1965 * 1966 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1967 * Description 1968 * Recompute the layer 3 (e.g. IP) checksum for the packet 1969 * associated to *skb*. Computation is incremental, so the helper 1970 * must know the former value of the header field that was 1971 * modified (*from*), the new value of this field (*to*), and the 1972 * number of bytes (2 or 4) for this field, stored in *size*. 1973 * Alternatively, it is possible to store the difference between 1974 * the previous and the new values of the header field in *to*, by 1975 * setting *from* and *size* to 0. For both methods, *offset* 1976 * indicates the location of the IP checksum within the packet. 1977 * 1978 * This helper works in combination with **bpf_csum_diff**\ (), 1979 * which does not update the checksum in-place, but offers more 1980 * flexibility and can handle sizes larger than 2 or 4 for the 1981 * checksum to update. 1982 * 1983 * A call to this helper is susceptible to change the underlying 1984 * packet buffer. Therefore, at load time, all checks on pointers 1985 * previously done by the verifier are invalidated and must be 1986 * performed again, if the helper is used in combination with 1987 * direct packet access. 1988 * Return 1989 * 0 on success, or a negative error in case of failure. 1990 * 1991 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1992 * Description 1993 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1994 * packet associated to *skb*. Computation is incremental, so the 1995 * helper must know the former value of the header field that was 1996 * modified (*from*), the new value of this field (*to*), and the 1997 * number of bytes (2 or 4) for this field, stored on the lowest 1998 * four bits of *flags*. Alternatively, it is possible to store 1999 * the difference between the previous and the new values of the 2000 * header field in *to*, by setting *from* and the four lowest 2001 * bits of *flags* to 0. For both methods, *offset* indicates the 2002 * location of the IP checksum within the packet. In addition to 2003 * the size of the field, *flags* can be added (bitwise OR) actual 2004 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 2005 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 2006 * for updates resulting in a null checksum the value is set to 2007 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 2008 * the checksum is to be computed against a pseudo-header. 2009 * 2010 * This helper works in combination with **bpf_csum_diff**\ (), 2011 * which does not update the checksum in-place, but offers more 2012 * flexibility and can handle sizes larger than 2 or 4 for the 2013 * checksum to update. 2014 * 2015 * A call to this helper is susceptible to change the underlying 2016 * packet buffer. Therefore, at load time, all checks on pointers 2017 * previously done by the verifier are invalidated and must be 2018 * performed again, if the helper is used in combination with 2019 * direct packet access. 2020 * Return 2021 * 0 on success, or a negative error in case of failure. 2022 * 2023 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2024 * Description 2025 * This special helper is used to trigger a "tail call", or in 2026 * other words, to jump into another eBPF program. The same stack 2027 * frame is used (but values on stack and in registers for the 2028 * caller are not accessible to the callee). This mechanism allows 2029 * for program chaining, either for raising the maximum number of 2030 * available eBPF instructions, or to execute given programs in 2031 * conditional blocks. For security reasons, there is an upper 2032 * limit to the number of successive tail calls that can be 2033 * performed. 2034 * 2035 * Upon call of this helper, the program attempts to jump into a 2036 * program referenced at index *index* in *prog_array_map*, a 2037 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2038 * *ctx*, a pointer to the context. 2039 * 2040 * If the call succeeds, the kernel immediately runs the first 2041 * instruction of the new program. This is not a function call, 2042 * and it never returns to the previous program. If the call 2043 * fails, then the helper has no effect, and the caller continues 2044 * to run its subsequent instructions. A call can fail if the 2045 * destination program for the jump does not exist (i.e. *index* 2046 * is superior to the number of entries in *prog_array_map*), or 2047 * if the maximum number of tail calls has been reached for this 2048 * chain of programs. This limit is defined in the kernel by the 2049 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2050 * which is currently set to 33. 2051 * Return 2052 * 0 on success, or a negative error in case of failure. 2053 * 2054 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2055 * Description 2056 * Clone and redirect the packet associated to *skb* to another 2057 * net device of index *ifindex*. Both ingress and egress 2058 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2059 * value in *flags* is used to make the distinction (ingress path 2060 * is selected if the flag is present, egress path otherwise). 2061 * This is the only flag supported for now. 2062 * 2063 * In comparison with **bpf_redirect**\ () helper, 2064 * **bpf_clone_redirect**\ () has the associated cost of 2065 * duplicating the packet buffer, but this can be executed out of 2066 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2067 * efficient, but it is handled through an action code where the 2068 * redirection happens only after the eBPF program has returned. 2069 * 2070 * A call to this helper is susceptible to change the underlying 2071 * packet buffer. Therefore, at load time, all checks on pointers 2072 * previously done by the verifier are invalidated and must be 2073 * performed again, if the helper is used in combination with 2074 * direct packet access. 2075 * Return 2076 * 0 on success, or a negative error in case of failure. Positive 2077 * error indicates a potential drop or congestion in the target 2078 * device. The particular positive error codes are not defined. 2079 * 2080 * u64 bpf_get_current_pid_tgid(void) 2081 * Description 2082 * Get the current pid and tgid. 2083 * Return 2084 * A 64-bit integer containing the current tgid and pid, and 2085 * created as such: 2086 * *current_task*\ **->tgid << 32 \|** 2087 * *current_task*\ **->pid**. 2088 * 2089 * u64 bpf_get_current_uid_gid(void) 2090 * Description 2091 * Get the current uid and gid. 2092 * Return 2093 * A 64-bit integer containing the current GID and UID, and 2094 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2095 * 2096 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2097 * Description 2098 * Copy the **comm** attribute of the current task into *buf* of 2099 * *size_of_buf*. The **comm** attribute contains the name of 2100 * the executable (excluding the path) for the current task. The 2101 * *size_of_buf* must be strictly positive. On success, the 2102 * helper makes sure that the *buf* is NUL-terminated. On failure, 2103 * it is filled with zeroes. 2104 * Return 2105 * 0 on success, or a negative error in case of failure. 2106 * 2107 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2108 * Description 2109 * Retrieve the classid for the current task, i.e. for the net_cls 2110 * cgroup to which *skb* belongs. 2111 * 2112 * This helper can be used on TC egress path, but not on ingress. 2113 * 2114 * The net_cls cgroup provides an interface to tag network packets 2115 * based on a user-provided identifier for all traffic coming from 2116 * the tasks belonging to the related cgroup. See also the related 2117 * kernel documentation, available from the Linux sources in file 2118 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2119 * 2120 * The Linux kernel has two versions for cgroups: there are 2121 * cgroups v1 and cgroups v2. Both are available to users, who can 2122 * use a mixture of them, but note that the net_cls cgroup is for 2123 * cgroup v1 only. This makes it incompatible with BPF programs 2124 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2125 * only hold data for one version of cgroups at a time). 2126 * 2127 * This helper is only available is the kernel was compiled with 2128 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2129 * "**y**" or to "**m**". 2130 * Return 2131 * The classid, or 0 for the default unconfigured classid. 2132 * 2133 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2134 * Description 2135 * Push a *vlan_tci* (VLAN tag control information) of protocol 2136 * *vlan_proto* to the packet associated to *skb*, then update 2137 * the checksum. Note that if *vlan_proto* is different from 2138 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2139 * be **ETH_P_8021Q**. 2140 * 2141 * A call to this helper is susceptible to change the underlying 2142 * packet buffer. Therefore, at load time, all checks on pointers 2143 * previously done by the verifier are invalidated and must be 2144 * performed again, if the helper is used in combination with 2145 * direct packet access. 2146 * Return 2147 * 0 on success, or a negative error in case of failure. 2148 * 2149 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2150 * Description 2151 * Pop a VLAN header from the packet associated to *skb*. 2152 * 2153 * A call to this helper is susceptible to change the underlying 2154 * packet buffer. Therefore, at load time, all checks on pointers 2155 * previously done by the verifier are invalidated and must be 2156 * performed again, if the helper is used in combination with 2157 * direct packet access. 2158 * Return 2159 * 0 on success, or a negative error in case of failure. 2160 * 2161 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2162 * Description 2163 * Get tunnel metadata. This helper takes a pointer *key* to an 2164 * empty **struct bpf_tunnel_key** of **size**, that will be 2165 * filled with tunnel metadata for the packet associated to *skb*. 2166 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2167 * indicates that the tunnel is based on IPv6 protocol instead of 2168 * IPv4. 2169 * 2170 * The **struct bpf_tunnel_key** is an object that generalizes the 2171 * principal parameters used by various tunneling protocols into a 2172 * single struct. This way, it can be used to easily make a 2173 * decision based on the contents of the encapsulation header, 2174 * "summarized" in this struct. In particular, it holds the IP 2175 * address of the remote end (IPv4 or IPv6, depending on the case) 2176 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2177 * this struct exposes the *key*\ **->tunnel_id**, which is 2178 * generally mapped to a VNI (Virtual Network Identifier), making 2179 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2180 * () helper. 2181 * 2182 * Let's imagine that the following code is part of a program 2183 * attached to the TC ingress interface, on one end of a GRE 2184 * tunnel, and is supposed to filter out all messages coming from 2185 * remote ends with IPv4 address other than 10.0.0.1: 2186 * 2187 * :: 2188 * 2189 * int ret; 2190 * struct bpf_tunnel_key key = {}; 2191 * 2192 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2193 * if (ret < 0) 2194 * return TC_ACT_SHOT; // drop packet 2195 * 2196 * if (key.remote_ipv4 != 0x0a000001) 2197 * return TC_ACT_SHOT; // drop packet 2198 * 2199 * return TC_ACT_OK; // accept packet 2200 * 2201 * This interface can also be used with all encapsulation devices 2202 * that can operate in "collect metadata" mode: instead of having 2203 * one network device per specific configuration, the "collect 2204 * metadata" mode only requires a single device where the 2205 * configuration can be extracted from this helper. 2206 * 2207 * This can be used together with various tunnels such as VXLan, 2208 * Geneve, GRE or IP in IP (IPIP). 2209 * Return 2210 * 0 on success, or a negative error in case of failure. 2211 * 2212 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2213 * Description 2214 * Populate tunnel metadata for packet associated to *skb.* The 2215 * tunnel metadata is set to the contents of *key*, of *size*. The 2216 * *flags* can be set to a combination of the following values: 2217 * 2218 * **BPF_F_TUNINFO_IPV6** 2219 * Indicate that the tunnel is based on IPv6 protocol 2220 * instead of IPv4. 2221 * **BPF_F_ZERO_CSUM_TX** 2222 * For IPv4 packets, add a flag to tunnel metadata 2223 * indicating that checksum computation should be skipped 2224 * and checksum set to zeroes. 2225 * **BPF_F_DONT_FRAGMENT** 2226 * Add a flag to tunnel metadata indicating that the 2227 * packet should not be fragmented. 2228 * **BPF_F_SEQ_NUMBER** 2229 * Add a flag to tunnel metadata indicating that a 2230 * sequence number should be added to tunnel header before 2231 * sending the packet. This flag was added for GRE 2232 * encapsulation, but might be used with other protocols 2233 * as well in the future. 2234 * **BPF_F_NO_TUNNEL_KEY** 2235 * Add a flag to tunnel metadata indicating that no tunnel 2236 * key should be set in the resulting tunnel header. 2237 * 2238 * Here is a typical usage on the transmit path: 2239 * 2240 * :: 2241 * 2242 * struct bpf_tunnel_key key; 2243 * populate key ... 2244 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2245 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2246 * 2247 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2248 * helper for additional information. 2249 * Return 2250 * 0 on success, or a negative error in case of failure. 2251 * 2252 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2253 * Description 2254 * Read the value of a perf event counter. This helper relies on a 2255 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2256 * the perf event counter is selected when *map* is updated with 2257 * perf event file descriptors. The *map* is an array whose size 2258 * is the number of available CPUs, and each cell contains a value 2259 * relative to one CPU. The value to retrieve is indicated by 2260 * *flags*, that contains the index of the CPU to look up, masked 2261 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2262 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2263 * current CPU should be retrieved. 2264 * 2265 * Note that before Linux 4.13, only hardware perf event can be 2266 * retrieved. 2267 * 2268 * Also, be aware that the newer helper 2269 * **bpf_perf_event_read_value**\ () is recommended over 2270 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2271 * quirks where error and counter value are used as a return code 2272 * (which is wrong to do since ranges may overlap). This issue is 2273 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2274 * time provides more features over the **bpf_perf_event_read**\ 2275 * () interface. Please refer to the description of 2276 * **bpf_perf_event_read_value**\ () for details. 2277 * Return 2278 * The value of the perf event counter read from the map, or a 2279 * negative error code in case of failure. 2280 * 2281 * long bpf_redirect(u32 ifindex, u64 flags) 2282 * Description 2283 * Redirect the packet to another net device of index *ifindex*. 2284 * This helper is somewhat similar to **bpf_clone_redirect**\ 2285 * (), except that the packet is not cloned, which provides 2286 * increased performance. 2287 * 2288 * Except for XDP, both ingress and egress interfaces can be used 2289 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2290 * to make the distinction (ingress path is selected if the flag 2291 * is present, egress path otherwise). Currently, XDP only 2292 * supports redirection to the egress interface, and accepts no 2293 * flag at all. 2294 * 2295 * The same effect can also be attained with the more generic 2296 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2297 * redirect target instead of providing it directly to the helper. 2298 * Return 2299 * For XDP, the helper returns **XDP_REDIRECT** on success or 2300 * **XDP_ABORTED** on error. For other program types, the values 2301 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2302 * error. 2303 * 2304 * u32 bpf_get_route_realm(struct sk_buff *skb) 2305 * Description 2306 * Retrieve the realm or the route, that is to say the 2307 * **tclassid** field of the destination for the *skb*. The 2308 * identifier retrieved is a user-provided tag, similar to the 2309 * one used with the net_cls cgroup (see description for 2310 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2311 * held by a route (a destination entry), not by a task. 2312 * 2313 * Retrieving this identifier works with the clsact TC egress hook 2314 * (see also **tc-bpf(8)**), or alternatively on conventional 2315 * classful egress qdiscs, but not on TC ingress path. In case of 2316 * clsact TC egress hook, this has the advantage that, internally, 2317 * the destination entry has not been dropped yet in the transmit 2318 * path. Therefore, the destination entry does not need to be 2319 * artificially held via **netif_keep_dst**\ () for a classful 2320 * qdisc until the *skb* is freed. 2321 * 2322 * This helper is available only if the kernel was compiled with 2323 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2324 * Return 2325 * The realm of the route for the packet associated to *skb*, or 0 2326 * if none was found. 2327 * 2328 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2329 * Description 2330 * Write raw *data* blob into a special BPF perf event held by 2331 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2332 * event must have the following attributes: **PERF_SAMPLE_RAW** 2333 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2334 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2335 * 2336 * The *flags* are used to indicate the index in *map* for which 2337 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2338 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2339 * to indicate that the index of the current CPU core should be 2340 * used. 2341 * 2342 * The value to write, of *size*, is passed through eBPF stack and 2343 * pointed by *data*. 2344 * 2345 * The context of the program *ctx* needs also be passed to the 2346 * helper. 2347 * 2348 * On user space, a program willing to read the values needs to 2349 * call **perf_event_open**\ () on the perf event (either for 2350 * one or for all CPUs) and to store the file descriptor into the 2351 * *map*. This must be done before the eBPF program can send data 2352 * into it. An example is available in file 2353 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2354 * tree (the eBPF program counterpart is in 2355 * *samples/bpf/trace_output_kern.c*). 2356 * 2357 * **bpf_perf_event_output**\ () achieves better performance 2358 * than **bpf_trace_printk**\ () for sharing data with user 2359 * space, and is much better suitable for streaming data from eBPF 2360 * programs. 2361 * 2362 * Note that this helper is not restricted to tracing use cases 2363 * and can be used with programs attached to TC or XDP as well, 2364 * where it allows for passing data to user space listeners. Data 2365 * can be: 2366 * 2367 * * Only custom structs, 2368 * * Only the packet payload, or 2369 * * A combination of both. 2370 * Return 2371 * 0 on success, or a negative error in case of failure. 2372 * 2373 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2374 * Description 2375 * This helper was provided as an easy way to load data from a 2376 * packet. It can be used to load *len* bytes from *offset* from 2377 * the packet associated to *skb*, into the buffer pointed by 2378 * *to*. 2379 * 2380 * Since Linux 4.7, usage of this helper has mostly been replaced 2381 * by "direct packet access", enabling packet data to be 2382 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2383 * pointing respectively to the first byte of packet data and to 2384 * the byte after the last byte of packet data. However, it 2385 * remains useful if one wishes to read large quantities of data 2386 * at once from a packet into the eBPF stack. 2387 * Return 2388 * 0 on success, or a negative error in case of failure. 2389 * 2390 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2391 * Description 2392 * Walk a user or a kernel stack and return its id. To achieve 2393 * this, the helper needs *ctx*, which is a pointer to the context 2394 * on which the tracing program is executed, and a pointer to a 2395 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2396 * 2397 * The last argument, *flags*, holds the number of stack frames to 2398 * skip (from 0 to 255), masked with 2399 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2400 * a combination of the following flags: 2401 * 2402 * **BPF_F_USER_STACK** 2403 * Collect a user space stack instead of a kernel stack. 2404 * **BPF_F_FAST_STACK_CMP** 2405 * Compare stacks by hash only. 2406 * **BPF_F_REUSE_STACKID** 2407 * If two different stacks hash into the same *stackid*, 2408 * discard the old one. 2409 * 2410 * The stack id retrieved is a 32 bit long integer handle which 2411 * can be further combined with other data (including other stack 2412 * ids) and used as a key into maps. This can be useful for 2413 * generating a variety of graphs (such as flame graphs or off-cpu 2414 * graphs). 2415 * 2416 * For walking a stack, this helper is an improvement over 2417 * **bpf_probe_read**\ (), which can be used with unrolled loops 2418 * but is not efficient and consumes a lot of eBPF instructions. 2419 * Instead, **bpf_get_stackid**\ () can collect up to 2420 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2421 * this limit can be controlled with the **sysctl** program, and 2422 * that it should be manually increased in order to profile long 2423 * user stacks (such as stacks for Java programs). To do so, use: 2424 * 2425 * :: 2426 * 2427 * # sysctl kernel.perf_event_max_stack=<new value> 2428 * Return 2429 * The positive or null stack id on success, or a negative error 2430 * in case of failure. 2431 * 2432 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2433 * Description 2434 * Compute a checksum difference, from the raw buffer pointed by 2435 * *from*, of length *from_size* (that must be a multiple of 4), 2436 * towards the raw buffer pointed by *to*, of size *to_size* 2437 * (same remark). An optional *seed* can be added to the value 2438 * (this can be cascaded, the seed may come from a previous call 2439 * to the helper). 2440 * 2441 * This is flexible enough to be used in several ways: 2442 * 2443 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2444 * checksum, it can be used when pushing new data. 2445 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2446 * checksum, it can be used when removing data from a packet. 2447 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2448 * can be used to compute a diff. Note that *from_size* and 2449 * *to_size* do not need to be equal. 2450 * 2451 * This helper can be used in combination with 2452 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2453 * which one can feed in the difference computed with 2454 * **bpf_csum_diff**\ (). 2455 * Return 2456 * The checksum result, or a negative error code in case of 2457 * failure. 2458 * 2459 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2460 * Description 2461 * Retrieve tunnel options metadata for the packet associated to 2462 * *skb*, and store the raw tunnel option data to the buffer *opt* 2463 * of *size*. 2464 * 2465 * This helper can be used with encapsulation devices that can 2466 * operate in "collect metadata" mode (please refer to the related 2467 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2468 * more details). A particular example where this can be used is 2469 * in combination with the Geneve encapsulation protocol, where it 2470 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2471 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2472 * the eBPF program. This allows for full customization of these 2473 * headers. 2474 * Return 2475 * The size of the option data retrieved. 2476 * 2477 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2478 * Description 2479 * Set tunnel options metadata for the packet associated to *skb* 2480 * to the option data contained in the raw buffer *opt* of *size*. 2481 * 2482 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2483 * helper for additional information. 2484 * Return 2485 * 0 on success, or a negative error in case of failure. 2486 * 2487 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2488 * Description 2489 * Change the protocol of the *skb* to *proto*. Currently 2490 * supported are transition from IPv4 to IPv6, and from IPv6 to 2491 * IPv4. The helper takes care of the groundwork for the 2492 * transition, including resizing the socket buffer. The eBPF 2493 * program is expected to fill the new headers, if any, via 2494 * **skb_store_bytes**\ () and to recompute the checksums with 2495 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2496 * (). The main case for this helper is to perform NAT64 2497 * operations out of an eBPF program. 2498 * 2499 * Internally, the GSO type is marked as dodgy so that headers are 2500 * checked and segments are recalculated by the GSO/GRO engine. 2501 * The size for GSO target is adapted as well. 2502 * 2503 * All values for *flags* are reserved for future usage, and must 2504 * be left at zero. 2505 * 2506 * A call to this helper is susceptible to change the underlying 2507 * packet buffer. Therefore, at load time, all checks on pointers 2508 * previously done by the verifier are invalidated and must be 2509 * performed again, if the helper is used in combination with 2510 * direct packet access. 2511 * Return 2512 * 0 on success, or a negative error in case of failure. 2513 * 2514 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2515 * Description 2516 * Change the packet type for the packet associated to *skb*. This 2517 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2518 * the eBPF program does not have a write access to *skb*\ 2519 * **->pkt_type** beside this helper. Using a helper here allows 2520 * for graceful handling of errors. 2521 * 2522 * The major use case is to change incoming *skb*s to 2523 * **PACKET_HOST** in a programmatic way instead of having to 2524 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2525 * example. 2526 * 2527 * Note that *type* only allows certain values. At this time, they 2528 * are: 2529 * 2530 * **PACKET_HOST** 2531 * Packet is for us. 2532 * **PACKET_BROADCAST** 2533 * Send packet to all. 2534 * **PACKET_MULTICAST** 2535 * Send packet to group. 2536 * **PACKET_OTHERHOST** 2537 * Send packet to someone else. 2538 * Return 2539 * 0 on success, or a negative error in case of failure. 2540 * 2541 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2542 * Description 2543 * Check whether *skb* is a descendant of the cgroup2 held by 2544 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2545 * Return 2546 * The return value depends on the result of the test, and can be: 2547 * 2548 * * 0, if the *skb* failed the cgroup2 descendant test. 2549 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2550 * * A negative error code, if an error occurred. 2551 * 2552 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2553 * Description 2554 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2555 * not set, in particular if the hash was cleared due to mangling, 2556 * recompute this hash. Later accesses to the hash can be done 2557 * directly with *skb*\ **->hash**. 2558 * 2559 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2560 * prototype with **bpf_skb_change_proto**\ (), or calling 2561 * **bpf_skb_store_bytes**\ () with the 2562 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2563 * the hash and to trigger a new computation for the next call to 2564 * **bpf_get_hash_recalc**\ (). 2565 * Return 2566 * The 32-bit hash. 2567 * 2568 * u64 bpf_get_current_task(void) 2569 * Description 2570 * Get the current task. 2571 * Return 2572 * A pointer to the current task struct. 2573 * 2574 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2575 * Description 2576 * Attempt in a safe way to write *len* bytes from the buffer 2577 * *src* to *dst* in memory. It only works for threads that are in 2578 * user context, and *dst* must be a valid user space address. 2579 * 2580 * This helper should not be used to implement any kind of 2581 * security mechanism because of TOC-TOU attacks, but rather to 2582 * debug, divert, and manipulate execution of semi-cooperative 2583 * processes. 2584 * 2585 * Keep in mind that this feature is meant for experiments, and it 2586 * has a risk of crashing the system and running programs. 2587 * Therefore, when an eBPF program using this helper is attached, 2588 * a warning including PID and process name is printed to kernel 2589 * logs. 2590 * Return 2591 * 0 on success, or a negative error in case of failure. 2592 * 2593 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2594 * Description 2595 * Check whether the probe is being run is the context of a given 2596 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2597 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2598 * Return 2599 * The return value depends on the result of the test, and can be: 2600 * 2601 * * 1, if current task belongs to the cgroup2. 2602 * * 0, if current task does not belong to the cgroup2. 2603 * * A negative error code, if an error occurred. 2604 * 2605 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2606 * Description 2607 * Resize (trim or grow) the packet associated to *skb* to the 2608 * new *len*. The *flags* are reserved for future usage, and must 2609 * be left at zero. 2610 * 2611 * The basic idea is that the helper performs the needed work to 2612 * change the size of the packet, then the eBPF program rewrites 2613 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2614 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2615 * and others. This helper is a slow path utility intended for 2616 * replies with control messages. And because it is targeted for 2617 * slow path, the helper itself can afford to be slow: it 2618 * implicitly linearizes, unclones and drops offloads from the 2619 * *skb*. 2620 * 2621 * A call to this helper is susceptible to change the underlying 2622 * packet buffer. Therefore, at load time, all checks on pointers 2623 * previously done by the verifier are invalidated and must be 2624 * performed again, if the helper is used in combination with 2625 * direct packet access. 2626 * Return 2627 * 0 on success, or a negative error in case of failure. 2628 * 2629 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2630 * Description 2631 * Pull in non-linear data in case the *skb* is non-linear and not 2632 * all of *len* are part of the linear section. Make *len* bytes 2633 * from *skb* readable and writable. If a zero value is passed for 2634 * *len*, then all bytes in the linear part of *skb* will be made 2635 * readable and writable. 2636 * 2637 * This helper is only needed for reading and writing with direct 2638 * packet access. 2639 * 2640 * For direct packet access, testing that offsets to access 2641 * are within packet boundaries (test on *skb*\ **->data_end**) is 2642 * susceptible to fail if offsets are invalid, or if the requested 2643 * data is in non-linear parts of the *skb*. On failure the 2644 * program can just bail out, or in the case of a non-linear 2645 * buffer, use a helper to make the data available. The 2646 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2647 * the data. Another one consists in using **bpf_skb_pull_data** 2648 * to pull in once the non-linear parts, then retesting and 2649 * eventually access the data. 2650 * 2651 * At the same time, this also makes sure the *skb* is uncloned, 2652 * which is a necessary condition for direct write. As this needs 2653 * to be an invariant for the write part only, the verifier 2654 * detects writes and adds a prologue that is calling 2655 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2656 * the very beginning in case it is indeed cloned. 2657 * 2658 * A call to this helper is susceptible to change the underlying 2659 * packet buffer. Therefore, at load time, all checks on pointers 2660 * previously done by the verifier are invalidated and must be 2661 * performed again, if the helper is used in combination with 2662 * direct packet access. 2663 * Return 2664 * 0 on success, or a negative error in case of failure. 2665 * 2666 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2667 * Description 2668 * Add the checksum *csum* into *skb*\ **->csum** in case the 2669 * driver has supplied a checksum for the entire packet into that 2670 * field. Return an error otherwise. This helper is intended to be 2671 * used in combination with **bpf_csum_diff**\ (), in particular 2672 * when the checksum needs to be updated after data has been 2673 * written into the packet through direct packet access. 2674 * Return 2675 * The checksum on success, or a negative error code in case of 2676 * failure. 2677 * 2678 * void bpf_set_hash_invalid(struct sk_buff *skb) 2679 * Description 2680 * Invalidate the current *skb*\ **->hash**. It can be used after 2681 * mangling on headers through direct packet access, in order to 2682 * indicate that the hash is outdated and to trigger a 2683 * recalculation the next time the kernel tries to access this 2684 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2685 * Return 2686 * void. 2687 * 2688 * long bpf_get_numa_node_id(void) 2689 * Description 2690 * Return the id of the current NUMA node. The primary use case 2691 * for this helper is the selection of sockets for the local NUMA 2692 * node, when the program is attached to sockets using the 2693 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2694 * but the helper is also available to other eBPF program types, 2695 * similarly to **bpf_get_smp_processor_id**\ (). 2696 * Return 2697 * The id of current NUMA node. 2698 * 2699 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2700 * Description 2701 * Grows headroom of packet associated to *skb* and adjusts the 2702 * offset of the MAC header accordingly, adding *len* bytes of 2703 * space. It automatically extends and reallocates memory as 2704 * required. 2705 * 2706 * This helper can be used on a layer 3 *skb* to push a MAC header 2707 * for redirection into a layer 2 device. 2708 * 2709 * All values for *flags* are reserved for future usage, and must 2710 * be left at zero. 2711 * 2712 * A call to this helper is susceptible to change the underlying 2713 * packet buffer. Therefore, at load time, all checks on pointers 2714 * previously done by the verifier are invalidated and must be 2715 * performed again, if the helper is used in combination with 2716 * direct packet access. 2717 * Return 2718 * 0 on success, or a negative error in case of failure. 2719 * 2720 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2721 * Description 2722 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2723 * it is possible to use a negative value for *delta*. This helper 2724 * can be used to prepare the packet for pushing or popping 2725 * headers. 2726 * 2727 * A call to this helper is susceptible to change the underlying 2728 * packet buffer. Therefore, at load time, all checks on pointers 2729 * previously done by the verifier are invalidated and must be 2730 * performed again, if the helper is used in combination with 2731 * direct packet access. 2732 * Return 2733 * 0 on success, or a negative error in case of failure. 2734 * 2735 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2736 * Description 2737 * Copy a NUL terminated string from an unsafe kernel address 2738 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2739 * more details. 2740 * 2741 * Generally, use **bpf_probe_read_user_str**\ () or 2742 * **bpf_probe_read_kernel_str**\ () instead. 2743 * Return 2744 * On success, the strictly positive length of the string, 2745 * including the trailing NUL character. On error, a negative 2746 * value. 2747 * 2748 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2749 * Description 2750 * If the **struct sk_buff** pointed by *skb* has a known socket, 2751 * retrieve the cookie (generated by the kernel) of this socket. 2752 * If no cookie has been set yet, generate a new cookie. Once 2753 * generated, the socket cookie remains stable for the life of the 2754 * socket. This helper can be useful for monitoring per socket 2755 * networking traffic statistics as it provides a global socket 2756 * identifier that can be assumed unique. 2757 * Return 2758 * A 8-byte long unique number on success, or 0 if the socket 2759 * field is missing inside *skb*. 2760 * 2761 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2762 * Description 2763 * Equivalent to bpf_get_socket_cookie() helper that accepts 2764 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2765 * Return 2766 * A 8-byte long unique number. 2767 * 2768 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2769 * Description 2770 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2771 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2772 * Return 2773 * A 8-byte long unique number. 2774 * 2775 * u64 bpf_get_socket_cookie(struct sock *sk) 2776 * Description 2777 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2778 * *sk*, but gets socket from a BTF **struct sock**. This helper 2779 * also works for sleepable programs. 2780 * Return 2781 * A 8-byte long unique number or 0 if *sk* is NULL. 2782 * 2783 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2784 * Description 2785 * Get the owner UID of the socked associated to *skb*. 2786 * Return 2787 * The owner UID of the socket associated to *skb*. If the socket 2788 * is **NULL**, or if it is not a full socket (i.e. if it is a 2789 * time-wait or a request socket instead), **overflowuid** value 2790 * is returned (note that **overflowuid** might also be the actual 2791 * UID value for the socket). 2792 * 2793 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2794 * Description 2795 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2796 * to value *hash*. 2797 * Return 2798 * 0 2799 * 2800 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2801 * Description 2802 * Emulate a call to **setsockopt()** on the socket associated to 2803 * *bpf_socket*, which must be a full socket. The *level* at 2804 * which the option resides and the name *optname* of the option 2805 * must be specified, see **setsockopt(2)** for more information. 2806 * The option value of length *optlen* is pointed by *optval*. 2807 * 2808 * *bpf_socket* should be one of the following: 2809 * 2810 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2811 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2812 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2813 * 2814 * This helper actually implements a subset of **setsockopt()**. 2815 * It supports the following *level*\ s: 2816 * 2817 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2818 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2819 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2820 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2821 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2822 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2823 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2824 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2825 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2826 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2827 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2828 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2829 * **TCP_BPF_RTO_MIN**. 2830 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2831 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2832 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2833 * Return 2834 * 0 on success, or a negative error in case of failure. 2835 * 2836 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2837 * Description 2838 * Grow or shrink the room for data in the packet associated to 2839 * *skb* by *len_diff*, and according to the selected *mode*. 2840 * 2841 * By default, the helper will reset any offloaded checksum 2842 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2843 * by the following flag: 2844 * 2845 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2846 * checksum data of the skb to CHECKSUM_NONE. 2847 * 2848 * There are two supported modes at this time: 2849 * 2850 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2851 * (room space is added or removed between the layer 2 and 2852 * layer 3 headers). 2853 * 2854 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2855 * (room space is added or removed between the layer 3 and 2856 * layer 4 headers). 2857 * 2858 * The following flags are supported at this time: 2859 * 2860 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2861 * Adjusting mss in this way is not allowed for datagrams. 2862 * 2863 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2864 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2865 * Any new space is reserved to hold a tunnel header. 2866 * Configure skb offsets and other fields accordingly. 2867 * 2868 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2869 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2870 * Use with ENCAP_L3 flags to further specify the tunnel type. 2871 * 2872 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2873 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2874 * type; *len* is the length of the inner MAC header. 2875 * 2876 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2877 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2878 * L2 type as Ethernet. 2879 * 2880 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2881 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2882 * Indicate the new IP header version after decapsulating the outer 2883 * IP header. Used when the inner and outer IP versions are different. 2884 * 2885 * A call to this helper is susceptible to change the underlying 2886 * packet buffer. Therefore, at load time, all checks on pointers 2887 * previously done by the verifier are invalidated and must be 2888 * performed again, if the helper is used in combination with 2889 * direct packet access. 2890 * Return 2891 * 0 on success, or a negative error in case of failure. 2892 * 2893 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2894 * Description 2895 * Redirect the packet to the endpoint referenced by *map* at 2896 * index *key*. Depending on its type, this *map* can contain 2897 * references to net devices (for forwarding packets through other 2898 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2899 * but this is only implemented for native XDP (with driver 2900 * support) as of this writing). 2901 * 2902 * The lower two bits of *flags* are used as the return code if 2903 * the map lookup fails. This is so that the return value can be 2904 * one of the XDP program return codes up to **XDP_TX**, as chosen 2905 * by the caller. The higher bits of *flags* can be set to 2906 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2907 * 2908 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2909 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2910 * interface will be excluded when do broadcasting. 2911 * 2912 * See also **bpf_redirect**\ (), which only supports redirecting 2913 * to an ifindex, but doesn't require a map to do so. 2914 * Return 2915 * **XDP_REDIRECT** on success, or the value of the two lower bits 2916 * of the *flags* argument on error. 2917 * 2918 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2919 * Description 2920 * Redirect the packet to the socket referenced by *map* (of type 2921 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2922 * egress interfaces can be used for redirection. The 2923 * **BPF_F_INGRESS** value in *flags* is used to make the 2924 * distinction (ingress path is selected if the flag is present, 2925 * egress path otherwise). This is the only flag supported for now. 2926 * Return 2927 * **SK_PASS** on success, or **SK_DROP** on error. 2928 * 2929 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2930 * Description 2931 * Add an entry to, or update a *map* referencing sockets. The 2932 * *skops* is used as a new value for the entry associated to 2933 * *key*. *flags* is one of: 2934 * 2935 * **BPF_NOEXIST** 2936 * The entry for *key* must not exist in the map. 2937 * **BPF_EXIST** 2938 * The entry for *key* must already exist in the map. 2939 * **BPF_ANY** 2940 * No condition on the existence of the entry for *key*. 2941 * 2942 * If the *map* has eBPF programs (parser and verdict), those will 2943 * be inherited by the socket being added. If the socket is 2944 * already attached to eBPF programs, this results in an error. 2945 * Return 2946 * 0 on success, or a negative error in case of failure. 2947 * 2948 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2949 * Description 2950 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2951 * *delta* (which can be positive or negative). Note that this 2952 * operation modifies the address stored in *xdp_md*\ **->data**, 2953 * so the latter must be loaded only after the helper has been 2954 * called. 2955 * 2956 * The use of *xdp_md*\ **->data_meta** is optional and programs 2957 * are not required to use it. The rationale is that when the 2958 * packet is processed with XDP (e.g. as DoS filter), it is 2959 * possible to push further meta data along with it before passing 2960 * to the stack, and to give the guarantee that an ingress eBPF 2961 * program attached as a TC classifier on the same device can pick 2962 * this up for further post-processing. Since TC works with socket 2963 * buffers, it remains possible to set from XDP the **mark** or 2964 * **priority** pointers, or other pointers for the socket buffer. 2965 * Having this scratch space generic and programmable allows for 2966 * more flexibility as the user is free to store whatever meta 2967 * data they need. 2968 * 2969 * A call to this helper is susceptible to change the underlying 2970 * packet buffer. Therefore, at load time, all checks on pointers 2971 * previously done by the verifier are invalidated and must be 2972 * performed again, if the helper is used in combination with 2973 * direct packet access. 2974 * Return 2975 * 0 on success, or a negative error in case of failure. 2976 * 2977 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2978 * Description 2979 * Read the value of a perf event counter, and store it into *buf* 2980 * of size *buf_size*. This helper relies on a *map* of type 2981 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2982 * counter is selected when *map* is updated with perf event file 2983 * descriptors. The *map* is an array whose size is the number of 2984 * available CPUs, and each cell contains a value relative to one 2985 * CPU. The value to retrieve is indicated by *flags*, that 2986 * contains the index of the CPU to look up, masked with 2987 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2988 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2989 * current CPU should be retrieved. 2990 * 2991 * This helper behaves in a way close to 2992 * **bpf_perf_event_read**\ () helper, save that instead of 2993 * just returning the value observed, it fills the *buf* 2994 * structure. This allows for additional data to be retrieved: in 2995 * particular, the enabled and running times (in *buf*\ 2996 * **->enabled** and *buf*\ **->running**, respectively) are 2997 * copied. In general, **bpf_perf_event_read_value**\ () is 2998 * recommended over **bpf_perf_event_read**\ (), which has some 2999 * ABI issues and provides fewer functionalities. 3000 * 3001 * These values are interesting, because hardware PMU (Performance 3002 * Monitoring Unit) counters are limited resources. When there are 3003 * more PMU based perf events opened than available counters, 3004 * kernel will multiplex these events so each event gets certain 3005 * percentage (but not all) of the PMU time. In case that 3006 * multiplexing happens, the number of samples or counter value 3007 * will not reflect the case compared to when no multiplexing 3008 * occurs. This makes comparison between different runs difficult. 3009 * Typically, the counter value should be normalized before 3010 * comparing to other experiments. The usual normalization is done 3011 * as follows. 3012 * 3013 * :: 3014 * 3015 * normalized_counter = counter * t_enabled / t_running 3016 * 3017 * Where t_enabled is the time enabled for event and t_running is 3018 * the time running for event since last normalization. The 3019 * enabled and running times are accumulated since the perf event 3020 * open. To achieve scaling factor between two invocations of an 3021 * eBPF program, users can use CPU id as the key (which is 3022 * typical for perf array usage model) to remember the previous 3023 * value and do the calculation inside the eBPF program. 3024 * Return 3025 * 0 on success, or a negative error in case of failure. 3026 * 3027 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3028 * Description 3029 * For an eBPF program attached to a perf event, retrieve the 3030 * value of the event counter associated to *ctx* and store it in 3031 * the structure pointed by *buf* and of size *buf_size*. Enabled 3032 * and running times are also stored in the structure (see 3033 * description of helper **bpf_perf_event_read_value**\ () for 3034 * more details). 3035 * Return 3036 * 0 on success, or a negative error in case of failure. 3037 * 3038 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3039 * Description 3040 * Emulate a call to **getsockopt()** on the socket associated to 3041 * *bpf_socket*, which must be a full socket. The *level* at 3042 * which the option resides and the name *optname* of the option 3043 * must be specified, see **getsockopt(2)** for more information. 3044 * The retrieved value is stored in the structure pointed by 3045 * *opval* and of length *optlen*. 3046 * 3047 * *bpf_socket* should be one of the following: 3048 * 3049 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3050 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3051 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3052 * 3053 * This helper actually implements a subset of **getsockopt()**. 3054 * It supports the same set of *optname*\ s that is supported by 3055 * the **bpf_setsockopt**\ () helper. The exceptions are 3056 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3057 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3058 * Return 3059 * 0 on success, or a negative error in case of failure. 3060 * 3061 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3062 * Description 3063 * Used for error injection, this helper uses kprobes to override 3064 * the return value of the probed function, and to set it to *rc*. 3065 * The first argument is the context *regs* on which the kprobe 3066 * works. 3067 * 3068 * This helper works by setting the PC (program counter) 3069 * to an override function which is run in place of the original 3070 * probed function. This means the probed function is not run at 3071 * all. The replacement function just returns with the required 3072 * value. 3073 * 3074 * This helper has security implications, and thus is subject to 3075 * restrictions. It is only available if the kernel was compiled 3076 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3077 * option, and in this case it only works on functions tagged with 3078 * **ALLOW_ERROR_INJECTION** in the kernel code. 3079 * 3080 * Also, the helper is only available for the architectures having 3081 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 3082 * x86 architecture is the only one to support this feature. 3083 * Return 3084 * 0 3085 * 3086 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3087 * Description 3088 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3089 * for the full TCP socket associated to *bpf_sock_ops* to 3090 * *argval*. 3091 * 3092 * The primary use of this field is to determine if there should 3093 * be calls to eBPF programs of type 3094 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3095 * code. A program of the same type can change its value, per 3096 * connection and as necessary, when the connection is 3097 * established. This field is directly accessible for reading, but 3098 * this helper must be used for updates in order to return an 3099 * error if an eBPF program tries to set a callback that is not 3100 * supported in the current kernel. 3101 * 3102 * *argval* is a flag array which can combine these flags: 3103 * 3104 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3105 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3106 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3107 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3108 * 3109 * Therefore, this function can be used to clear a callback flag by 3110 * setting the appropriate bit to zero. e.g. to disable the RTO 3111 * callback: 3112 * 3113 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3114 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3115 * 3116 * Here are some examples of where one could call such eBPF 3117 * program: 3118 * 3119 * * When RTO fires. 3120 * * When a packet is retransmitted. 3121 * * When the connection terminates. 3122 * * When a packet is sent. 3123 * * When a packet is received. 3124 * Return 3125 * Code **-EINVAL** if the socket is not a full TCP socket; 3126 * otherwise, a positive number containing the bits that could not 3127 * be set is returned (which comes down to 0 if all bits were set 3128 * as required). 3129 * 3130 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3131 * Description 3132 * This helper is used in programs implementing policies at the 3133 * socket level. If the message *msg* is allowed to pass (i.e. if 3134 * the verdict eBPF program returns **SK_PASS**), redirect it to 3135 * the socket referenced by *map* (of type 3136 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3137 * egress interfaces can be used for redirection. The 3138 * **BPF_F_INGRESS** value in *flags* is used to make the 3139 * distinction (ingress path is selected if the flag is present, 3140 * egress path otherwise). This is the only flag supported for now. 3141 * Return 3142 * **SK_PASS** on success, or **SK_DROP** on error. 3143 * 3144 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3145 * Description 3146 * For socket policies, apply the verdict of the eBPF program to 3147 * the next *bytes* (number of bytes) of message *msg*. 3148 * 3149 * For example, this helper can be used in the following cases: 3150 * 3151 * * A single **sendmsg**\ () or **sendfile**\ () system call 3152 * contains multiple logical messages that the eBPF program is 3153 * supposed to read and for which it should apply a verdict. 3154 * * An eBPF program only cares to read the first *bytes* of a 3155 * *msg*. If the message has a large payload, then setting up 3156 * and calling the eBPF program repeatedly for all bytes, even 3157 * though the verdict is already known, would create unnecessary 3158 * overhead. 3159 * 3160 * When called from within an eBPF program, the helper sets a 3161 * counter internal to the BPF infrastructure, that is used to 3162 * apply the last verdict to the next *bytes*. If *bytes* is 3163 * smaller than the current data being processed from a 3164 * **sendmsg**\ () or **sendfile**\ () system call, the first 3165 * *bytes* will be sent and the eBPF program will be re-run with 3166 * the pointer for start of data pointing to byte number *bytes* 3167 * **+ 1**. If *bytes* is larger than the current data being 3168 * processed, then the eBPF verdict will be applied to multiple 3169 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3170 * consumed. 3171 * 3172 * Note that if a socket closes with the internal counter holding 3173 * a non-zero value, this is not a problem because data is not 3174 * being buffered for *bytes* and is sent as it is received. 3175 * Return 3176 * 0 3177 * 3178 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3179 * Description 3180 * For socket policies, prevent the execution of the verdict eBPF 3181 * program for message *msg* until *bytes* (byte number) have been 3182 * accumulated. 3183 * 3184 * This can be used when one needs a specific number of bytes 3185 * before a verdict can be assigned, even if the data spans 3186 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3187 * case would be a user calling **sendmsg**\ () repeatedly with 3188 * 1-byte long message segments. Obviously, this is bad for 3189 * performance, but it is still valid. If the eBPF program needs 3190 * *bytes* bytes to validate a header, this helper can be used to 3191 * prevent the eBPF program to be called again until *bytes* have 3192 * been accumulated. 3193 * Return 3194 * 0 3195 * 3196 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3197 * Description 3198 * For socket policies, pull in non-linear data from user space 3199 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3200 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3201 * respectively. 3202 * 3203 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3204 * *msg* it can only parse data that the (**data**, **data_end**) 3205 * pointers have already consumed. For **sendmsg**\ () hooks this 3206 * is likely the first scatterlist element. But for calls relying 3207 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3208 * be the range (**0**, **0**) because the data is shared with 3209 * user space and by default the objective is to avoid allowing 3210 * user space to modify data while (or after) eBPF verdict is 3211 * being decided. This helper can be used to pull in data and to 3212 * set the start and end pointer to given values. Data will be 3213 * copied if necessary (i.e. if data was not linear and if start 3214 * and end pointers do not point to the same chunk). 3215 * 3216 * A call to this helper is susceptible to change the underlying 3217 * packet buffer. Therefore, at load time, all checks on pointers 3218 * previously done by the verifier are invalidated and must be 3219 * performed again, if the helper is used in combination with 3220 * direct packet access. 3221 * 3222 * All values for *flags* are reserved for future usage, and must 3223 * be left at zero. 3224 * Return 3225 * 0 on success, or a negative error in case of failure. 3226 * 3227 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3228 * Description 3229 * Bind the socket associated to *ctx* to the address pointed by 3230 * *addr*, of length *addr_len*. This allows for making outgoing 3231 * connection from the desired IP address, which can be useful for 3232 * example when all processes inside a cgroup should use one 3233 * single IP address on a host that has multiple IP configured. 3234 * 3235 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3236 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3237 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3238 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3239 * behavior and lets the kernel efficiently pick up an unused 3240 * port as long as 4-tuple is unique. Passing non-zero port might 3241 * lead to degraded performance. 3242 * Return 3243 * 0 on success, or a negative error in case of failure. 3244 * 3245 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3246 * Description 3247 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3248 * possible to both shrink and grow the packet tail. 3249 * Shrink done via *delta* being a negative integer. 3250 * 3251 * A call to this helper is susceptible to change the underlying 3252 * packet buffer. Therefore, at load time, all checks on pointers 3253 * previously done by the verifier are invalidated and must be 3254 * performed again, if the helper is used in combination with 3255 * direct packet access. 3256 * Return 3257 * 0 on success, or a negative error in case of failure. 3258 * 3259 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3260 * Description 3261 * Retrieve the XFRM state (IP transform framework, see also 3262 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3263 * 3264 * The retrieved value is stored in the **struct bpf_xfrm_state** 3265 * pointed by *xfrm_state* and of length *size*. 3266 * 3267 * All values for *flags* are reserved for future usage, and must 3268 * be left at zero. 3269 * 3270 * This helper is available only if the kernel was compiled with 3271 * **CONFIG_XFRM** configuration option. 3272 * Return 3273 * 0 on success, or a negative error in case of failure. 3274 * 3275 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3276 * Description 3277 * Return a user or a kernel stack in bpf program provided buffer. 3278 * To achieve this, the helper needs *ctx*, which is a pointer 3279 * to the context on which the tracing program is executed. 3280 * To store the stacktrace, the bpf program provides *buf* with 3281 * a nonnegative *size*. 3282 * 3283 * The last argument, *flags*, holds the number of stack frames to 3284 * skip (from 0 to 255), masked with 3285 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3286 * the following flags: 3287 * 3288 * **BPF_F_USER_STACK** 3289 * Collect a user space stack instead of a kernel stack. 3290 * **BPF_F_USER_BUILD_ID** 3291 * Collect (build_id, file_offset) instead of ips for user 3292 * stack, only valid if **BPF_F_USER_STACK** is also 3293 * specified. 3294 * 3295 * *file_offset* is an offset relative to the beginning 3296 * of the executable or shared object file backing the vma 3297 * which the *ip* falls in. It is *not* an offset relative 3298 * to that object's base address. Accordingly, it must be 3299 * adjusted by adding (sh_addr - sh_offset), where 3300 * sh_{addr,offset} correspond to the executable section 3301 * containing *file_offset* in the object, for comparisons 3302 * to symbols' st_value to be valid. 3303 * 3304 * **bpf_get_stack**\ () can collect up to 3305 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3306 * to sufficient large buffer size. Note that 3307 * this limit can be controlled with the **sysctl** program, and 3308 * that it should be manually increased in order to profile long 3309 * user stacks (such as stacks for Java programs). To do so, use: 3310 * 3311 * :: 3312 * 3313 * # sysctl kernel.perf_event_max_stack=<new value> 3314 * Return 3315 * The non-negative copied *buf* length equal to or less than 3316 * *size* on success, or a negative error in case of failure. 3317 * 3318 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3319 * Description 3320 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3321 * it provides an easy way to load *len* bytes from *offset* 3322 * from the packet associated to *skb*, into the buffer pointed 3323 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3324 * a fifth argument *start_header* exists in order to select a 3325 * base offset to start from. *start_header* can be one of: 3326 * 3327 * **BPF_HDR_START_MAC** 3328 * Base offset to load data from is *skb*'s mac header. 3329 * **BPF_HDR_START_NET** 3330 * Base offset to load data from is *skb*'s network header. 3331 * 3332 * In general, "direct packet access" is the preferred method to 3333 * access packet data, however, this helper is in particular useful 3334 * in socket filters where *skb*\ **->data** does not always point 3335 * to the start of the mac header and where "direct packet access" 3336 * is not available. 3337 * Return 3338 * 0 on success, or a negative error in case of failure. 3339 * 3340 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3341 * Description 3342 * Do FIB lookup in kernel tables using parameters in *params*. 3343 * If lookup is successful and result shows packet is to be 3344 * forwarded, the neighbor tables are searched for the nexthop. 3345 * If successful (ie., FIB lookup shows forwarding and nexthop 3346 * is resolved), the nexthop address is returned in ipv4_dst 3347 * or ipv6_dst based on family, smac is set to mac address of 3348 * egress device, dmac is set to nexthop mac address, rt_metric 3349 * is set to metric from route (IPv4/IPv6 only), and ifindex 3350 * is set to the device index of the nexthop from the FIB lookup. 3351 * 3352 * *plen* argument is the size of the passed in struct. 3353 * *flags* argument can be a combination of one or more of the 3354 * following values: 3355 * 3356 * **BPF_FIB_LOOKUP_DIRECT** 3357 * Do a direct table lookup vs full lookup using FIB 3358 * rules. 3359 * **BPF_FIB_LOOKUP_TBID** 3360 * Used with BPF_FIB_LOOKUP_DIRECT. 3361 * Use the routing table ID present in *params*->tbid 3362 * for the fib lookup. 3363 * **BPF_FIB_LOOKUP_OUTPUT** 3364 * Perform lookup from an egress perspective (default is 3365 * ingress). 3366 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3367 * Skip the neighbour table lookup. *params*->dmac 3368 * and *params*->smac will not be set as output. A common 3369 * use case is to call **bpf_redirect_neigh**\ () after 3370 * doing **bpf_fib_lookup**\ (). 3371 * **BPF_FIB_LOOKUP_SRC** 3372 * Derive and set source IP addr in *params*->ipv{4,6}_src 3373 * for the nexthop. If the src addr cannot be derived, 3374 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3375 * case, *params*->dmac and *params*->smac are not set either. 3376 * 3377 * *ctx* is either **struct xdp_md** for XDP programs or 3378 * **struct sk_buff** tc cls_act programs. 3379 * Return 3380 * * < 0 if any input argument is invalid 3381 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3382 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3383 * packet is not forwarded or needs assist from full stack 3384 * 3385 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3386 * was exceeded and output params->mtu_result contains the MTU. 3387 * 3388 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3389 * Description 3390 * Add an entry to, or update a sockhash *map* referencing sockets. 3391 * The *skops* is used as a new value for the entry associated to 3392 * *key*. *flags* is one of: 3393 * 3394 * **BPF_NOEXIST** 3395 * The entry for *key* must not exist in the map. 3396 * **BPF_EXIST** 3397 * The entry for *key* must already exist in the map. 3398 * **BPF_ANY** 3399 * No condition on the existence of the entry for *key*. 3400 * 3401 * If the *map* has eBPF programs (parser and verdict), those will 3402 * be inherited by the socket being added. If the socket is 3403 * already attached to eBPF programs, this results in an error. 3404 * Return 3405 * 0 on success, or a negative error in case of failure. 3406 * 3407 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3408 * Description 3409 * This helper is used in programs implementing policies at the 3410 * socket level. If the message *msg* is allowed to pass (i.e. if 3411 * the verdict eBPF program returns **SK_PASS**), redirect it to 3412 * the socket referenced by *map* (of type 3413 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3414 * egress interfaces can be used for redirection. The 3415 * **BPF_F_INGRESS** value in *flags* is used to make the 3416 * distinction (ingress path is selected if the flag is present, 3417 * egress path otherwise). This is the only flag supported for now. 3418 * Return 3419 * **SK_PASS** on success, or **SK_DROP** on error. 3420 * 3421 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3422 * Description 3423 * This helper is used in programs implementing policies at the 3424 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3425 * if the verdict eBPF program returns **SK_PASS**), redirect it 3426 * to the socket referenced by *map* (of type 3427 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3428 * egress interfaces can be used for redirection. The 3429 * **BPF_F_INGRESS** value in *flags* is used to make the 3430 * distinction (ingress path is selected if the flag is present, 3431 * egress otherwise). This is the only flag supported for now. 3432 * Return 3433 * **SK_PASS** on success, or **SK_DROP** on error. 3434 * 3435 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3436 * Description 3437 * Encapsulate the packet associated to *skb* within a Layer 3 3438 * protocol header. This header is provided in the buffer at 3439 * address *hdr*, with *len* its size in bytes. *type* indicates 3440 * the protocol of the header and can be one of: 3441 * 3442 * **BPF_LWT_ENCAP_SEG6** 3443 * IPv6 encapsulation with Segment Routing Header 3444 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3445 * the IPv6 header is computed by the kernel. 3446 * **BPF_LWT_ENCAP_SEG6_INLINE** 3447 * Only works if *skb* contains an IPv6 packet. Insert a 3448 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3449 * the IPv6 header. 3450 * **BPF_LWT_ENCAP_IP** 3451 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3452 * must be IPv4 or IPv6, followed by zero or more 3453 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3454 * total bytes in all prepended headers. Please note that 3455 * if **skb_is_gso**\ (*skb*) is true, no more than two 3456 * headers can be prepended, and the inner header, if 3457 * present, should be either GRE or UDP/GUE. 3458 * 3459 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3460 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3461 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3462 * **BPF_PROG_TYPE_LWT_XMIT**. 3463 * 3464 * A call to this helper is susceptible to change the underlying 3465 * packet buffer. Therefore, at load time, all checks on pointers 3466 * previously done by the verifier are invalidated and must be 3467 * performed again, if the helper is used in combination with 3468 * direct packet access. 3469 * Return 3470 * 0 on success, or a negative error in case of failure. 3471 * 3472 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3473 * Description 3474 * Store *len* bytes from address *from* into the packet 3475 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3476 * inside the outermost IPv6 Segment Routing Header can be 3477 * modified through this helper. 3478 * 3479 * A call to this helper is susceptible to change the underlying 3480 * packet buffer. Therefore, at load time, all checks on pointers 3481 * previously done by the verifier are invalidated and must be 3482 * performed again, if the helper is used in combination with 3483 * direct packet access. 3484 * Return 3485 * 0 on success, or a negative error in case of failure. 3486 * 3487 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3488 * Description 3489 * Adjust the size allocated to TLVs in the outermost IPv6 3490 * Segment Routing Header contained in the packet associated to 3491 * *skb*, at position *offset* by *delta* bytes. Only offsets 3492 * after the segments are accepted. *delta* can be as well 3493 * positive (growing) as negative (shrinking). 3494 * 3495 * A call to this helper is susceptible to change the underlying 3496 * packet buffer. Therefore, at load time, all checks on pointers 3497 * previously done by the verifier are invalidated and must be 3498 * performed again, if the helper is used in combination with 3499 * direct packet access. 3500 * Return 3501 * 0 on success, or a negative error in case of failure. 3502 * 3503 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3504 * Description 3505 * Apply an IPv6 Segment Routing action of type *action* to the 3506 * packet associated to *skb*. Each action takes a parameter 3507 * contained at address *param*, and of length *param_len* bytes. 3508 * *action* can be one of: 3509 * 3510 * **SEG6_LOCAL_ACTION_END_X** 3511 * End.X action: Endpoint with Layer-3 cross-connect. 3512 * Type of *param*: **struct in6_addr**. 3513 * **SEG6_LOCAL_ACTION_END_T** 3514 * End.T action: Endpoint with specific IPv6 table lookup. 3515 * Type of *param*: **int**. 3516 * **SEG6_LOCAL_ACTION_END_B6** 3517 * End.B6 action: Endpoint bound to an SRv6 policy. 3518 * Type of *param*: **struct ipv6_sr_hdr**. 3519 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3520 * End.B6.Encap action: Endpoint bound to an SRv6 3521 * encapsulation policy. 3522 * Type of *param*: **struct ipv6_sr_hdr**. 3523 * 3524 * A call to this helper is susceptible to change the underlying 3525 * packet buffer. Therefore, at load time, all checks on pointers 3526 * previously done by the verifier are invalidated and must be 3527 * performed again, if the helper is used in combination with 3528 * direct packet access. 3529 * Return 3530 * 0 on success, or a negative error in case of failure. 3531 * 3532 * long bpf_rc_repeat(void *ctx) 3533 * Description 3534 * This helper is used in programs implementing IR decoding, to 3535 * report a successfully decoded repeat key message. This delays 3536 * the generation of a key up event for previously generated 3537 * key down event. 3538 * 3539 * Some IR protocols like NEC have a special IR message for 3540 * repeating last button, for when a button is held down. 3541 * 3542 * The *ctx* should point to the lirc sample as passed into 3543 * the program. 3544 * 3545 * This helper is only available is the kernel was compiled with 3546 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3547 * "**y**". 3548 * Return 3549 * 0 3550 * 3551 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3552 * Description 3553 * This helper is used in programs implementing IR decoding, to 3554 * report a successfully decoded key press with *scancode*, 3555 * *toggle* value in the given *protocol*. The scancode will be 3556 * translated to a keycode using the rc keymap, and reported as 3557 * an input key down event. After a period a key up event is 3558 * generated. This period can be extended by calling either 3559 * **bpf_rc_keydown**\ () again with the same values, or calling 3560 * **bpf_rc_repeat**\ (). 3561 * 3562 * Some protocols include a toggle bit, in case the button was 3563 * released and pressed again between consecutive scancodes. 3564 * 3565 * The *ctx* should point to the lirc sample as passed into 3566 * the program. 3567 * 3568 * The *protocol* is the decoded protocol number (see 3569 * **enum rc_proto** for some predefined values). 3570 * 3571 * This helper is only available is the kernel was compiled with 3572 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3573 * "**y**". 3574 * Return 3575 * 0 3576 * 3577 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3578 * Description 3579 * Return the cgroup v2 id of the socket associated with the *skb*. 3580 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3581 * helper for cgroup v1 by providing a tag resp. identifier that 3582 * can be matched on or used for map lookups e.g. to implement 3583 * policy. The cgroup v2 id of a given path in the hierarchy is 3584 * exposed in user space through the f_handle API in order to get 3585 * to the same 64-bit id. 3586 * 3587 * This helper can be used on TC egress path, but not on ingress, 3588 * and is available only if the kernel was compiled with the 3589 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3590 * Return 3591 * The id is returned or 0 in case the id could not be retrieved. 3592 * 3593 * u64 bpf_get_current_cgroup_id(void) 3594 * Description 3595 * Get the current cgroup id based on the cgroup within which 3596 * the current task is running. 3597 * Return 3598 * A 64-bit integer containing the current cgroup id based 3599 * on the cgroup within which the current task is running. 3600 * 3601 * void *bpf_get_local_storage(void *map, u64 flags) 3602 * Description 3603 * Get the pointer to the local storage area. 3604 * The type and the size of the local storage is defined 3605 * by the *map* argument. 3606 * The *flags* meaning is specific for each map type, 3607 * and has to be 0 for cgroup local storage. 3608 * 3609 * Depending on the BPF program type, a local storage area 3610 * can be shared between multiple instances of the BPF program, 3611 * running simultaneously. 3612 * 3613 * A user should care about the synchronization by himself. 3614 * For example, by using the **BPF_ATOMIC** instructions to alter 3615 * the shared data. 3616 * Return 3617 * A pointer to the local storage area. 3618 * 3619 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3620 * Description 3621 * Select a **SO_REUSEPORT** socket from a 3622 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3623 * It checks the selected socket is matching the incoming 3624 * request in the socket buffer. 3625 * Return 3626 * 0 on success, or a negative error in case of failure. 3627 * 3628 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3629 * Description 3630 * Return id of cgroup v2 that is ancestor of cgroup associated 3631 * with the *skb* at the *ancestor_level*. The root cgroup is at 3632 * *ancestor_level* zero and each step down the hierarchy 3633 * increments the level. If *ancestor_level* == level of cgroup 3634 * associated with *skb*, then return value will be same as that 3635 * of **bpf_skb_cgroup_id**\ (). 3636 * 3637 * The helper is useful to implement policies based on cgroups 3638 * that are upper in hierarchy than immediate cgroup associated 3639 * with *skb*. 3640 * 3641 * The format of returned id and helper limitations are same as in 3642 * **bpf_skb_cgroup_id**\ (). 3643 * Return 3644 * The id is returned or 0 in case the id could not be retrieved. 3645 * 3646 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3647 * Description 3648 * Look for TCP socket matching *tuple*, optionally in a child 3649 * network namespace *netns*. The return value must be checked, 3650 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3651 * 3652 * The *ctx* should point to the context of the program, such as 3653 * the skb or socket (depending on the hook in use). This is used 3654 * to determine the base network namespace for the lookup. 3655 * 3656 * *tuple_size* must be one of: 3657 * 3658 * **sizeof**\ (*tuple*\ **->ipv4**) 3659 * Look for an IPv4 socket. 3660 * **sizeof**\ (*tuple*\ **->ipv6**) 3661 * Look for an IPv6 socket. 3662 * 3663 * If the *netns* is a negative signed 32-bit integer, then the 3664 * socket lookup table in the netns associated with the *ctx* 3665 * will be used. For the TC hooks, this is the netns of the device 3666 * in the skb. For socket hooks, this is the netns of the socket. 3667 * If *netns* is any other signed 32-bit value greater than or 3668 * equal to zero then it specifies the ID of the netns relative to 3669 * the netns associated with the *ctx*. *netns* values beyond the 3670 * range of 32-bit integers are reserved for future use. 3671 * 3672 * All values for *flags* are reserved for future usage, and must 3673 * be left at zero. 3674 * 3675 * This helper is available only if the kernel was compiled with 3676 * **CONFIG_NET** configuration option. 3677 * Return 3678 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3679 * For sockets with reuseport option, the **struct bpf_sock** 3680 * result is from *reuse*\ **->socks**\ [] using the hash of the 3681 * tuple. 3682 * 3683 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3684 * Description 3685 * Look for UDP socket matching *tuple*, optionally in a child 3686 * network namespace *netns*. The return value must be checked, 3687 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3688 * 3689 * The *ctx* should point to the context of the program, such as 3690 * the skb or socket (depending on the hook in use). This is used 3691 * to determine the base network namespace for the lookup. 3692 * 3693 * *tuple_size* must be one of: 3694 * 3695 * **sizeof**\ (*tuple*\ **->ipv4**) 3696 * Look for an IPv4 socket. 3697 * **sizeof**\ (*tuple*\ **->ipv6**) 3698 * Look for an IPv6 socket. 3699 * 3700 * If the *netns* is a negative signed 32-bit integer, then the 3701 * socket lookup table in the netns associated with the *ctx* 3702 * will be used. For the TC hooks, this is the netns of the device 3703 * in the skb. For socket hooks, this is the netns of the socket. 3704 * If *netns* is any other signed 32-bit value greater than or 3705 * equal to zero then it specifies the ID of the netns relative to 3706 * the netns associated with the *ctx*. *netns* values beyond the 3707 * range of 32-bit integers are reserved for future use. 3708 * 3709 * All values for *flags* are reserved for future usage, and must 3710 * be left at zero. 3711 * 3712 * This helper is available only if the kernel was compiled with 3713 * **CONFIG_NET** configuration option. 3714 * Return 3715 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3716 * For sockets with reuseport option, the **struct bpf_sock** 3717 * result is from *reuse*\ **->socks**\ [] using the hash of the 3718 * tuple. 3719 * 3720 * long bpf_sk_release(void *sock) 3721 * Description 3722 * Release the reference held by *sock*. *sock* must be a 3723 * non-**NULL** pointer that was returned from 3724 * **bpf_sk_lookup_xxx**\ (). 3725 * Return 3726 * 0 on success, or a negative error in case of failure. 3727 * 3728 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3729 * Description 3730 * Push an element *value* in *map*. *flags* is one of: 3731 * 3732 * **BPF_EXIST** 3733 * If the queue/stack is full, the oldest element is 3734 * removed to make room for this. 3735 * Return 3736 * 0 on success, or a negative error in case of failure. 3737 * 3738 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3739 * Description 3740 * Pop an element from *map*. 3741 * Return 3742 * 0 on success, or a negative error in case of failure. 3743 * 3744 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3745 * Description 3746 * Get an element from *map* without removing it. 3747 * Return 3748 * 0 on success, or a negative error in case of failure. 3749 * 3750 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3751 * Description 3752 * For socket policies, insert *len* bytes into *msg* at offset 3753 * *start*. 3754 * 3755 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3756 * *msg* it may want to insert metadata or options into the *msg*. 3757 * This can later be read and used by any of the lower layer BPF 3758 * hooks. 3759 * 3760 * This helper may fail if under memory pressure (a malloc 3761 * fails) in these cases BPF programs will get an appropriate 3762 * error and BPF programs will need to handle them. 3763 * Return 3764 * 0 on success, or a negative error in case of failure. 3765 * 3766 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3767 * Description 3768 * Will remove *len* bytes from a *msg* starting at byte *start*. 3769 * This may result in **ENOMEM** errors under certain situations if 3770 * an allocation and copy are required due to a full ring buffer. 3771 * However, the helper will try to avoid doing the allocation 3772 * if possible. Other errors can occur if input parameters are 3773 * invalid either due to *start* byte not being valid part of *msg* 3774 * payload and/or *pop* value being to large. 3775 * Return 3776 * 0 on success, or a negative error in case of failure. 3777 * 3778 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3779 * Description 3780 * This helper is used in programs implementing IR decoding, to 3781 * report a successfully decoded pointer movement. 3782 * 3783 * The *ctx* should point to the lirc sample as passed into 3784 * the program. 3785 * 3786 * This helper is only available is the kernel was compiled with 3787 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3788 * "**y**". 3789 * Return 3790 * 0 3791 * 3792 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3793 * Description 3794 * Acquire a spinlock represented by the pointer *lock*, which is 3795 * stored as part of a value of a map. Taking the lock allows to 3796 * safely update the rest of the fields in that value. The 3797 * spinlock can (and must) later be released with a call to 3798 * **bpf_spin_unlock**\ (\ *lock*\ ). 3799 * 3800 * Spinlocks in BPF programs come with a number of restrictions 3801 * and constraints: 3802 * 3803 * * **bpf_spin_lock** objects are only allowed inside maps of 3804 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3805 * list could be extended in the future). 3806 * * BTF description of the map is mandatory. 3807 * * The BPF program can take ONE lock at a time, since taking two 3808 * or more could cause dead locks. 3809 * * Only one **struct bpf_spin_lock** is allowed per map element. 3810 * * When the lock is taken, calls (either BPF to BPF or helpers) 3811 * are not allowed. 3812 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3813 * allowed inside a spinlock-ed region. 3814 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3815 * the lock, on all execution paths, before it returns. 3816 * * The BPF program can access **struct bpf_spin_lock** only via 3817 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3818 * helpers. Loading or storing data into the **struct 3819 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3820 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3821 * of the map value must be a struct and have **struct 3822 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3823 * Nested lock inside another struct is not allowed. 3824 * * The **struct bpf_spin_lock** *lock* field in a map value must 3825 * be aligned on a multiple of 4 bytes in that value. 3826 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3827 * the **bpf_spin_lock** field to user space. 3828 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3829 * a BPF program, do not update the **bpf_spin_lock** field. 3830 * * **bpf_spin_lock** cannot be on the stack or inside a 3831 * networking packet (it can only be inside of a map values). 3832 * * **bpf_spin_lock** is available to root only. 3833 * * Tracing programs and socket filter programs cannot use 3834 * **bpf_spin_lock**\ () due to insufficient preemption checks 3835 * (but this may change in the future). 3836 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3837 * Return 3838 * 0 3839 * 3840 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3841 * Description 3842 * Release the *lock* previously locked by a call to 3843 * **bpf_spin_lock**\ (\ *lock*\ ). 3844 * Return 3845 * 0 3846 * 3847 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3848 * Description 3849 * This helper gets a **struct bpf_sock** pointer such 3850 * that all the fields in this **bpf_sock** can be accessed. 3851 * Return 3852 * A **struct bpf_sock** pointer on success, or **NULL** in 3853 * case of failure. 3854 * 3855 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3856 * Description 3857 * This helper gets a **struct bpf_tcp_sock** pointer from a 3858 * **struct bpf_sock** pointer. 3859 * Return 3860 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3861 * case of failure. 3862 * 3863 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3864 * Description 3865 * Set ECN (Explicit Congestion Notification) field of IP header 3866 * to **CE** (Congestion Encountered) if current value is **ECT** 3867 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3868 * and IPv4. 3869 * Return 3870 * 1 if the **CE** flag is set (either by the current helper call 3871 * or because it was already present), 0 if it is not set. 3872 * 3873 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3874 * Description 3875 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3876 * **bpf_sk_release**\ () is unnecessary and not allowed. 3877 * Return 3878 * A **struct bpf_sock** pointer on success, or **NULL** in 3879 * case of failure. 3880 * 3881 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3882 * Description 3883 * Look for TCP socket matching *tuple*, optionally in a child 3884 * network namespace *netns*. The return value must be checked, 3885 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3886 * 3887 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3888 * that it also returns timewait or request sockets. Use 3889 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3890 * full structure. 3891 * 3892 * This helper is available only if the kernel was compiled with 3893 * **CONFIG_NET** configuration option. 3894 * Return 3895 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3896 * For sockets with reuseport option, the **struct bpf_sock** 3897 * result is from *reuse*\ **->socks**\ [] using the hash of the 3898 * tuple. 3899 * 3900 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3901 * Description 3902 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3903 * the listening socket in *sk*. 3904 * 3905 * *iph* points to the start of the IPv4 or IPv6 header, while 3906 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3907 * **sizeof**\ (**struct ipv6hdr**). 3908 * 3909 * *th* points to the start of the TCP header, while *th_len* 3910 * contains the length of the TCP header (at least 3911 * **sizeof**\ (**struct tcphdr**)). 3912 * Return 3913 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3914 * error otherwise. 3915 * 3916 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3917 * Description 3918 * Get name of sysctl in /proc/sys/ and copy it into provided by 3919 * program buffer *buf* of size *buf_len*. 3920 * 3921 * The buffer is always NUL terminated, unless it's zero-sized. 3922 * 3923 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3924 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3925 * only (e.g. "tcp_mem"). 3926 * Return 3927 * Number of character copied (not including the trailing NUL). 3928 * 3929 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3930 * truncated name in this case). 3931 * 3932 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3933 * Description 3934 * Get current value of sysctl as it is presented in /proc/sys 3935 * (incl. newline, etc), and copy it as a string into provided 3936 * by program buffer *buf* of size *buf_len*. 3937 * 3938 * The whole value is copied, no matter what file position user 3939 * space issued e.g. sys_read at. 3940 * 3941 * The buffer is always NUL terminated, unless it's zero-sized. 3942 * Return 3943 * Number of character copied (not including the trailing NUL). 3944 * 3945 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3946 * truncated name in this case). 3947 * 3948 * **-EINVAL** if current value was unavailable, e.g. because 3949 * sysctl is uninitialized and read returns -EIO for it. 3950 * 3951 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3952 * Description 3953 * Get new value being written by user space to sysctl (before 3954 * the actual write happens) and copy it as a string into 3955 * provided by program buffer *buf* of size *buf_len*. 3956 * 3957 * User space may write new value at file position > 0. 3958 * 3959 * The buffer is always NUL terminated, unless it's zero-sized. 3960 * Return 3961 * Number of character copied (not including the trailing NUL). 3962 * 3963 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3964 * truncated name in this case). 3965 * 3966 * **-EINVAL** if sysctl is being read. 3967 * 3968 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3969 * Description 3970 * Override new value being written by user space to sysctl with 3971 * value provided by program in buffer *buf* of size *buf_len*. 3972 * 3973 * *buf* should contain a string in same form as provided by user 3974 * space on sysctl write. 3975 * 3976 * User space may write new value at file position > 0. To override 3977 * the whole sysctl value file position should be set to zero. 3978 * Return 3979 * 0 on success. 3980 * 3981 * **-E2BIG** if the *buf_len* is too big. 3982 * 3983 * **-EINVAL** if sysctl is being read. 3984 * 3985 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3986 * Description 3987 * Convert the initial part of the string from buffer *buf* of 3988 * size *buf_len* to a long integer according to the given base 3989 * and save the result in *res*. 3990 * 3991 * The string may begin with an arbitrary amount of white space 3992 * (as determined by **isspace**\ (3)) followed by a single 3993 * optional '**-**' sign. 3994 * 3995 * Five least significant bits of *flags* encode base, other bits 3996 * are currently unused. 3997 * 3998 * Base must be either 8, 10, 16 or 0 to detect it automatically 3999 * similar to user space **strtol**\ (3). 4000 * Return 4001 * Number of characters consumed on success. Must be positive but 4002 * no more than *buf_len*. 4003 * 4004 * **-EINVAL** if no valid digits were found or unsupported base 4005 * was provided. 4006 * 4007 * **-ERANGE** if resulting value was out of range. 4008 * 4009 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 4010 * Description 4011 * Convert the initial part of the string from buffer *buf* of 4012 * size *buf_len* to an unsigned long integer according to the 4013 * given base and save the result in *res*. 4014 * 4015 * The string may begin with an arbitrary amount of white space 4016 * (as determined by **isspace**\ (3)). 4017 * 4018 * Five least significant bits of *flags* encode base, other bits 4019 * are currently unused. 4020 * 4021 * Base must be either 8, 10, 16 or 0 to detect it automatically 4022 * similar to user space **strtoul**\ (3). 4023 * Return 4024 * Number of characters consumed on success. Must be positive but 4025 * no more than *buf_len*. 4026 * 4027 * **-EINVAL** if no valid digits were found or unsupported base 4028 * was provided. 4029 * 4030 * **-ERANGE** if resulting value was out of range. 4031 * 4032 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4033 * Description 4034 * Get a bpf-local-storage from a *sk*. 4035 * 4036 * Logically, it could be thought of getting the value from 4037 * a *map* with *sk* as the **key**. From this 4038 * perspective, the usage is not much different from 4039 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4040 * helper enforces the key must be a full socket and the map must 4041 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4042 * 4043 * Underneath, the value is stored locally at *sk* instead of 4044 * the *map*. The *map* is used as the bpf-local-storage 4045 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4046 * searched against all bpf-local-storages residing at *sk*. 4047 * 4048 * *sk* is a kernel **struct sock** pointer for LSM program. 4049 * *sk* is a **struct bpf_sock** pointer for other program types. 4050 * 4051 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4052 * used such that a new bpf-local-storage will be 4053 * created if one does not exist. *value* can be used 4054 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4055 * the initial value of a bpf-local-storage. If *value* is 4056 * **NULL**, the new bpf-local-storage will be zero initialized. 4057 * Return 4058 * A bpf-local-storage pointer is returned on success. 4059 * 4060 * **NULL** if not found or there was an error in adding 4061 * a new bpf-local-storage. 4062 * 4063 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4064 * Description 4065 * Delete a bpf-local-storage from a *sk*. 4066 * Return 4067 * 0 on success. 4068 * 4069 * **-ENOENT** if the bpf-local-storage cannot be found. 4070 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4071 * 4072 * long bpf_send_signal(u32 sig) 4073 * Description 4074 * Send signal *sig* to the process of the current task. 4075 * The signal may be delivered to any of this process's threads. 4076 * Return 4077 * 0 on success or successfully queued. 4078 * 4079 * **-EBUSY** if work queue under nmi is full. 4080 * 4081 * **-EINVAL** if *sig* is invalid. 4082 * 4083 * **-EPERM** if no permission to send the *sig*. 4084 * 4085 * **-EAGAIN** if bpf program can try again. 4086 * 4087 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4088 * Description 4089 * Try to issue a SYN cookie for the packet with corresponding 4090 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4091 * 4092 * *iph* points to the start of the IPv4 or IPv6 header, while 4093 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4094 * **sizeof**\ (**struct ipv6hdr**). 4095 * 4096 * *th* points to the start of the TCP header, while *th_len* 4097 * contains the length of the TCP header with options (at least 4098 * **sizeof**\ (**struct tcphdr**)). 4099 * Return 4100 * On success, lower 32 bits hold the generated SYN cookie in 4101 * followed by 16 bits which hold the MSS value for that cookie, 4102 * and the top 16 bits are unused. 4103 * 4104 * On failure, the returned value is one of the following: 4105 * 4106 * **-EINVAL** SYN cookie cannot be issued due to error 4107 * 4108 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4109 * 4110 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4111 * 4112 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4113 * 4114 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4115 * Description 4116 * Write raw *data* blob into a special BPF perf event held by 4117 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4118 * event must have the following attributes: **PERF_SAMPLE_RAW** 4119 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4120 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4121 * 4122 * The *flags* are used to indicate the index in *map* for which 4123 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4124 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4125 * to indicate that the index of the current CPU core should be 4126 * used. 4127 * 4128 * The value to write, of *size*, is passed through eBPF stack and 4129 * pointed by *data*. 4130 * 4131 * *ctx* is a pointer to in-kernel struct sk_buff. 4132 * 4133 * This helper is similar to **bpf_perf_event_output**\ () but 4134 * restricted to raw_tracepoint bpf programs. 4135 * Return 4136 * 0 on success, or a negative error in case of failure. 4137 * 4138 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4139 * Description 4140 * Safely attempt to read *size* bytes from user space address 4141 * *unsafe_ptr* and store the data in *dst*. 4142 * Return 4143 * 0 on success, or a negative error in case of failure. 4144 * 4145 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4146 * Description 4147 * Safely attempt to read *size* bytes from kernel space address 4148 * *unsafe_ptr* and store the data in *dst*. 4149 * Return 4150 * 0 on success, or a negative error in case of failure. 4151 * 4152 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4153 * Description 4154 * Copy a NUL terminated string from an unsafe user address 4155 * *unsafe_ptr* to *dst*. The *size* should include the 4156 * terminating NUL byte. In case the string length is smaller than 4157 * *size*, the target is not padded with further NUL bytes. If the 4158 * string length is larger than *size*, just *size*-1 bytes are 4159 * copied and the last byte is set to NUL. 4160 * 4161 * On success, returns the number of bytes that were written, 4162 * including the terminal NUL. This makes this helper useful in 4163 * tracing programs for reading strings, and more importantly to 4164 * get its length at runtime. See the following snippet: 4165 * 4166 * :: 4167 * 4168 * SEC("kprobe/sys_open") 4169 * void bpf_sys_open(struct pt_regs *ctx) 4170 * { 4171 * char buf[PATHLEN]; // PATHLEN is defined to 256 4172 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4173 * ctx->di); 4174 * 4175 * // Consume buf, for example push it to 4176 * // userspace via bpf_perf_event_output(); we 4177 * // can use res (the string length) as event 4178 * // size, after checking its boundaries. 4179 * } 4180 * 4181 * In comparison, using **bpf_probe_read_user**\ () helper here 4182 * instead to read the string would require to estimate the length 4183 * at compile time, and would often result in copying more memory 4184 * than necessary. 4185 * 4186 * Another useful use case is when parsing individual process 4187 * arguments or individual environment variables navigating 4188 * *current*\ **->mm->arg_start** and *current*\ 4189 * **->mm->env_start**: using this helper and the return value, 4190 * one can quickly iterate at the right offset of the memory area. 4191 * Return 4192 * On success, the strictly positive length of the output string, 4193 * including the trailing NUL character. On error, a negative 4194 * value. 4195 * 4196 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4197 * Description 4198 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4199 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4200 * Return 4201 * On success, the strictly positive length of the string, including 4202 * the trailing NUL character. On error, a negative value. 4203 * 4204 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4205 * Description 4206 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4207 * *rcv_nxt* is the ack_seq to be sent out. 4208 * Return 4209 * 0 on success, or a negative error in case of failure. 4210 * 4211 * long bpf_send_signal_thread(u32 sig) 4212 * Description 4213 * Send signal *sig* to the thread corresponding to the current task. 4214 * Return 4215 * 0 on success or successfully queued. 4216 * 4217 * **-EBUSY** if work queue under nmi is full. 4218 * 4219 * **-EINVAL** if *sig* is invalid. 4220 * 4221 * **-EPERM** if no permission to send the *sig*. 4222 * 4223 * **-EAGAIN** if bpf program can try again. 4224 * 4225 * u64 bpf_jiffies64(void) 4226 * Description 4227 * Obtain the 64bit jiffies 4228 * Return 4229 * The 64 bit jiffies 4230 * 4231 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4232 * Description 4233 * For an eBPF program attached to a perf event, retrieve the 4234 * branch records (**struct perf_branch_entry**) associated to *ctx* 4235 * and store it in the buffer pointed by *buf* up to size 4236 * *size* bytes. 4237 * Return 4238 * On success, number of bytes written to *buf*. On error, a 4239 * negative value. 4240 * 4241 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4242 * instead return the number of bytes required to store all the 4243 * branch entries. If this flag is set, *buf* may be NULL. 4244 * 4245 * **-EINVAL** if arguments invalid or **size** not a multiple 4246 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4247 * 4248 * **-ENOENT** if architecture does not support branch records. 4249 * 4250 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4251 * Description 4252 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4253 * *namespace* will be returned in *nsdata*. 4254 * Return 4255 * 0 on success, or one of the following in case of failure: 4256 * 4257 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4258 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4259 * 4260 * **-ENOENT** if pidns does not exists for the current task. 4261 * 4262 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4263 * Description 4264 * Write raw *data* blob into a special BPF perf event held by 4265 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4266 * event must have the following attributes: **PERF_SAMPLE_RAW** 4267 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4268 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4269 * 4270 * The *flags* are used to indicate the index in *map* for which 4271 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4272 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4273 * to indicate that the index of the current CPU core should be 4274 * used. 4275 * 4276 * The value to write, of *size*, is passed through eBPF stack and 4277 * pointed by *data*. 4278 * 4279 * *ctx* is a pointer to in-kernel struct xdp_buff. 4280 * 4281 * This helper is similar to **bpf_perf_eventoutput**\ () but 4282 * restricted to raw_tracepoint bpf programs. 4283 * Return 4284 * 0 on success, or a negative error in case of failure. 4285 * 4286 * u64 bpf_get_netns_cookie(void *ctx) 4287 * Description 4288 * Retrieve the cookie (generated by the kernel) of the network 4289 * namespace the input *ctx* is associated with. The network 4290 * namespace cookie remains stable for its lifetime and provides 4291 * a global identifier that can be assumed unique. If *ctx* is 4292 * NULL, then the helper returns the cookie for the initial 4293 * network namespace. The cookie itself is very similar to that 4294 * of **bpf_get_socket_cookie**\ () helper, but for network 4295 * namespaces instead of sockets. 4296 * Return 4297 * A 8-byte long opaque number. 4298 * 4299 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4300 * Description 4301 * Return id of cgroup v2 that is ancestor of the cgroup associated 4302 * with the current task at the *ancestor_level*. The root cgroup 4303 * is at *ancestor_level* zero and each step down the hierarchy 4304 * increments the level. If *ancestor_level* == level of cgroup 4305 * associated with the current task, then return value will be the 4306 * same as that of **bpf_get_current_cgroup_id**\ (). 4307 * 4308 * The helper is useful to implement policies based on cgroups 4309 * that are upper in hierarchy than immediate cgroup associated 4310 * with the current task. 4311 * 4312 * The format of returned id and helper limitations are same as in 4313 * **bpf_get_current_cgroup_id**\ (). 4314 * Return 4315 * The id is returned or 0 in case the id could not be retrieved. 4316 * 4317 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4318 * Description 4319 * Helper is overloaded depending on BPF program type. This 4320 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4321 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4322 * 4323 * Assign the *sk* to the *skb*. When combined with appropriate 4324 * routing configuration to receive the packet towards the socket, 4325 * will cause *skb* to be delivered to the specified socket. 4326 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4327 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4328 * interfere with successful delivery to the socket. 4329 * 4330 * This operation is only valid from TC ingress path. 4331 * 4332 * The *flags* argument must be zero. 4333 * Return 4334 * 0 on success, or a negative error in case of failure: 4335 * 4336 * **-EINVAL** if specified *flags* are not supported. 4337 * 4338 * **-ENOENT** if the socket is unavailable for assignment. 4339 * 4340 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4341 * 4342 * **-EOPNOTSUPP** if the operation is not supported, for example 4343 * a call from outside of TC ingress. 4344 * 4345 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4346 * Description 4347 * Helper is overloaded depending on BPF program type. This 4348 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4349 * 4350 * Select the *sk* as a result of a socket lookup. 4351 * 4352 * For the operation to succeed passed socket must be compatible 4353 * with the packet description provided by the *ctx* object. 4354 * 4355 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4356 * be an exact match. While IP family (**AF_INET** or 4357 * **AF_INET6**) must be compatible, that is IPv6 sockets 4358 * that are not v6-only can be selected for IPv4 packets. 4359 * 4360 * Only TCP listeners and UDP unconnected sockets can be 4361 * selected. *sk* can also be NULL to reset any previous 4362 * selection. 4363 * 4364 * *flags* argument can combination of following values: 4365 * 4366 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4367 * socket selection, potentially done by a BPF program 4368 * that ran before us. 4369 * 4370 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4371 * load-balancing within reuseport group for the socket 4372 * being selected. 4373 * 4374 * On success *ctx->sk* will point to the selected socket. 4375 * 4376 * Return 4377 * 0 on success, or a negative errno in case of failure. 4378 * 4379 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4380 * not compatible with packet family (*ctx->family*). 4381 * 4382 * * **-EEXIST** if socket has been already selected, 4383 * potentially by another program, and 4384 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4385 * 4386 * * **-EINVAL** if unsupported flags were specified. 4387 * 4388 * * **-EPROTOTYPE** if socket L4 protocol 4389 * (*sk->protocol*) doesn't match packet protocol 4390 * (*ctx->protocol*). 4391 * 4392 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4393 * state (TCP listening or UDP unconnected). 4394 * 4395 * u64 bpf_ktime_get_boot_ns(void) 4396 * Description 4397 * Return the time elapsed since system boot, in nanoseconds. 4398 * Does include the time the system was suspended. 4399 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4400 * Return 4401 * Current *ktime*. 4402 * 4403 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4404 * Description 4405 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4406 * out the format string. 4407 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4408 * the format string itself. The *data* and *data_len* are format string 4409 * arguments. The *data* are a **u64** array and corresponding format string 4410 * values are stored in the array. For strings and pointers where pointees 4411 * are accessed, only the pointer values are stored in the *data* array. 4412 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4413 * 4414 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4415 * Reading kernel memory may fail due to either invalid address or 4416 * valid address but requiring a major memory fault. If reading kernel memory 4417 * fails, the string for **%s** will be an empty string, and the ip 4418 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4419 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4420 * Return 4421 * 0 on success, or a negative error in case of failure: 4422 * 4423 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4424 * by returning 1 from bpf program. 4425 * 4426 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4427 * 4428 * **-E2BIG** if *fmt* contains too many format specifiers. 4429 * 4430 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4431 * 4432 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4433 * Description 4434 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4435 * The *m* represents the seq_file. The *data* and *len* represent the 4436 * data to write in bytes. 4437 * Return 4438 * 0 on success, or a negative error in case of failure: 4439 * 4440 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4441 * 4442 * u64 bpf_sk_cgroup_id(void *sk) 4443 * Description 4444 * Return the cgroup v2 id of the socket *sk*. 4445 * 4446 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4447 * returned from **bpf_sk_lookup_xxx**\ (), 4448 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4449 * same as in **bpf_skb_cgroup_id**\ (). 4450 * 4451 * This helper is available only if the kernel was compiled with 4452 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4453 * Return 4454 * The id is returned or 0 in case the id could not be retrieved. 4455 * 4456 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4457 * Description 4458 * Return id of cgroup v2 that is ancestor of cgroup associated 4459 * with the *sk* at the *ancestor_level*. The root cgroup is at 4460 * *ancestor_level* zero and each step down the hierarchy 4461 * increments the level. If *ancestor_level* == level of cgroup 4462 * associated with *sk*, then return value will be same as that 4463 * of **bpf_sk_cgroup_id**\ (). 4464 * 4465 * The helper is useful to implement policies based on cgroups 4466 * that are upper in hierarchy than immediate cgroup associated 4467 * with *sk*. 4468 * 4469 * The format of returned id and helper limitations are same as in 4470 * **bpf_sk_cgroup_id**\ (). 4471 * Return 4472 * The id is returned or 0 in case the id could not be retrieved. 4473 * 4474 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4475 * Description 4476 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4477 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4478 * of new data availability is sent. 4479 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4480 * of new data availability is sent unconditionally. 4481 * If **0** is specified in *flags*, an adaptive notification 4482 * of new data availability is sent. 4483 * 4484 * An adaptive notification is a notification sent whenever the user-space 4485 * process has caught up and consumed all available payloads. In case the user-space 4486 * process is still processing a previous payload, then no notification is needed 4487 * as it will process the newly added payload automatically. 4488 * Return 4489 * 0 on success, or a negative error in case of failure. 4490 * 4491 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4492 * Description 4493 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4494 * *flags* must be 0. 4495 * Return 4496 * Valid pointer with *size* bytes of memory available; NULL, 4497 * otherwise. 4498 * 4499 * void bpf_ringbuf_submit(void *data, u64 flags) 4500 * Description 4501 * Submit reserved ring buffer sample, pointed to by *data*. 4502 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4503 * of new data availability is sent. 4504 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4505 * of new data availability is sent unconditionally. 4506 * If **0** is specified in *flags*, an adaptive notification 4507 * of new data availability is sent. 4508 * 4509 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4510 * Return 4511 * Nothing. Always succeeds. 4512 * 4513 * void bpf_ringbuf_discard(void *data, u64 flags) 4514 * Description 4515 * Discard reserved ring buffer sample, pointed to by *data*. 4516 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4517 * of new data availability is sent. 4518 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4519 * of new data availability is sent unconditionally. 4520 * If **0** is specified in *flags*, an adaptive notification 4521 * of new data availability is sent. 4522 * 4523 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4524 * Return 4525 * Nothing. Always succeeds. 4526 * 4527 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4528 * Description 4529 * Query various characteristics of provided ring buffer. What 4530 * exactly is queries is determined by *flags*: 4531 * 4532 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4533 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4534 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4535 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4536 * 4537 * Data returned is just a momentary snapshot of actual values 4538 * and could be inaccurate, so this facility should be used to 4539 * power heuristics and for reporting, not to make 100% correct 4540 * calculation. 4541 * Return 4542 * Requested value, or 0, if *flags* are not recognized. 4543 * 4544 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4545 * Description 4546 * Change the skbs checksum level by one layer up or down, or 4547 * reset it entirely to none in order to have the stack perform 4548 * checksum validation. The level is applicable to the following 4549 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4550 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4551 * through **bpf_skb_adjust_room**\ () helper with passing in 4552 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4553 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4554 * the UDP header is removed. Similarly, an encap of the latter 4555 * into the former could be accompanied by a helper call to 4556 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4557 * skb is still intended to be processed in higher layers of the 4558 * stack instead of just egressing at tc. 4559 * 4560 * There are three supported level settings at this time: 4561 * 4562 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4563 * with CHECKSUM_UNNECESSARY. 4564 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4565 * with CHECKSUM_UNNECESSARY. 4566 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4567 * sets CHECKSUM_NONE to force checksum validation by the stack. 4568 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4569 * skb->csum_level. 4570 * Return 4571 * 0 on success, or a negative error in case of failure. In the 4572 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4573 * is returned or the error code -EACCES in case the skb is not 4574 * subject to CHECKSUM_UNNECESSARY. 4575 * 4576 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4577 * Description 4578 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4579 * Return 4580 * *sk* if casting is valid, or **NULL** otherwise. 4581 * 4582 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4583 * Description 4584 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4585 * Return 4586 * *sk* if casting is valid, or **NULL** otherwise. 4587 * 4588 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4589 * Description 4590 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4591 * Return 4592 * *sk* if casting is valid, or **NULL** otherwise. 4593 * 4594 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4595 * Description 4596 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4597 * Return 4598 * *sk* if casting is valid, or **NULL** otherwise. 4599 * 4600 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4601 * Description 4602 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4603 * Return 4604 * *sk* if casting is valid, or **NULL** otherwise. 4605 * 4606 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4607 * Description 4608 * Return a user or a kernel stack in bpf program provided buffer. 4609 * Note: the user stack will only be populated if the *task* is 4610 * the current task; all other tasks will return -EOPNOTSUPP. 4611 * To achieve this, the helper needs *task*, which is a valid 4612 * pointer to **struct task_struct**. To store the stacktrace, the 4613 * bpf program provides *buf* with a nonnegative *size*. 4614 * 4615 * The last argument, *flags*, holds the number of stack frames to 4616 * skip (from 0 to 255), masked with 4617 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4618 * the following flags: 4619 * 4620 * **BPF_F_USER_STACK** 4621 * Collect a user space stack instead of a kernel stack. 4622 * The *task* must be the current task. 4623 * **BPF_F_USER_BUILD_ID** 4624 * Collect buildid+offset instead of ips for user stack, 4625 * only valid if **BPF_F_USER_STACK** is also specified. 4626 * 4627 * **bpf_get_task_stack**\ () can collect up to 4628 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4629 * to sufficient large buffer size. Note that 4630 * this limit can be controlled with the **sysctl** program, and 4631 * that it should be manually increased in order to profile long 4632 * user stacks (such as stacks for Java programs). To do so, use: 4633 * 4634 * :: 4635 * 4636 * # sysctl kernel.perf_event_max_stack=<new value> 4637 * Return 4638 * The non-negative copied *buf* length equal to or less than 4639 * *size* on success, or a negative error in case of failure. 4640 * 4641 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4642 * Description 4643 * Load header option. Support reading a particular TCP header 4644 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4645 * 4646 * If *flags* is 0, it will search the option from the 4647 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4648 * has details on what skb_data contains under different 4649 * *skops*\ **->op**. 4650 * 4651 * The first byte of the *searchby_res* specifies the 4652 * kind that it wants to search. 4653 * 4654 * If the searching kind is an experimental kind 4655 * (i.e. 253 or 254 according to RFC6994). It also 4656 * needs to specify the "magic" which is either 4657 * 2 bytes or 4 bytes. It then also needs to 4658 * specify the size of the magic by using 4659 * the 2nd byte which is "kind-length" of a TCP 4660 * header option and the "kind-length" also 4661 * includes the first 2 bytes "kind" and "kind-length" 4662 * itself as a normal TCP header option also does. 4663 * 4664 * For example, to search experimental kind 254 with 4665 * 2 byte magic 0xeB9F, the searchby_res should be 4666 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4667 * 4668 * To search for the standard window scale option (3), 4669 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4670 * Note, kind-length must be 0 for regular option. 4671 * 4672 * Searching for No-Op (0) and End-of-Option-List (1) are 4673 * not supported. 4674 * 4675 * *len* must be at least 2 bytes which is the minimal size 4676 * of a header option. 4677 * 4678 * Supported flags: 4679 * 4680 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4681 * saved_syn packet or the just-received syn packet. 4682 * 4683 * Return 4684 * > 0 when found, the header option is copied to *searchby_res*. 4685 * The return value is the total length copied. On failure, a 4686 * negative error code is returned: 4687 * 4688 * **-EINVAL** if a parameter is invalid. 4689 * 4690 * **-ENOMSG** if the option is not found. 4691 * 4692 * **-ENOENT** if no syn packet is available when 4693 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4694 * 4695 * **-ENOSPC** if there is not enough space. Only *len* number of 4696 * bytes are copied. 4697 * 4698 * **-EFAULT** on failure to parse the header options in the 4699 * packet. 4700 * 4701 * **-EPERM** if the helper cannot be used under the current 4702 * *skops*\ **->op**. 4703 * 4704 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4705 * Description 4706 * Store header option. The data will be copied 4707 * from buffer *from* with length *len* to the TCP header. 4708 * 4709 * The buffer *from* should have the whole option that 4710 * includes the kind, kind-length, and the actual 4711 * option data. The *len* must be at least kind-length 4712 * long. The kind-length does not have to be 4 byte 4713 * aligned. The kernel will take care of the padding 4714 * and setting the 4 bytes aligned value to th->doff. 4715 * 4716 * This helper will check for duplicated option 4717 * by searching the same option in the outgoing skb. 4718 * 4719 * This helper can only be called during 4720 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4721 * 4722 * Return 4723 * 0 on success, or negative error in case of failure: 4724 * 4725 * **-EINVAL** If param is invalid. 4726 * 4727 * **-ENOSPC** if there is not enough space in the header. 4728 * Nothing has been written 4729 * 4730 * **-EEXIST** if the option already exists. 4731 * 4732 * **-EFAULT** on failure to parse the existing header options. 4733 * 4734 * **-EPERM** if the helper cannot be used under the current 4735 * *skops*\ **->op**. 4736 * 4737 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4738 * Description 4739 * Reserve *len* bytes for the bpf header option. The 4740 * space will be used by **bpf_store_hdr_opt**\ () later in 4741 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4742 * 4743 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4744 * the total number of bytes will be reserved. 4745 * 4746 * This helper can only be called during 4747 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4748 * 4749 * Return 4750 * 0 on success, or negative error in case of failure: 4751 * 4752 * **-EINVAL** if a parameter is invalid. 4753 * 4754 * **-ENOSPC** if there is not enough space in the header. 4755 * 4756 * **-EPERM** if the helper cannot be used under the current 4757 * *skops*\ **->op**. 4758 * 4759 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4760 * Description 4761 * Get a bpf_local_storage from an *inode*. 4762 * 4763 * Logically, it could be thought of as getting the value from 4764 * a *map* with *inode* as the **key**. From this 4765 * perspective, the usage is not much different from 4766 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4767 * helper enforces the key must be an inode and the map must also 4768 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4769 * 4770 * Underneath, the value is stored locally at *inode* instead of 4771 * the *map*. The *map* is used as the bpf-local-storage 4772 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4773 * searched against all bpf_local_storage residing at *inode*. 4774 * 4775 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4776 * used such that a new bpf_local_storage will be 4777 * created if one does not exist. *value* can be used 4778 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4779 * the initial value of a bpf_local_storage. If *value* is 4780 * **NULL**, the new bpf_local_storage will be zero initialized. 4781 * Return 4782 * A bpf_local_storage pointer is returned on success. 4783 * 4784 * **NULL** if not found or there was an error in adding 4785 * a new bpf_local_storage. 4786 * 4787 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4788 * Description 4789 * Delete a bpf_local_storage from an *inode*. 4790 * Return 4791 * 0 on success. 4792 * 4793 * **-ENOENT** if the bpf_local_storage cannot be found. 4794 * 4795 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4796 * Description 4797 * Return full path for given **struct path** object, which 4798 * needs to be the kernel BTF *path* object. The path is 4799 * returned in the provided buffer *buf* of size *sz* and 4800 * is zero terminated. 4801 * 4802 * Return 4803 * On success, the strictly positive length of the string, 4804 * including the trailing NUL character. On error, a negative 4805 * value. 4806 * 4807 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4808 * Description 4809 * Read *size* bytes from user space address *user_ptr* and store 4810 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4811 * Return 4812 * 0 on success, or a negative error in case of failure. 4813 * 4814 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4815 * Description 4816 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4817 * using *ptr*->type_id. This value should specify the type 4818 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4819 * can be used to look up vmlinux BTF type ids. Traversing the 4820 * data structure using BTF, the type information and values are 4821 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4822 * the pointer data is carried out to avoid kernel crashes during 4823 * operation. Smaller types can use string space on the stack; 4824 * larger programs can use map data to store the string 4825 * representation. 4826 * 4827 * The string can be subsequently shared with userspace via 4828 * bpf_perf_event_output() or ring buffer interfaces. 4829 * bpf_trace_printk() is to be avoided as it places too small 4830 * a limit on string size to be useful. 4831 * 4832 * *flags* is a combination of 4833 * 4834 * **BTF_F_COMPACT** 4835 * no formatting around type information 4836 * **BTF_F_NONAME** 4837 * no struct/union member names/types 4838 * **BTF_F_PTR_RAW** 4839 * show raw (unobfuscated) pointer values; 4840 * equivalent to printk specifier %px. 4841 * **BTF_F_ZERO** 4842 * show zero-valued struct/union members; they 4843 * are not displayed by default 4844 * 4845 * Return 4846 * The number of bytes that were written (or would have been 4847 * written if output had to be truncated due to string size), 4848 * or a negative error in cases of failure. 4849 * 4850 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4851 * Description 4852 * Use BTF to write to seq_write a string representation of 4853 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4854 * *flags* are identical to those used for bpf_snprintf_btf. 4855 * Return 4856 * 0 on success or a negative error in case of failure. 4857 * 4858 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4859 * Description 4860 * See **bpf_get_cgroup_classid**\ () for the main description. 4861 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4862 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4863 * associated socket instead of the current process. 4864 * Return 4865 * The id is returned or 0 in case the id could not be retrieved. 4866 * 4867 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4868 * Description 4869 * Redirect the packet to another net device of index *ifindex* 4870 * and fill in L2 addresses from neighboring subsystem. This helper 4871 * is somewhat similar to **bpf_redirect**\ (), except that it 4872 * populates L2 addresses as well, meaning, internally, the helper 4873 * relies on the neighbor lookup for the L2 address of the nexthop. 4874 * 4875 * The helper will perform a FIB lookup based on the skb's 4876 * networking header to get the address of the next hop, unless 4877 * this is supplied by the caller in the *params* argument. The 4878 * *plen* argument indicates the len of *params* and should be set 4879 * to 0 if *params* is NULL. 4880 * 4881 * The *flags* argument is reserved and must be 0. The helper is 4882 * currently only supported for tc BPF program types, and enabled 4883 * for IPv4 and IPv6 protocols. 4884 * Return 4885 * The helper returns **TC_ACT_REDIRECT** on success or 4886 * **TC_ACT_SHOT** on error. 4887 * 4888 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4889 * Description 4890 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4891 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4892 * extern variable decorated with '__ksym'. For ksym, there is a 4893 * global var (either static or global) defined of the same name 4894 * in the kernel. The ksym is percpu if the global var is percpu. 4895 * The returned pointer points to the global percpu var on *cpu*. 4896 * 4897 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4898 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4899 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4900 * bpf_per_cpu_ptr() must check the returned value. 4901 * Return 4902 * A pointer pointing to the kernel percpu variable on *cpu*, or 4903 * NULL, if *cpu* is invalid. 4904 * 4905 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4906 * Description 4907 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4908 * pointer to the percpu kernel variable on this cpu. See the 4909 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4910 * 4911 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4912 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4913 * never return NULL. 4914 * Return 4915 * A pointer pointing to the kernel percpu variable on this cpu. 4916 * 4917 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4918 * Description 4919 * Redirect the packet to another net device of index *ifindex*. 4920 * This helper is somewhat similar to **bpf_redirect**\ (), except 4921 * that the redirection happens to the *ifindex*' peer device and 4922 * the netns switch takes place from ingress to ingress without 4923 * going through the CPU's backlog queue. 4924 * 4925 * The *flags* argument is reserved and must be 0. The helper is 4926 * currently only supported for tc BPF program types at the 4927 * ingress hook and for veth and netkit target device types. The 4928 * peer device must reside in a different network namespace. 4929 * Return 4930 * The helper returns **TC_ACT_REDIRECT** on success or 4931 * **TC_ACT_SHOT** on error. 4932 * 4933 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4934 * Description 4935 * Get a bpf_local_storage from the *task*. 4936 * 4937 * Logically, it could be thought of as getting the value from 4938 * a *map* with *task* as the **key**. From this 4939 * perspective, the usage is not much different from 4940 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4941 * helper enforces the key must be a task_struct and the map must also 4942 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4943 * 4944 * Underneath, the value is stored locally at *task* instead of 4945 * the *map*. The *map* is used as the bpf-local-storage 4946 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4947 * searched against all bpf_local_storage residing at *task*. 4948 * 4949 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4950 * used such that a new bpf_local_storage will be 4951 * created if one does not exist. *value* can be used 4952 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4953 * the initial value of a bpf_local_storage. If *value* is 4954 * **NULL**, the new bpf_local_storage will be zero initialized. 4955 * Return 4956 * A bpf_local_storage pointer is returned on success. 4957 * 4958 * **NULL** if not found or there was an error in adding 4959 * a new bpf_local_storage. 4960 * 4961 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4962 * Description 4963 * Delete a bpf_local_storage from a *task*. 4964 * Return 4965 * 0 on success. 4966 * 4967 * **-ENOENT** if the bpf_local_storage cannot be found. 4968 * 4969 * struct task_struct *bpf_get_current_task_btf(void) 4970 * Description 4971 * Return a BTF pointer to the "current" task. 4972 * This pointer can also be used in helpers that accept an 4973 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4974 * Return 4975 * Pointer to the current task. 4976 * 4977 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4978 * Description 4979 * Set or clear certain options on *bprm*: 4980 * 4981 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4982 * which sets the **AT_SECURE** auxv for glibc. The bit 4983 * is cleared if the flag is not specified. 4984 * Return 4985 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4986 * 4987 * u64 bpf_ktime_get_coarse_ns(void) 4988 * Description 4989 * Return a coarse-grained version of the time elapsed since 4990 * system boot, in nanoseconds. Does not include time the system 4991 * was suspended. 4992 * 4993 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4994 * Return 4995 * Current *ktime*. 4996 * 4997 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4998 * Description 4999 * Returns the stored IMA hash of the *inode* (if it's available). 5000 * If the hash is larger than *size*, then only *size* 5001 * bytes will be copied to *dst* 5002 * Return 5003 * The **hash_algo** is returned on success, 5004 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 5005 * invalid arguments are passed. 5006 * 5007 * struct socket *bpf_sock_from_file(struct file *file) 5008 * Description 5009 * If the given file represents a socket, returns the associated 5010 * socket. 5011 * Return 5012 * A pointer to a struct socket on success or NULL if the file is 5013 * not a socket. 5014 * 5015 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 5016 * Description 5017 * Check packet size against exceeding MTU of net device (based 5018 * on *ifindex*). This helper will likely be used in combination 5019 * with helpers that adjust/change the packet size. 5020 * 5021 * The argument *len_diff* can be used for querying with a planned 5022 * size change. This allows to check MTU prior to changing packet 5023 * ctx. Providing a *len_diff* adjustment that is larger than the 5024 * actual packet size (resulting in negative packet size) will in 5025 * principle not exceed the MTU, which is why it is not considered 5026 * a failure. Other BPF helpers are needed for performing the 5027 * planned size change; therefore the responsibility for catching 5028 * a negative packet size belongs in those helpers. 5029 * 5030 * Specifying *ifindex* zero means the MTU check is performed 5031 * against the current net device. This is practical if this isn't 5032 * used prior to redirect. 5033 * 5034 * On input *mtu_len* must be a valid pointer, else verifier will 5035 * reject BPF program. If the value *mtu_len* is initialized to 5036 * zero then the ctx packet size is use. When value *mtu_len* is 5037 * provided as input this specify the L3 length that the MTU check 5038 * is done against. Remember XDP and TC length operate at L2, but 5039 * this value is L3 as this correlate to MTU and IP-header tot_len 5040 * values which are L3 (similar behavior as bpf_fib_lookup). 5041 * 5042 * The Linux kernel route table can configure MTUs on a more 5043 * specific per route level, which is not provided by this helper. 5044 * For route level MTU checks use the **bpf_fib_lookup**\ () 5045 * helper. 5046 * 5047 * *ctx* is either **struct xdp_md** for XDP programs or 5048 * **struct sk_buff** for tc cls_act programs. 5049 * 5050 * The *flags* argument can be a combination of one or more of the 5051 * following values: 5052 * 5053 * **BPF_MTU_CHK_SEGS** 5054 * This flag will only works for *ctx* **struct sk_buff**. 5055 * If packet context contains extra packet segment buffers 5056 * (often knows as GSO skb), then MTU check is harder to 5057 * check at this point, because in transmit path it is 5058 * possible for the skb packet to get re-segmented 5059 * (depending on net device features). This could still be 5060 * a MTU violation, so this flag enables performing MTU 5061 * check against segments, with a different violation 5062 * return code to tell it apart. Check cannot use len_diff. 5063 * 5064 * On return *mtu_len* pointer contains the MTU value of the net 5065 * device. Remember the net device configured MTU is the L3 size, 5066 * which is returned here and XDP and TC length operate at L2. 5067 * Helper take this into account for you, but remember when using 5068 * MTU value in your BPF-code. 5069 * 5070 * Return 5071 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5072 * 5073 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5074 * 5075 * MTU violations return positive values, but also populate MTU 5076 * value in *mtu_len* pointer, as this can be needed for 5077 * implementing PMTU handing: 5078 * 5079 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5080 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5081 * 5082 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5083 * Description 5084 * For each element in **map**, call **callback_fn** function with 5085 * **map**, **callback_ctx** and other map-specific parameters. 5086 * The **callback_fn** should be a static function and 5087 * the **callback_ctx** should be a pointer to the stack. 5088 * The **flags** is used to control certain aspects of the helper. 5089 * Currently, the **flags** must be 0. 5090 * 5091 * The following are a list of supported map types and their 5092 * respective expected callback signatures: 5093 * 5094 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5095 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5096 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5097 * 5098 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5099 * 5100 * For per_cpu maps, the map_value is the value on the cpu where the 5101 * bpf_prog is running. 5102 * 5103 * If **callback_fn** return 0, the helper will continue to the next 5104 * element. If return value is 1, the helper will skip the rest of 5105 * elements and return. Other return values are not used now. 5106 * 5107 * Return 5108 * The number of traversed map elements for success, **-EINVAL** for 5109 * invalid **flags**. 5110 * 5111 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5112 * Description 5113 * Outputs a string into the **str** buffer of size **str_size** 5114 * based on a format string stored in a read-only map pointed by 5115 * **fmt**. 5116 * 5117 * Each format specifier in **fmt** corresponds to one u64 element 5118 * in the **data** array. For strings and pointers where pointees 5119 * are accessed, only the pointer values are stored in the *data* 5120 * array. The *data_len* is the size of *data* in bytes - must be 5121 * a multiple of 8. 5122 * 5123 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5124 * memory. Reading kernel memory may fail due to either invalid 5125 * address or valid address but requiring a major memory fault. If 5126 * reading kernel memory fails, the string for **%s** will be an 5127 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5128 * Not returning error to bpf program is consistent with what 5129 * **bpf_trace_printk**\ () does for now. 5130 * 5131 * Return 5132 * The strictly positive length of the formatted string, including 5133 * the trailing zero character. If the return value is greater than 5134 * **str_size**, **str** contains a truncated string, guaranteed to 5135 * be zero-terminated except when **str_size** is 0. 5136 * 5137 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5138 * 5139 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5140 * Description 5141 * Execute bpf syscall with given arguments. 5142 * Return 5143 * A syscall result. 5144 * 5145 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5146 * Description 5147 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5148 * Return 5149 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5150 * 5151 * long bpf_sys_close(u32 fd) 5152 * Description 5153 * Execute close syscall for given FD. 5154 * Return 5155 * A syscall result. 5156 * 5157 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5158 * Description 5159 * Initialize the timer. 5160 * First 4 bits of *flags* specify clockid. 5161 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5162 * All other bits of *flags* are reserved. 5163 * The verifier will reject the program if *timer* is not from 5164 * the same *map*. 5165 * Return 5166 * 0 on success. 5167 * **-EBUSY** if *timer* is already initialized. 5168 * **-EINVAL** if invalid *flags* are passed. 5169 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5170 * The user space should either hold a file descriptor to a map with timers 5171 * or pin such map in bpffs. When map is unpinned or file descriptor is 5172 * closed all timers in the map will be cancelled and freed. 5173 * 5174 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5175 * Description 5176 * Configure the timer to call *callback_fn* static function. 5177 * Return 5178 * 0 on success. 5179 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5180 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5181 * The user space should either hold a file descriptor to a map with timers 5182 * or pin such map in bpffs. When map is unpinned or file descriptor is 5183 * closed all timers in the map will be cancelled and freed. 5184 * 5185 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5186 * Description 5187 * Set timer expiration N nanoseconds from the current time. The 5188 * configured callback will be invoked in soft irq context on some cpu 5189 * and will not repeat unless another bpf_timer_start() is made. 5190 * In such case the next invocation can migrate to a different cpu. 5191 * Since struct bpf_timer is a field inside map element the map 5192 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5193 * of BPF program to make sure that callback_fn code stays valid. 5194 * When user space reference to a map reaches zero all timers 5195 * in a map are cancelled and corresponding program's refcnts are 5196 * decremented. This is done to make sure that Ctrl-C of a user 5197 * process doesn't leave any timers running. If map is pinned in 5198 * bpffs the callback_fn can re-arm itself indefinitely. 5199 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5200 * cancel and free the timer in the given map element. 5201 * The map can contain timers that invoke callback_fn-s from different 5202 * programs. The same callback_fn can serve different timers from 5203 * different maps if key/value layout matches across maps. 5204 * Every bpf_timer_set_callback() can have different callback_fn. 5205 * 5206 * *flags* can be one of: 5207 * 5208 * **BPF_F_TIMER_ABS** 5209 * Start the timer in absolute expire value instead of the 5210 * default relative one. 5211 * **BPF_F_TIMER_CPU_PIN** 5212 * Timer will be pinned to the CPU of the caller. 5213 * 5214 * Return 5215 * 0 on success. 5216 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5217 * or invalid *flags* are passed. 5218 * 5219 * long bpf_timer_cancel(struct bpf_timer *timer) 5220 * Description 5221 * Cancel the timer and wait for callback_fn to finish if it was running. 5222 * Return 5223 * 0 if the timer was not active. 5224 * 1 if the timer was active. 5225 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5226 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5227 * own timer which would have led to a deadlock otherwise. 5228 * 5229 * u64 bpf_get_func_ip(void *ctx) 5230 * Description 5231 * Get address of the traced function (for tracing and kprobe programs). 5232 * 5233 * When called for kprobe program attached as uprobe it returns 5234 * probe address for both entry and return uprobe. 5235 * 5236 * Return 5237 * Address of the traced function for kprobe. 5238 * 0 for kprobes placed within the function (not at the entry). 5239 * Address of the probe for uprobe and return uprobe. 5240 * 5241 * u64 bpf_get_attach_cookie(void *ctx) 5242 * Description 5243 * Get bpf_cookie value provided (optionally) during the program 5244 * attachment. It might be different for each individual 5245 * attachment, even if BPF program itself is the same. 5246 * Expects BPF program context *ctx* as a first argument. 5247 * 5248 * Supported for the following program types: 5249 * - kprobe/uprobe; 5250 * - tracepoint; 5251 * - perf_event. 5252 * Return 5253 * Value specified by user at BPF link creation/attachment time 5254 * or 0, if it was not specified. 5255 * 5256 * long bpf_task_pt_regs(struct task_struct *task) 5257 * Description 5258 * Get the struct pt_regs associated with **task**. 5259 * Return 5260 * A pointer to struct pt_regs. 5261 * 5262 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5263 * Description 5264 * Get branch trace from hardware engines like Intel LBR. The 5265 * hardware engine is stopped shortly after the helper is 5266 * called. Therefore, the user need to filter branch entries 5267 * based on the actual use case. To capture branch trace 5268 * before the trigger point of the BPF program, the helper 5269 * should be called at the beginning of the BPF program. 5270 * 5271 * The data is stored as struct perf_branch_entry into output 5272 * buffer *entries*. *size* is the size of *entries* in bytes. 5273 * *flags* is reserved for now and must be zero. 5274 * 5275 * Return 5276 * On success, number of bytes written to *buf*. On error, a 5277 * negative value. 5278 * 5279 * **-EINVAL** if *flags* is not zero. 5280 * 5281 * **-ENOENT** if architecture does not support branch records. 5282 * 5283 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5284 * Description 5285 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5286 * to format and can handle more format args as a result. 5287 * 5288 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5289 * Return 5290 * The number of bytes written to the buffer, or a negative error 5291 * in case of failure. 5292 * 5293 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5294 * Description 5295 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5296 * Return 5297 * *sk* if casting is valid, or **NULL** otherwise. 5298 * 5299 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5300 * Description 5301 * Get the address of a kernel symbol, returned in *res*. *res* is 5302 * set to 0 if the symbol is not found. 5303 * Return 5304 * On success, zero. On error, a negative value. 5305 * 5306 * **-EINVAL** if *flags* is not zero. 5307 * 5308 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5309 * 5310 * **-ENOENT** if symbol is not found. 5311 * 5312 * **-EPERM** if caller does not have permission to obtain kernel address. 5313 * 5314 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5315 * Description 5316 * Find vma of *task* that contains *addr*, call *callback_fn* 5317 * function with *task*, *vma*, and *callback_ctx*. 5318 * The *callback_fn* should be a static function and 5319 * the *callback_ctx* should be a pointer to the stack. 5320 * The *flags* is used to control certain aspects of the helper. 5321 * Currently, the *flags* must be 0. 5322 * 5323 * The expected callback signature is 5324 * 5325 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5326 * 5327 * Return 5328 * 0 on success. 5329 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5330 * **-EBUSY** if failed to try lock mmap_lock. 5331 * **-EINVAL** for invalid **flags**. 5332 * 5333 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5334 * Description 5335 * For **nr_loops**, call **callback_fn** function 5336 * with **callback_ctx** as the context parameter. 5337 * The **callback_fn** should be a static function and 5338 * the **callback_ctx** should be a pointer to the stack. 5339 * The **flags** is used to control certain aspects of the helper. 5340 * Currently, the **flags** must be 0. Currently, nr_loops is 5341 * limited to 1 << 23 (~8 million) loops. 5342 * 5343 * long (\*callback_fn)(u32 index, void \*ctx); 5344 * 5345 * where **index** is the current index in the loop. The index 5346 * is zero-indexed. 5347 * 5348 * If **callback_fn** returns 0, the helper will continue to the next 5349 * loop. If return value is 1, the helper will skip the rest of 5350 * the loops and return. Other return values are not used now, 5351 * and will be rejected by the verifier. 5352 * 5353 * Return 5354 * The number of loops performed, **-EINVAL** for invalid **flags**, 5355 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5356 * 5357 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5358 * Description 5359 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5360 * to be null-terminated and **s1_sz** is the maximum storage 5361 * size of **s1**. **s2** must be a read-only string. 5362 * Return 5363 * An integer less than, equal to, or greater than zero 5364 * if the first **s1_sz** bytes of **s1** is found to be 5365 * less than, to match, or be greater than **s2**. 5366 * 5367 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5368 * Description 5369 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5370 * returned in **value**. 5371 * 5372 * Return 5373 * 0 on success. 5374 * **-EINVAL** if n >= argument register count of traced function. 5375 * 5376 * long bpf_get_func_ret(void *ctx, u64 *value) 5377 * Description 5378 * Get return value of the traced function (for tracing programs) 5379 * in **value**. 5380 * 5381 * Return 5382 * 0 on success. 5383 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5384 * 5385 * long bpf_get_func_arg_cnt(void *ctx) 5386 * Description 5387 * Get number of registers of the traced function (for tracing programs) where 5388 * function arguments are stored in these registers. 5389 * 5390 * Return 5391 * The number of argument registers of the traced function. 5392 * 5393 * int bpf_get_retval(void) 5394 * Description 5395 * Get the BPF program's return value that will be returned to the upper layers. 5396 * 5397 * This helper is currently supported by cgroup programs and only by the hooks 5398 * where BPF program's return value is returned to the userspace via errno. 5399 * Return 5400 * The BPF program's return value. 5401 * 5402 * int bpf_set_retval(int retval) 5403 * Description 5404 * Set the BPF program's return value that will be returned to the upper layers. 5405 * 5406 * This helper is currently supported by cgroup programs and only by the hooks 5407 * where BPF program's return value is returned to the userspace via errno. 5408 * 5409 * Note that there is the following corner case where the program exports an error 5410 * via bpf_set_retval but signals success via 'return 1': 5411 * 5412 * bpf_set_retval(-EPERM); 5413 * return 1; 5414 * 5415 * In this case, the BPF program's return value will use helper's -EPERM. This 5416 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5417 * 5418 * Return 5419 * 0 on success, or a negative error in case of failure. 5420 * 5421 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5422 * Description 5423 * Get the total size of a given xdp buff (linear and paged area) 5424 * Return 5425 * The total size of a given xdp buffer. 5426 * 5427 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5428 * Description 5429 * This helper is provided as an easy way to load data from a 5430 * xdp buffer. It can be used to load *len* bytes from *offset* from 5431 * the frame associated to *xdp_md*, into the buffer pointed by 5432 * *buf*. 5433 * Return 5434 * 0 on success, or a negative error in case of failure. 5435 * 5436 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5437 * Description 5438 * Store *len* bytes from buffer *buf* into the frame 5439 * associated to *xdp_md*, at *offset*. 5440 * Return 5441 * 0 on success, or a negative error in case of failure. 5442 * 5443 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5444 * Description 5445 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5446 * address space, and stores the data in *dst*. *flags* is not 5447 * used yet and is provided for future extensibility. This helper 5448 * can only be used by sleepable programs. 5449 * Return 5450 * 0 on success, or a negative error in case of failure. On error 5451 * *dst* buffer is zeroed out. 5452 * 5453 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5454 * Description 5455 * Change the __sk_buff->tstamp_type to *tstamp_type* 5456 * and set *tstamp* to the __sk_buff->tstamp together. 5457 * 5458 * If there is no need to change the __sk_buff->tstamp_type, 5459 * the tstamp value can be directly written to __sk_buff->tstamp 5460 * instead. 5461 * 5462 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5463 * will be kept during bpf_redirect_*(). A non zero 5464 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5465 * *tstamp_type*. 5466 * 5467 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5468 * with a zero *tstamp*. 5469 * 5470 * Only IPv4 and IPv6 skb->protocol are supported. 5471 * 5472 * This function is most useful when it needs to set a 5473 * mono delivery time to __sk_buff->tstamp and then 5474 * bpf_redirect_*() to the egress of an iface. For example, 5475 * changing the (rcv) timestamp in __sk_buff->tstamp at 5476 * ingress to a mono delivery time and then bpf_redirect_*() 5477 * to sch_fq@phy-dev. 5478 * Return 5479 * 0 on success. 5480 * **-EINVAL** for invalid input 5481 * **-EOPNOTSUPP** for unsupported protocol 5482 * 5483 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5484 * Description 5485 * Returns a calculated IMA hash of the *file*. 5486 * If the hash is larger than *size*, then only *size* 5487 * bytes will be copied to *dst* 5488 * Return 5489 * The **hash_algo** is returned on success, 5490 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5491 * invalid arguments are passed. 5492 * 5493 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5494 * Description 5495 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5496 * old value. *ptr* can be NULL, otherwise it must be a referenced 5497 * pointer which will be released when this helper is called. 5498 * Return 5499 * The old value of kptr (which can be NULL). The returned pointer 5500 * if not NULL, is a reference which must be released using its 5501 * corresponding release function, or moved into a BPF map before 5502 * program exit. 5503 * 5504 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5505 * Description 5506 * Perform a lookup in *percpu map* for an entry associated to 5507 * *key* on *cpu*. 5508 * Return 5509 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5510 * was found or *cpu* is invalid. 5511 * 5512 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5513 * Description 5514 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5515 * Return 5516 * *sk* if casting is valid, or **NULL** otherwise. 5517 * 5518 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5519 * Description 5520 * Get a dynptr to local memory *data*. 5521 * 5522 * *data* must be a ptr to a map value. 5523 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5524 * *flags* is currently unused. 5525 * Return 5526 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5527 * -EINVAL if flags is not 0. 5528 * 5529 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5530 * Description 5531 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5532 * through the dynptr interface. *flags* must be 0. 5533 * 5534 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5535 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5536 * reservation fails. This is enforced by the verifier. 5537 * Return 5538 * 0 on success, or a negative error in case of failure. 5539 * 5540 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5541 * Description 5542 * Submit reserved ring buffer sample, pointed to by *data*, 5543 * through the dynptr interface. This is a no-op if the dynptr is 5544 * invalid/null. 5545 * 5546 * For more information on *flags*, please see 5547 * 'bpf_ringbuf_submit'. 5548 * Return 5549 * Nothing. Always succeeds. 5550 * 5551 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5552 * Description 5553 * Discard reserved ring buffer sample through the dynptr 5554 * interface. This is a no-op if the dynptr is invalid/null. 5555 * 5556 * For more information on *flags*, please see 5557 * 'bpf_ringbuf_discard'. 5558 * Return 5559 * Nothing. Always succeeds. 5560 * 5561 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5562 * Description 5563 * Read *len* bytes from *src* into *dst*, starting from *offset* 5564 * into *src*. 5565 * *flags* is currently unused. 5566 * Return 5567 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5568 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5569 * *flags* is not 0. 5570 * 5571 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5572 * Description 5573 * Write *len* bytes from *src* into *dst*, starting from *offset* 5574 * into *dst*. 5575 * 5576 * *flags* must be 0 except for skb-type dynptrs. 5577 * 5578 * For skb-type dynptrs: 5579 * * All data slices of the dynptr are automatically 5580 * invalidated after **bpf_dynptr_write**\ (). This is 5581 * because writing may pull the skb and change the 5582 * underlying packet buffer. 5583 * 5584 * * For *flags*, please see the flags accepted by 5585 * **bpf_skb_store_bytes**\ (). 5586 * Return 5587 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5588 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5589 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5590 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5591 * 5592 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5593 * Description 5594 * Get a pointer to the underlying dynptr data. 5595 * 5596 * *len* must be a statically known value. The returned data slice 5597 * is invalidated whenever the dynptr is invalidated. 5598 * 5599 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5600 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5601 * Return 5602 * Pointer to the underlying dynptr data, NULL if the dynptr is 5603 * read-only, if the dynptr is invalid, or if the offset and length 5604 * is out of bounds. 5605 * 5606 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5607 * Description 5608 * Try to issue a SYN cookie for the packet with corresponding 5609 * IPv4/TCP headers, *iph* and *th*, without depending on a 5610 * listening socket. 5611 * 5612 * *iph* points to the IPv4 header. 5613 * 5614 * *th* points to the start of the TCP header, while *th_len* 5615 * contains the length of the TCP header (at least 5616 * **sizeof**\ (**struct tcphdr**)). 5617 * Return 5618 * On success, lower 32 bits hold the generated SYN cookie in 5619 * followed by 16 bits which hold the MSS value for that cookie, 5620 * and the top 16 bits are unused. 5621 * 5622 * On failure, the returned value is one of the following: 5623 * 5624 * **-EINVAL** if *th_len* is invalid. 5625 * 5626 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5627 * Description 5628 * Try to issue a SYN cookie for the packet with corresponding 5629 * IPv6/TCP headers, *iph* and *th*, without depending on a 5630 * listening socket. 5631 * 5632 * *iph* points to the IPv6 header. 5633 * 5634 * *th* points to the start of the TCP header, while *th_len* 5635 * contains the length of the TCP header (at least 5636 * **sizeof**\ (**struct tcphdr**)). 5637 * Return 5638 * On success, lower 32 bits hold the generated SYN cookie in 5639 * followed by 16 bits which hold the MSS value for that cookie, 5640 * and the top 16 bits are unused. 5641 * 5642 * On failure, the returned value is one of the following: 5643 * 5644 * **-EINVAL** if *th_len* is invalid. 5645 * 5646 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5647 * 5648 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5649 * Description 5650 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5651 * without depending on a listening socket. 5652 * 5653 * *iph* points to the IPv4 header. 5654 * 5655 * *th* points to the TCP header. 5656 * Return 5657 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5658 * 5659 * On failure, the returned value is one of the following: 5660 * 5661 * **-EACCES** if the SYN cookie is not valid. 5662 * 5663 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5664 * Description 5665 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5666 * without depending on a listening socket. 5667 * 5668 * *iph* points to the IPv6 header. 5669 * 5670 * *th* points to the TCP header. 5671 * Return 5672 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5673 * 5674 * On failure, the returned value is one of the following: 5675 * 5676 * **-EACCES** if the SYN cookie is not valid. 5677 * 5678 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5679 * 5680 * u64 bpf_ktime_get_tai_ns(void) 5681 * Description 5682 * A nonsettable system-wide clock derived from wall-clock time but 5683 * ignoring leap seconds. This clock does not experience 5684 * discontinuities and backwards jumps caused by NTP inserting leap 5685 * seconds as CLOCK_REALTIME does. 5686 * 5687 * See: **clock_gettime**\ (**CLOCK_TAI**) 5688 * Return 5689 * Current *ktime*. 5690 * 5691 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5692 * Description 5693 * Drain samples from the specified user ring buffer, and invoke 5694 * the provided callback for each such sample: 5695 * 5696 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5697 * 5698 * If **callback_fn** returns 0, the helper will continue to try 5699 * and drain the next sample, up to a maximum of 5700 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5701 * the helper will skip the rest of the samples and return. Other 5702 * return values are not used now, and will be rejected by the 5703 * verifier. 5704 * Return 5705 * The number of drained samples if no error was encountered while 5706 * draining samples, or 0 if no samples were present in the ring 5707 * buffer. If a user-space producer was epoll-waiting on this map, 5708 * and at least one sample was drained, they will receive an event 5709 * notification notifying them of available space in the ring 5710 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5711 * function, no wakeup notification will be sent. If the 5712 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5713 * be sent even if no sample was drained. 5714 * 5715 * On failure, the returned value is one of the following: 5716 * 5717 * **-EBUSY** if the ring buffer is contended, and another calling 5718 * context was concurrently draining the ring buffer. 5719 * 5720 * **-EINVAL** if user-space is not properly tracking the ring 5721 * buffer due to the producer position not being aligned to 8 5722 * bytes, a sample not being aligned to 8 bytes, or the producer 5723 * position not matching the advertised length of a sample. 5724 * 5725 * **-E2BIG** if user-space has tried to publish a sample which is 5726 * larger than the size of the ring buffer, or which cannot fit 5727 * within a struct bpf_dynptr. 5728 * 5729 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5730 * Description 5731 * Get a bpf_local_storage from the *cgroup*. 5732 * 5733 * Logically, it could be thought of as getting the value from 5734 * a *map* with *cgroup* as the **key**. From this 5735 * perspective, the usage is not much different from 5736 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5737 * helper enforces the key must be a cgroup struct and the map must also 5738 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5739 * 5740 * In reality, the local-storage value is embedded directly inside of the 5741 * *cgroup* object itself, rather than being located in the 5742 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5743 * queried for some *map* on a *cgroup* object, the kernel will perform an 5744 * O(n) iteration over all of the live local-storage values for that 5745 * *cgroup* object until the local-storage value for the *map* is found. 5746 * 5747 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5748 * used such that a new bpf_local_storage will be 5749 * created if one does not exist. *value* can be used 5750 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5751 * the initial value of a bpf_local_storage. If *value* is 5752 * **NULL**, the new bpf_local_storage will be zero initialized. 5753 * Return 5754 * A bpf_local_storage pointer is returned on success. 5755 * 5756 * **NULL** if not found or there was an error in adding 5757 * a new bpf_local_storage. 5758 * 5759 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5760 * Description 5761 * Delete a bpf_local_storage from a *cgroup*. 5762 * Return 5763 * 0 on success. 5764 * 5765 * **-ENOENT** if the bpf_local_storage cannot be found. 5766 */ 5767 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5768 FN(unspec, 0, ##ctx) \ 5769 FN(map_lookup_elem, 1, ##ctx) \ 5770 FN(map_update_elem, 2, ##ctx) \ 5771 FN(map_delete_elem, 3, ##ctx) \ 5772 FN(probe_read, 4, ##ctx) \ 5773 FN(ktime_get_ns, 5, ##ctx) \ 5774 FN(trace_printk, 6, ##ctx) \ 5775 FN(get_prandom_u32, 7, ##ctx) \ 5776 FN(get_smp_processor_id, 8, ##ctx) \ 5777 FN(skb_store_bytes, 9, ##ctx) \ 5778 FN(l3_csum_replace, 10, ##ctx) \ 5779 FN(l4_csum_replace, 11, ##ctx) \ 5780 FN(tail_call, 12, ##ctx) \ 5781 FN(clone_redirect, 13, ##ctx) \ 5782 FN(get_current_pid_tgid, 14, ##ctx) \ 5783 FN(get_current_uid_gid, 15, ##ctx) \ 5784 FN(get_current_comm, 16, ##ctx) \ 5785 FN(get_cgroup_classid, 17, ##ctx) \ 5786 FN(skb_vlan_push, 18, ##ctx) \ 5787 FN(skb_vlan_pop, 19, ##ctx) \ 5788 FN(skb_get_tunnel_key, 20, ##ctx) \ 5789 FN(skb_set_tunnel_key, 21, ##ctx) \ 5790 FN(perf_event_read, 22, ##ctx) \ 5791 FN(redirect, 23, ##ctx) \ 5792 FN(get_route_realm, 24, ##ctx) \ 5793 FN(perf_event_output, 25, ##ctx) \ 5794 FN(skb_load_bytes, 26, ##ctx) \ 5795 FN(get_stackid, 27, ##ctx) \ 5796 FN(csum_diff, 28, ##ctx) \ 5797 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5798 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5799 FN(skb_change_proto, 31, ##ctx) \ 5800 FN(skb_change_type, 32, ##ctx) \ 5801 FN(skb_under_cgroup, 33, ##ctx) \ 5802 FN(get_hash_recalc, 34, ##ctx) \ 5803 FN(get_current_task, 35, ##ctx) \ 5804 FN(probe_write_user, 36, ##ctx) \ 5805 FN(current_task_under_cgroup, 37, ##ctx) \ 5806 FN(skb_change_tail, 38, ##ctx) \ 5807 FN(skb_pull_data, 39, ##ctx) \ 5808 FN(csum_update, 40, ##ctx) \ 5809 FN(set_hash_invalid, 41, ##ctx) \ 5810 FN(get_numa_node_id, 42, ##ctx) \ 5811 FN(skb_change_head, 43, ##ctx) \ 5812 FN(xdp_adjust_head, 44, ##ctx) \ 5813 FN(probe_read_str, 45, ##ctx) \ 5814 FN(get_socket_cookie, 46, ##ctx) \ 5815 FN(get_socket_uid, 47, ##ctx) \ 5816 FN(set_hash, 48, ##ctx) \ 5817 FN(setsockopt, 49, ##ctx) \ 5818 FN(skb_adjust_room, 50, ##ctx) \ 5819 FN(redirect_map, 51, ##ctx) \ 5820 FN(sk_redirect_map, 52, ##ctx) \ 5821 FN(sock_map_update, 53, ##ctx) \ 5822 FN(xdp_adjust_meta, 54, ##ctx) \ 5823 FN(perf_event_read_value, 55, ##ctx) \ 5824 FN(perf_prog_read_value, 56, ##ctx) \ 5825 FN(getsockopt, 57, ##ctx) \ 5826 FN(override_return, 58, ##ctx) \ 5827 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5828 FN(msg_redirect_map, 60, ##ctx) \ 5829 FN(msg_apply_bytes, 61, ##ctx) \ 5830 FN(msg_cork_bytes, 62, ##ctx) \ 5831 FN(msg_pull_data, 63, ##ctx) \ 5832 FN(bind, 64, ##ctx) \ 5833 FN(xdp_adjust_tail, 65, ##ctx) \ 5834 FN(skb_get_xfrm_state, 66, ##ctx) \ 5835 FN(get_stack, 67, ##ctx) \ 5836 FN(skb_load_bytes_relative, 68, ##ctx) \ 5837 FN(fib_lookup, 69, ##ctx) \ 5838 FN(sock_hash_update, 70, ##ctx) \ 5839 FN(msg_redirect_hash, 71, ##ctx) \ 5840 FN(sk_redirect_hash, 72, ##ctx) \ 5841 FN(lwt_push_encap, 73, ##ctx) \ 5842 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5843 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5844 FN(lwt_seg6_action, 76, ##ctx) \ 5845 FN(rc_repeat, 77, ##ctx) \ 5846 FN(rc_keydown, 78, ##ctx) \ 5847 FN(skb_cgroup_id, 79, ##ctx) \ 5848 FN(get_current_cgroup_id, 80, ##ctx) \ 5849 FN(get_local_storage, 81, ##ctx) \ 5850 FN(sk_select_reuseport, 82, ##ctx) \ 5851 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5852 FN(sk_lookup_tcp, 84, ##ctx) \ 5853 FN(sk_lookup_udp, 85, ##ctx) \ 5854 FN(sk_release, 86, ##ctx) \ 5855 FN(map_push_elem, 87, ##ctx) \ 5856 FN(map_pop_elem, 88, ##ctx) \ 5857 FN(map_peek_elem, 89, ##ctx) \ 5858 FN(msg_push_data, 90, ##ctx) \ 5859 FN(msg_pop_data, 91, ##ctx) \ 5860 FN(rc_pointer_rel, 92, ##ctx) \ 5861 FN(spin_lock, 93, ##ctx) \ 5862 FN(spin_unlock, 94, ##ctx) \ 5863 FN(sk_fullsock, 95, ##ctx) \ 5864 FN(tcp_sock, 96, ##ctx) \ 5865 FN(skb_ecn_set_ce, 97, ##ctx) \ 5866 FN(get_listener_sock, 98, ##ctx) \ 5867 FN(skc_lookup_tcp, 99, ##ctx) \ 5868 FN(tcp_check_syncookie, 100, ##ctx) \ 5869 FN(sysctl_get_name, 101, ##ctx) \ 5870 FN(sysctl_get_current_value, 102, ##ctx) \ 5871 FN(sysctl_get_new_value, 103, ##ctx) \ 5872 FN(sysctl_set_new_value, 104, ##ctx) \ 5873 FN(strtol, 105, ##ctx) \ 5874 FN(strtoul, 106, ##ctx) \ 5875 FN(sk_storage_get, 107, ##ctx) \ 5876 FN(sk_storage_delete, 108, ##ctx) \ 5877 FN(send_signal, 109, ##ctx) \ 5878 FN(tcp_gen_syncookie, 110, ##ctx) \ 5879 FN(skb_output, 111, ##ctx) \ 5880 FN(probe_read_user, 112, ##ctx) \ 5881 FN(probe_read_kernel, 113, ##ctx) \ 5882 FN(probe_read_user_str, 114, ##ctx) \ 5883 FN(probe_read_kernel_str, 115, ##ctx) \ 5884 FN(tcp_send_ack, 116, ##ctx) \ 5885 FN(send_signal_thread, 117, ##ctx) \ 5886 FN(jiffies64, 118, ##ctx) \ 5887 FN(read_branch_records, 119, ##ctx) \ 5888 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5889 FN(xdp_output, 121, ##ctx) \ 5890 FN(get_netns_cookie, 122, ##ctx) \ 5891 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5892 FN(sk_assign, 124, ##ctx) \ 5893 FN(ktime_get_boot_ns, 125, ##ctx) \ 5894 FN(seq_printf, 126, ##ctx) \ 5895 FN(seq_write, 127, ##ctx) \ 5896 FN(sk_cgroup_id, 128, ##ctx) \ 5897 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5898 FN(ringbuf_output, 130, ##ctx) \ 5899 FN(ringbuf_reserve, 131, ##ctx) \ 5900 FN(ringbuf_submit, 132, ##ctx) \ 5901 FN(ringbuf_discard, 133, ##ctx) \ 5902 FN(ringbuf_query, 134, ##ctx) \ 5903 FN(csum_level, 135, ##ctx) \ 5904 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5905 FN(skc_to_tcp_sock, 137, ##ctx) \ 5906 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5907 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5908 FN(skc_to_udp6_sock, 140, ##ctx) \ 5909 FN(get_task_stack, 141, ##ctx) \ 5910 FN(load_hdr_opt, 142, ##ctx) \ 5911 FN(store_hdr_opt, 143, ##ctx) \ 5912 FN(reserve_hdr_opt, 144, ##ctx) \ 5913 FN(inode_storage_get, 145, ##ctx) \ 5914 FN(inode_storage_delete, 146, ##ctx) \ 5915 FN(d_path, 147, ##ctx) \ 5916 FN(copy_from_user, 148, ##ctx) \ 5917 FN(snprintf_btf, 149, ##ctx) \ 5918 FN(seq_printf_btf, 150, ##ctx) \ 5919 FN(skb_cgroup_classid, 151, ##ctx) \ 5920 FN(redirect_neigh, 152, ##ctx) \ 5921 FN(per_cpu_ptr, 153, ##ctx) \ 5922 FN(this_cpu_ptr, 154, ##ctx) \ 5923 FN(redirect_peer, 155, ##ctx) \ 5924 FN(task_storage_get, 156, ##ctx) \ 5925 FN(task_storage_delete, 157, ##ctx) \ 5926 FN(get_current_task_btf, 158, ##ctx) \ 5927 FN(bprm_opts_set, 159, ##ctx) \ 5928 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5929 FN(ima_inode_hash, 161, ##ctx) \ 5930 FN(sock_from_file, 162, ##ctx) \ 5931 FN(check_mtu, 163, ##ctx) \ 5932 FN(for_each_map_elem, 164, ##ctx) \ 5933 FN(snprintf, 165, ##ctx) \ 5934 FN(sys_bpf, 166, ##ctx) \ 5935 FN(btf_find_by_name_kind, 167, ##ctx) \ 5936 FN(sys_close, 168, ##ctx) \ 5937 FN(timer_init, 169, ##ctx) \ 5938 FN(timer_set_callback, 170, ##ctx) \ 5939 FN(timer_start, 171, ##ctx) \ 5940 FN(timer_cancel, 172, ##ctx) \ 5941 FN(get_func_ip, 173, ##ctx) \ 5942 FN(get_attach_cookie, 174, ##ctx) \ 5943 FN(task_pt_regs, 175, ##ctx) \ 5944 FN(get_branch_snapshot, 176, ##ctx) \ 5945 FN(trace_vprintk, 177, ##ctx) \ 5946 FN(skc_to_unix_sock, 178, ##ctx) \ 5947 FN(kallsyms_lookup_name, 179, ##ctx) \ 5948 FN(find_vma, 180, ##ctx) \ 5949 FN(loop, 181, ##ctx) \ 5950 FN(strncmp, 182, ##ctx) \ 5951 FN(get_func_arg, 183, ##ctx) \ 5952 FN(get_func_ret, 184, ##ctx) \ 5953 FN(get_func_arg_cnt, 185, ##ctx) \ 5954 FN(get_retval, 186, ##ctx) \ 5955 FN(set_retval, 187, ##ctx) \ 5956 FN(xdp_get_buff_len, 188, ##ctx) \ 5957 FN(xdp_load_bytes, 189, ##ctx) \ 5958 FN(xdp_store_bytes, 190, ##ctx) \ 5959 FN(copy_from_user_task, 191, ##ctx) \ 5960 FN(skb_set_tstamp, 192, ##ctx) \ 5961 FN(ima_file_hash, 193, ##ctx) \ 5962 FN(kptr_xchg, 194, ##ctx) \ 5963 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5964 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5965 FN(dynptr_from_mem, 197, ##ctx) \ 5966 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5967 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5968 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5969 FN(dynptr_read, 201, ##ctx) \ 5970 FN(dynptr_write, 202, ##ctx) \ 5971 FN(dynptr_data, 203, ##ctx) \ 5972 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5973 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5974 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5975 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5976 FN(ktime_get_tai_ns, 208, ##ctx) \ 5977 FN(user_ringbuf_drain, 209, ##ctx) \ 5978 FN(cgrp_storage_get, 210, ##ctx) \ 5979 FN(cgrp_storage_delete, 211, ##ctx) \ 5980 /* */ 5981 5982 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 5983 * know or care about integer value that is now passed as second argument 5984 */ 5985 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 5986 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 5987 5988 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5989 * function eBPF program intends to call 5990 */ 5991 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 5992 enum bpf_func_id { 5993 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5994 __BPF_FUNC_MAX_ID, 5995 }; 5996 #undef __BPF_ENUM_FN 5997 5998 /* All flags used by eBPF helper functions, placed here. */ 5999 6000 /* BPF_FUNC_skb_store_bytes flags. */ 6001 enum { 6002 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 6003 BPF_F_INVALIDATE_HASH = (1ULL << 1), 6004 }; 6005 6006 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 6007 * First 4 bits are for passing the header field size. 6008 */ 6009 enum { 6010 BPF_F_HDR_FIELD_MASK = 0xfULL, 6011 }; 6012 6013 /* BPF_FUNC_l4_csum_replace flags. */ 6014 enum { 6015 BPF_F_PSEUDO_HDR = (1ULL << 4), 6016 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 6017 BPF_F_MARK_ENFORCE = (1ULL << 6), 6018 }; 6019 6020 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 6021 enum { 6022 BPF_F_INGRESS = (1ULL << 0), 6023 }; 6024 6025 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6026 enum { 6027 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6028 }; 6029 6030 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6031 enum { 6032 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6033 BPF_F_USER_STACK = (1ULL << 8), 6034 /* flags used by BPF_FUNC_get_stackid only. */ 6035 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6036 BPF_F_REUSE_STACKID = (1ULL << 10), 6037 /* flags used by BPF_FUNC_get_stack only. */ 6038 BPF_F_USER_BUILD_ID = (1ULL << 11), 6039 }; 6040 6041 /* BPF_FUNC_skb_set_tunnel_key flags. */ 6042 enum { 6043 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6044 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6045 BPF_F_SEQ_NUMBER = (1ULL << 3), 6046 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6047 }; 6048 6049 /* BPF_FUNC_skb_get_tunnel_key flags. */ 6050 enum { 6051 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6052 }; 6053 6054 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6055 * BPF_FUNC_perf_event_read_value flags. 6056 */ 6057 enum { 6058 BPF_F_INDEX_MASK = 0xffffffffULL, 6059 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6060 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 6061 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6062 }; 6063 6064 /* Current network namespace */ 6065 enum { 6066 BPF_F_CURRENT_NETNS = (-1L), 6067 }; 6068 6069 /* BPF_FUNC_csum_level level values. */ 6070 enum { 6071 BPF_CSUM_LEVEL_QUERY, 6072 BPF_CSUM_LEVEL_INC, 6073 BPF_CSUM_LEVEL_DEC, 6074 BPF_CSUM_LEVEL_RESET, 6075 }; 6076 6077 /* BPF_FUNC_skb_adjust_room flags. */ 6078 enum { 6079 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6080 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6081 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6082 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6083 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6084 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6085 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6086 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6087 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6088 }; 6089 6090 enum { 6091 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6092 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6093 }; 6094 6095 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6096 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6097 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6098 6099 /* BPF_FUNC_sysctl_get_name flags. */ 6100 enum { 6101 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6102 }; 6103 6104 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 6105 enum { 6106 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6107 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6108 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6109 */ 6110 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6111 }; 6112 6113 /* BPF_FUNC_read_branch_records flags. */ 6114 enum { 6115 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6116 }; 6117 6118 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6119 * BPF_FUNC_bpf_ringbuf_output flags. 6120 */ 6121 enum { 6122 BPF_RB_NO_WAKEUP = (1ULL << 0), 6123 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6124 }; 6125 6126 /* BPF_FUNC_bpf_ringbuf_query flags */ 6127 enum { 6128 BPF_RB_AVAIL_DATA = 0, 6129 BPF_RB_RING_SIZE = 1, 6130 BPF_RB_CONS_POS = 2, 6131 BPF_RB_PROD_POS = 3, 6132 }; 6133 6134 /* BPF ring buffer constants */ 6135 enum { 6136 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6137 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6138 BPF_RINGBUF_HDR_SZ = 8, 6139 }; 6140 6141 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6142 enum { 6143 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6144 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6145 }; 6146 6147 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6148 enum bpf_adj_room_mode { 6149 BPF_ADJ_ROOM_NET, 6150 BPF_ADJ_ROOM_MAC, 6151 }; 6152 6153 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6154 enum bpf_hdr_start_off { 6155 BPF_HDR_START_MAC, 6156 BPF_HDR_START_NET, 6157 }; 6158 6159 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6160 enum bpf_lwt_encap_mode { 6161 BPF_LWT_ENCAP_SEG6, 6162 BPF_LWT_ENCAP_SEG6_INLINE, 6163 BPF_LWT_ENCAP_IP, 6164 }; 6165 6166 /* Flags for bpf_bprm_opts_set helper */ 6167 enum { 6168 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6169 }; 6170 6171 /* Flags for bpf_redirect_map helper */ 6172 enum { 6173 BPF_F_BROADCAST = (1ULL << 3), 6174 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 6175 }; 6176 6177 #define __bpf_md_ptr(type, name) \ 6178 union { \ 6179 type name; \ 6180 __u64 :64; \ 6181 } __attribute__((aligned(8))) 6182 6183 enum { 6184 BPF_SKB_TSTAMP_UNSPEC, 6185 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 6186 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 6187 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 6188 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 6189 */ 6190 }; 6191 6192 /* user accessible mirror of in-kernel sk_buff. 6193 * new fields can only be added to the end of this structure 6194 */ 6195 struct __sk_buff { 6196 __u32 len; 6197 __u32 pkt_type; 6198 __u32 mark; 6199 __u32 queue_mapping; 6200 __u32 protocol; 6201 __u32 vlan_present; 6202 __u32 vlan_tci; 6203 __u32 vlan_proto; 6204 __u32 priority; 6205 __u32 ingress_ifindex; 6206 __u32 ifindex; 6207 __u32 tc_index; 6208 __u32 cb[5]; 6209 __u32 hash; 6210 __u32 tc_classid; 6211 __u32 data; 6212 __u32 data_end; 6213 __u32 napi_id; 6214 6215 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6216 __u32 family; 6217 __u32 remote_ip4; /* Stored in network byte order */ 6218 __u32 local_ip4; /* Stored in network byte order */ 6219 __u32 remote_ip6[4]; /* Stored in network byte order */ 6220 __u32 local_ip6[4]; /* Stored in network byte order */ 6221 __u32 remote_port; /* Stored in network byte order */ 6222 __u32 local_port; /* stored in host byte order */ 6223 /* ... here. */ 6224 6225 __u32 data_meta; 6226 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6227 __u64 tstamp; 6228 __u32 wire_len; 6229 __u32 gso_segs; 6230 __bpf_md_ptr(struct bpf_sock *, sk); 6231 __u32 gso_size; 6232 __u8 tstamp_type; 6233 __u32 :24; /* Padding, future use. */ 6234 __u64 hwtstamp; 6235 }; 6236 6237 struct bpf_tunnel_key { 6238 __u32 tunnel_id; 6239 union { 6240 __u32 remote_ipv4; 6241 __u32 remote_ipv6[4]; 6242 }; 6243 __u8 tunnel_tos; 6244 __u8 tunnel_ttl; 6245 union { 6246 __u16 tunnel_ext; /* compat */ 6247 __be16 tunnel_flags; 6248 }; 6249 __u32 tunnel_label; 6250 union { 6251 __u32 local_ipv4; 6252 __u32 local_ipv6[4]; 6253 }; 6254 }; 6255 6256 /* user accessible mirror of in-kernel xfrm_state. 6257 * new fields can only be added to the end of this structure 6258 */ 6259 struct bpf_xfrm_state { 6260 __u32 reqid; 6261 __u32 spi; /* Stored in network byte order */ 6262 __u16 family; 6263 __u16 ext; /* Padding, future use. */ 6264 union { 6265 __u32 remote_ipv4; /* Stored in network byte order */ 6266 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6267 }; 6268 }; 6269 6270 /* Generic BPF return codes which all BPF program types may support. 6271 * The values are binary compatible with their TC_ACT_* counter-part to 6272 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6273 * programs. 6274 * 6275 * XDP is handled seprately, see XDP_*. 6276 */ 6277 enum bpf_ret_code { 6278 BPF_OK = 0, 6279 /* 1 reserved */ 6280 BPF_DROP = 2, 6281 /* 3-6 reserved */ 6282 BPF_REDIRECT = 7, 6283 /* >127 are reserved for prog type specific return codes. 6284 * 6285 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6286 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6287 * changed and should be routed based on its new L3 header. 6288 * (This is an L3 redirect, as opposed to L2 redirect 6289 * represented by BPF_REDIRECT above). 6290 */ 6291 BPF_LWT_REROUTE = 128, 6292 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6293 * to indicate that no custom dissection was performed, and 6294 * fallback to standard dissector is requested. 6295 */ 6296 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6297 }; 6298 6299 struct bpf_sock { 6300 __u32 bound_dev_if; 6301 __u32 family; 6302 __u32 type; 6303 __u32 protocol; 6304 __u32 mark; 6305 __u32 priority; 6306 /* IP address also allows 1 and 2 bytes access */ 6307 __u32 src_ip4; 6308 __u32 src_ip6[4]; 6309 __u32 src_port; /* host byte order */ 6310 __be16 dst_port; /* network byte order */ 6311 __u16 :16; /* zero padding */ 6312 __u32 dst_ip4; 6313 __u32 dst_ip6[4]; 6314 __u32 state; 6315 __s32 rx_queue_mapping; 6316 }; 6317 6318 struct bpf_tcp_sock { 6319 __u32 snd_cwnd; /* Sending congestion window */ 6320 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6321 __u32 rtt_min; 6322 __u32 snd_ssthresh; /* Slow start size threshold */ 6323 __u32 rcv_nxt; /* What we want to receive next */ 6324 __u32 snd_nxt; /* Next sequence we send */ 6325 __u32 snd_una; /* First byte we want an ack for */ 6326 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6327 __u32 ecn_flags; /* ECN status bits. */ 6328 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6329 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6330 __u32 packets_out; /* Packets which are "in flight" */ 6331 __u32 retrans_out; /* Retransmitted packets out */ 6332 __u32 total_retrans; /* Total retransmits for entire connection */ 6333 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6334 * total number of segments in. 6335 */ 6336 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6337 * total number of data segments in. 6338 */ 6339 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6340 * The total number of segments sent. 6341 */ 6342 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6343 * total number of data segments sent. 6344 */ 6345 __u32 lost_out; /* Lost packets */ 6346 __u32 sacked_out; /* SACK'd packets */ 6347 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6348 * sum(delta(rcv_nxt)), or how many bytes 6349 * were acked. 6350 */ 6351 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6352 * sum(delta(snd_una)), or how many bytes 6353 * were acked. 6354 */ 6355 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6356 * total number of DSACK blocks received 6357 */ 6358 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6359 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6360 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6361 }; 6362 6363 struct bpf_sock_tuple { 6364 union { 6365 struct { 6366 __be32 saddr; 6367 __be32 daddr; 6368 __be16 sport; 6369 __be16 dport; 6370 } ipv4; 6371 struct { 6372 __be32 saddr[4]; 6373 __be32 daddr[4]; 6374 __be16 sport; 6375 __be16 dport; 6376 } ipv6; 6377 }; 6378 }; 6379 6380 /* (Simplified) user return codes for tcx prog type. 6381 * A valid tcx program must return one of these defined values. All other 6382 * return codes are reserved for future use. Must remain compatible with 6383 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6384 * return codes are mapped to TCX_NEXT. 6385 */ 6386 enum tcx_action_base { 6387 TCX_NEXT = -1, 6388 TCX_PASS = 0, 6389 TCX_DROP = 2, 6390 TCX_REDIRECT = 7, 6391 }; 6392 6393 struct bpf_xdp_sock { 6394 __u32 queue_id; 6395 }; 6396 6397 #define XDP_PACKET_HEADROOM 256 6398 6399 /* User return codes for XDP prog type. 6400 * A valid XDP program must return one of these defined values. All other 6401 * return codes are reserved for future use. Unknown return codes will 6402 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6403 */ 6404 enum xdp_action { 6405 XDP_ABORTED = 0, 6406 XDP_DROP, 6407 XDP_PASS, 6408 XDP_TX, 6409 XDP_REDIRECT, 6410 }; 6411 6412 /* user accessible metadata for XDP packet hook 6413 * new fields must be added to the end of this structure 6414 */ 6415 struct xdp_md { 6416 __u32 data; 6417 __u32 data_end; 6418 __u32 data_meta; 6419 /* Below access go through struct xdp_rxq_info */ 6420 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6421 __u32 rx_queue_index; /* rxq->queue_index */ 6422 6423 __u32 egress_ifindex; /* txq->dev->ifindex */ 6424 }; 6425 6426 /* DEVMAP map-value layout 6427 * 6428 * The struct data-layout of map-value is a configuration interface. 6429 * New members can only be added to the end of this structure. 6430 */ 6431 struct bpf_devmap_val { 6432 __u32 ifindex; /* device index */ 6433 union { 6434 int fd; /* prog fd on map write */ 6435 __u32 id; /* prog id on map read */ 6436 } bpf_prog; 6437 }; 6438 6439 /* CPUMAP map-value layout 6440 * 6441 * The struct data-layout of map-value is a configuration interface. 6442 * New members can only be added to the end of this structure. 6443 */ 6444 struct bpf_cpumap_val { 6445 __u32 qsize; /* queue size to remote target CPU */ 6446 union { 6447 int fd; /* prog fd on map write */ 6448 __u32 id; /* prog id on map read */ 6449 } bpf_prog; 6450 }; 6451 6452 enum sk_action { 6453 SK_DROP = 0, 6454 SK_PASS, 6455 }; 6456 6457 /* user accessible metadata for SK_MSG packet hook, new fields must 6458 * be added to the end of this structure 6459 */ 6460 struct sk_msg_md { 6461 __bpf_md_ptr(void *, data); 6462 __bpf_md_ptr(void *, data_end); 6463 6464 __u32 family; 6465 __u32 remote_ip4; /* Stored in network byte order */ 6466 __u32 local_ip4; /* Stored in network byte order */ 6467 __u32 remote_ip6[4]; /* Stored in network byte order */ 6468 __u32 local_ip6[4]; /* Stored in network byte order */ 6469 __u32 remote_port; /* Stored in network byte order */ 6470 __u32 local_port; /* stored in host byte order */ 6471 __u32 size; /* Total size of sk_msg */ 6472 6473 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6474 }; 6475 6476 struct sk_reuseport_md { 6477 /* 6478 * Start of directly accessible data. It begins from 6479 * the tcp/udp header. 6480 */ 6481 __bpf_md_ptr(void *, data); 6482 /* End of directly accessible data */ 6483 __bpf_md_ptr(void *, data_end); 6484 /* 6485 * Total length of packet (starting from the tcp/udp header). 6486 * Note that the directly accessible bytes (data_end - data) 6487 * could be less than this "len". Those bytes could be 6488 * indirectly read by a helper "bpf_skb_load_bytes()". 6489 */ 6490 __u32 len; 6491 /* 6492 * Eth protocol in the mac header (network byte order). e.g. 6493 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6494 */ 6495 __u32 eth_protocol; 6496 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6497 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6498 __u32 hash; /* A hash of the packet 4 tuples */ 6499 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6500 * new incoming connection request (e.g. selecting a listen sk for 6501 * the received SYN in the TCP case). reuse->sk is one of the sk 6502 * in the reuseport group. The bpf prog can use reuse->sk to learn 6503 * the local listening ip/port without looking into the skb. 6504 * 6505 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6506 * reuse->migrating_sk is the socket that needs to be migrated 6507 * to another listening socket. migrating_sk could be a fullsock 6508 * sk that is fully established or a reqsk that is in-the-middle 6509 * of 3-way handshake. 6510 */ 6511 __bpf_md_ptr(struct bpf_sock *, sk); 6512 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6513 }; 6514 6515 #define BPF_TAG_SIZE 8 6516 6517 struct bpf_prog_info { 6518 __u32 type; 6519 __u32 id; 6520 __u8 tag[BPF_TAG_SIZE]; 6521 __u32 jited_prog_len; 6522 __u32 xlated_prog_len; 6523 __aligned_u64 jited_prog_insns; 6524 __aligned_u64 xlated_prog_insns; 6525 __u64 load_time; /* ns since boottime */ 6526 __u32 created_by_uid; 6527 __u32 nr_map_ids; 6528 __aligned_u64 map_ids; 6529 char name[BPF_OBJ_NAME_LEN]; 6530 __u32 ifindex; 6531 __u32 gpl_compatible:1; 6532 __u32 :31; /* alignment pad */ 6533 __u64 netns_dev; 6534 __u64 netns_ino; 6535 __u32 nr_jited_ksyms; 6536 __u32 nr_jited_func_lens; 6537 __aligned_u64 jited_ksyms; 6538 __aligned_u64 jited_func_lens; 6539 __u32 btf_id; 6540 __u32 func_info_rec_size; 6541 __aligned_u64 func_info; 6542 __u32 nr_func_info; 6543 __u32 nr_line_info; 6544 __aligned_u64 line_info; 6545 __aligned_u64 jited_line_info; 6546 __u32 nr_jited_line_info; 6547 __u32 line_info_rec_size; 6548 __u32 jited_line_info_rec_size; 6549 __u32 nr_prog_tags; 6550 __aligned_u64 prog_tags; 6551 __u64 run_time_ns; 6552 __u64 run_cnt; 6553 __u64 recursion_misses; 6554 __u32 verified_insns; 6555 __u32 attach_btf_obj_id; 6556 __u32 attach_btf_id; 6557 } __attribute__((aligned(8))); 6558 6559 struct bpf_map_info { 6560 __u32 type; 6561 __u32 id; 6562 __u32 key_size; 6563 __u32 value_size; 6564 __u32 max_entries; 6565 __u32 map_flags; 6566 char name[BPF_OBJ_NAME_LEN]; 6567 __u32 ifindex; 6568 __u32 btf_vmlinux_value_type_id; 6569 __u64 netns_dev; 6570 __u64 netns_ino; 6571 __u32 btf_id; 6572 __u32 btf_key_type_id; 6573 __u32 btf_value_type_id; 6574 __u32 btf_vmlinux_id; 6575 __u64 map_extra; 6576 } __attribute__((aligned(8))); 6577 6578 struct bpf_btf_info { 6579 __aligned_u64 btf; 6580 __u32 btf_size; 6581 __u32 id; 6582 __aligned_u64 name; 6583 __u32 name_len; 6584 __u32 kernel_btf; 6585 } __attribute__((aligned(8))); 6586 6587 struct bpf_link_info { 6588 __u32 type; 6589 __u32 id; 6590 __u32 prog_id; 6591 union { 6592 struct { 6593 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6594 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6595 } raw_tracepoint; 6596 struct { 6597 __u32 attach_type; 6598 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6599 __u32 target_btf_id; /* BTF type id inside the object */ 6600 } tracing; 6601 struct { 6602 __u64 cgroup_id; 6603 __u32 attach_type; 6604 } cgroup; 6605 struct { 6606 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6607 __u32 target_name_len; /* in/out: target_name buffer len */ 6608 6609 /* If the iter specific field is 32 bits, it can be put 6610 * in the first or second union. Otherwise it should be 6611 * put in the second union. 6612 */ 6613 union { 6614 struct { 6615 __u32 map_id; 6616 } map; 6617 }; 6618 union { 6619 struct { 6620 __u64 cgroup_id; 6621 __u32 order; 6622 } cgroup; 6623 struct { 6624 __u32 tid; 6625 __u32 pid; 6626 } task; 6627 }; 6628 } iter; 6629 struct { 6630 __u32 netns_ino; 6631 __u32 attach_type; 6632 } netns; 6633 struct { 6634 __u32 ifindex; 6635 } xdp; 6636 struct { 6637 __u32 map_id; 6638 } struct_ops; 6639 struct { 6640 __u32 pf; 6641 __u32 hooknum; 6642 __s32 priority; 6643 __u32 flags; 6644 } netfilter; 6645 struct { 6646 __aligned_u64 addrs; 6647 __u32 count; /* in/out: kprobe_multi function count */ 6648 __u32 flags; 6649 __u64 missed; 6650 __aligned_u64 cookies; 6651 } kprobe_multi; 6652 struct { 6653 __aligned_u64 path; 6654 __aligned_u64 offsets; 6655 __aligned_u64 ref_ctr_offsets; 6656 __aligned_u64 cookies; 6657 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6658 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6659 __u32 flags; 6660 __u32 pid; 6661 } uprobe_multi; 6662 struct { 6663 __u32 type; /* enum bpf_perf_event_type */ 6664 __u32 :32; 6665 union { 6666 struct { 6667 __aligned_u64 file_name; /* in/out */ 6668 __u32 name_len; 6669 __u32 offset; /* offset from file_name */ 6670 __u64 cookie; 6671 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6672 struct { 6673 __aligned_u64 func_name; /* in/out */ 6674 __u32 name_len; 6675 __u32 offset; /* offset from func_name */ 6676 __u64 addr; 6677 __u64 missed; 6678 __u64 cookie; 6679 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6680 struct { 6681 __aligned_u64 tp_name; /* in/out */ 6682 __u32 name_len; 6683 __u32 :32; 6684 __u64 cookie; 6685 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6686 struct { 6687 __u64 config; 6688 __u32 type; 6689 __u32 :32; 6690 __u64 cookie; 6691 } event; /* BPF_PERF_EVENT_EVENT */ 6692 }; 6693 } perf_event; 6694 struct { 6695 __u32 ifindex; 6696 __u32 attach_type; 6697 } tcx; 6698 struct { 6699 __u32 ifindex; 6700 __u32 attach_type; 6701 } netkit; 6702 }; 6703 } __attribute__((aligned(8))); 6704 6705 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6706 * by user and intended to be used by socket (e.g. to bind to, depends on 6707 * attach type). 6708 */ 6709 struct bpf_sock_addr { 6710 __u32 user_family; /* Allows 4-byte read, but no write. */ 6711 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6712 * Stored in network byte order. 6713 */ 6714 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6715 * Stored in network byte order. 6716 */ 6717 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6718 * Stored in network byte order 6719 */ 6720 __u32 family; /* Allows 4-byte read, but no write */ 6721 __u32 type; /* Allows 4-byte read, but no write */ 6722 __u32 protocol; /* Allows 4-byte read, but no write */ 6723 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6724 * Stored in network byte order. 6725 */ 6726 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6727 * Stored in network byte order. 6728 */ 6729 __bpf_md_ptr(struct bpf_sock *, sk); 6730 }; 6731 6732 /* User bpf_sock_ops struct to access socket values and specify request ops 6733 * and their replies. 6734 * Some of this fields are in network (bigendian) byte order and may need 6735 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6736 * New fields can only be added at the end of this structure 6737 */ 6738 struct bpf_sock_ops { 6739 __u32 op; 6740 union { 6741 __u32 args[4]; /* Optionally passed to bpf program */ 6742 __u32 reply; /* Returned by bpf program */ 6743 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6744 }; 6745 __u32 family; 6746 __u32 remote_ip4; /* Stored in network byte order */ 6747 __u32 local_ip4; /* Stored in network byte order */ 6748 __u32 remote_ip6[4]; /* Stored in network byte order */ 6749 __u32 local_ip6[4]; /* Stored in network byte order */ 6750 __u32 remote_port; /* Stored in network byte order */ 6751 __u32 local_port; /* stored in host byte order */ 6752 __u32 is_fullsock; /* Some TCP fields are only valid if 6753 * there is a full socket. If not, the 6754 * fields read as zero. 6755 */ 6756 __u32 snd_cwnd; 6757 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6758 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6759 __u32 state; 6760 __u32 rtt_min; 6761 __u32 snd_ssthresh; 6762 __u32 rcv_nxt; 6763 __u32 snd_nxt; 6764 __u32 snd_una; 6765 __u32 mss_cache; 6766 __u32 ecn_flags; 6767 __u32 rate_delivered; 6768 __u32 rate_interval_us; 6769 __u32 packets_out; 6770 __u32 retrans_out; 6771 __u32 total_retrans; 6772 __u32 segs_in; 6773 __u32 data_segs_in; 6774 __u32 segs_out; 6775 __u32 data_segs_out; 6776 __u32 lost_out; 6777 __u32 sacked_out; 6778 __u32 sk_txhash; 6779 __u64 bytes_received; 6780 __u64 bytes_acked; 6781 __bpf_md_ptr(struct bpf_sock *, sk); 6782 /* [skb_data, skb_data_end) covers the whole TCP header. 6783 * 6784 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6785 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6786 * header has not been written. 6787 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6788 * been written so far. 6789 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6790 * the 3WHS. 6791 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6792 * the 3WHS. 6793 * 6794 * bpf_load_hdr_opt() can also be used to read a particular option. 6795 */ 6796 __bpf_md_ptr(void *, skb_data); 6797 __bpf_md_ptr(void *, skb_data_end); 6798 __u32 skb_len; /* The total length of a packet. 6799 * It includes the header, options, 6800 * and payload. 6801 */ 6802 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6803 * an easy way to check for tcp_flags 6804 * without parsing skb_data. 6805 * 6806 * In particular, the skb_tcp_flags 6807 * will still be available in 6808 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6809 * the outgoing header has not 6810 * been written yet. 6811 */ 6812 __u64 skb_hwtstamp; 6813 }; 6814 6815 /* Definitions for bpf_sock_ops_cb_flags */ 6816 enum { 6817 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6818 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6819 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6820 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6821 /* Call bpf for all received TCP headers. The bpf prog will be 6822 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6823 * 6824 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6825 * for the header option related helpers that will be useful 6826 * to the bpf programs. 6827 * 6828 * It could be used at the client/active side (i.e. connect() side) 6829 * when the server told it that the server was in syncookie 6830 * mode and required the active side to resend the bpf-written 6831 * options. The active side can keep writing the bpf-options until 6832 * it received a valid packet from the server side to confirm 6833 * the earlier packet (and options) has been received. The later 6834 * example patch is using it like this at the active side when the 6835 * server is in syncookie mode. 6836 * 6837 * The bpf prog will usually turn this off in the common cases. 6838 */ 6839 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6840 /* Call bpf when kernel has received a header option that 6841 * the kernel cannot handle. The bpf prog will be called under 6842 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6843 * 6844 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6845 * for the header option related helpers that will be useful 6846 * to the bpf programs. 6847 */ 6848 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6849 /* Call bpf when the kernel is writing header options for the 6850 * outgoing packet. The bpf prog will first be called 6851 * to reserve space in a skb under 6852 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6853 * the bpf prog will be called to write the header option(s) 6854 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6855 * 6856 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6857 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6858 * related helpers that will be useful to the bpf programs. 6859 * 6860 * The kernel gets its chance to reserve space and write 6861 * options first before the BPF program does. 6862 */ 6863 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6864 /* Mask of all currently supported cb flags */ 6865 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6866 }; 6867 6868 /* List of known BPF sock_ops operators. 6869 * New entries can only be added at the end 6870 */ 6871 enum { 6872 BPF_SOCK_OPS_VOID, 6873 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6874 * -1 if default value should be used 6875 */ 6876 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6877 * window (in packets) or -1 if default 6878 * value should be used 6879 */ 6880 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6881 * active connection is initialized 6882 */ 6883 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6884 * active connection is 6885 * established 6886 */ 6887 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6888 * passive connection is 6889 * established 6890 */ 6891 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6892 * needs ECN 6893 */ 6894 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6895 * based on the path and may be 6896 * dependent on the congestion control 6897 * algorithm. In general it indicates 6898 * a congestion threshold. RTTs above 6899 * this indicate congestion 6900 */ 6901 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6902 * Arg1: value of icsk_retransmits 6903 * Arg2: value of icsk_rto 6904 * Arg3: whether RTO has expired 6905 */ 6906 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6907 * Arg1: sequence number of 1st byte 6908 * Arg2: # segments 6909 * Arg3: return value of 6910 * tcp_transmit_skb (0 => success) 6911 */ 6912 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6913 * Arg1: old_state 6914 * Arg2: new_state 6915 */ 6916 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6917 * socket transition to LISTEN state. 6918 */ 6919 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6920 */ 6921 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6922 * It will be called to handle 6923 * the packets received at 6924 * an already established 6925 * connection. 6926 * 6927 * sock_ops->skb_data: 6928 * Referring to the received skb. 6929 * It covers the TCP header only. 6930 * 6931 * bpf_load_hdr_opt() can also 6932 * be used to search for a 6933 * particular option. 6934 */ 6935 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6936 * header option later in 6937 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6938 * Arg1: bool want_cookie. (in 6939 * writing SYNACK only) 6940 * 6941 * sock_ops->skb_data: 6942 * Not available because no header has 6943 * been written yet. 6944 * 6945 * sock_ops->skb_tcp_flags: 6946 * The tcp_flags of the 6947 * outgoing skb. (e.g. SYN, ACK, FIN). 6948 * 6949 * bpf_reserve_hdr_opt() should 6950 * be used to reserve space. 6951 */ 6952 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6953 * Arg1: bool want_cookie. (in 6954 * writing SYNACK only) 6955 * 6956 * sock_ops->skb_data: 6957 * Referring to the outgoing skb. 6958 * It covers the TCP header 6959 * that has already been written 6960 * by the kernel and the 6961 * earlier bpf-progs. 6962 * 6963 * sock_ops->skb_tcp_flags: 6964 * The tcp_flags of the outgoing 6965 * skb. (e.g. SYN, ACK, FIN). 6966 * 6967 * bpf_store_hdr_opt() should 6968 * be used to write the 6969 * option. 6970 * 6971 * bpf_load_hdr_opt() can also 6972 * be used to search for a 6973 * particular option that 6974 * has already been written 6975 * by the kernel or the 6976 * earlier bpf-progs. 6977 */ 6978 }; 6979 6980 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6981 * changes between the TCP and BPF versions. Ideally this should never happen. 6982 * If it does, we need to add code to convert them before calling 6983 * the BPF sock_ops function. 6984 */ 6985 enum { 6986 BPF_TCP_ESTABLISHED = 1, 6987 BPF_TCP_SYN_SENT, 6988 BPF_TCP_SYN_RECV, 6989 BPF_TCP_FIN_WAIT1, 6990 BPF_TCP_FIN_WAIT2, 6991 BPF_TCP_TIME_WAIT, 6992 BPF_TCP_CLOSE, 6993 BPF_TCP_CLOSE_WAIT, 6994 BPF_TCP_LAST_ACK, 6995 BPF_TCP_LISTEN, 6996 BPF_TCP_CLOSING, /* Now a valid state */ 6997 BPF_TCP_NEW_SYN_RECV, 6998 BPF_TCP_BOUND_INACTIVE, 6999 7000 BPF_TCP_MAX_STATES /* Leave at the end! */ 7001 }; 7002 7003 enum { 7004 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 7005 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 7006 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 7007 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 7008 /* Copy the SYN pkt to optval 7009 * 7010 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 7011 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 7012 * to only getting from the saved_syn. It can either get the 7013 * syn packet from: 7014 * 7015 * 1. the just-received SYN packet (only available when writing the 7016 * SYNACK). It will be useful when it is not necessary to 7017 * save the SYN packet for latter use. It is also the only way 7018 * to get the SYN during syncookie mode because the syn 7019 * packet cannot be saved during syncookie. 7020 * 7021 * OR 7022 * 7023 * 2. the earlier saved syn which was done by 7024 * bpf_setsockopt(TCP_SAVE_SYN). 7025 * 7026 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7027 * SYN packet is obtained. 7028 * 7029 * If the bpf-prog does not need the IP[46] header, the 7030 * bpf-prog can avoid parsing the IP header by using 7031 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7032 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7033 * 7034 * >0: Total number of bytes copied 7035 * -ENOSPC: Not enough space in optval. Only optlen number of 7036 * bytes is copied. 7037 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7038 * is not saved by setsockopt(TCP_SAVE_SYN). 7039 */ 7040 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7041 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7042 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7043 }; 7044 7045 enum { 7046 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7047 }; 7048 7049 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7050 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7051 */ 7052 enum { 7053 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7054 * total option spaces 7055 * required for an established 7056 * sk in order to calculate the 7057 * MSS. No skb is actually 7058 * sent. 7059 */ 7060 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7061 * when sending a SYN. 7062 */ 7063 }; 7064 7065 struct bpf_perf_event_value { 7066 __u64 counter; 7067 __u64 enabled; 7068 __u64 running; 7069 }; 7070 7071 enum { 7072 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7073 BPF_DEVCG_ACC_READ = (1ULL << 1), 7074 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7075 }; 7076 7077 enum { 7078 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7079 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7080 }; 7081 7082 struct bpf_cgroup_dev_ctx { 7083 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7084 __u32 access_type; 7085 __u32 major; 7086 __u32 minor; 7087 }; 7088 7089 struct bpf_raw_tracepoint_args { 7090 __u64 args[0]; 7091 }; 7092 7093 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7094 * OUTPUT: Do lookup from egress perspective; default is ingress 7095 */ 7096 enum { 7097 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7098 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7099 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7100 BPF_FIB_LOOKUP_TBID = (1U << 3), 7101 BPF_FIB_LOOKUP_SRC = (1U << 4), 7102 }; 7103 7104 enum { 7105 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7106 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7107 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7108 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7109 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7110 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7111 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7112 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7113 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7114 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7115 }; 7116 7117 struct bpf_fib_lookup { 7118 /* input: network family for lookup (AF_INET, AF_INET6) 7119 * output: network family of egress nexthop 7120 */ 7121 __u8 family; 7122 7123 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7124 __u8 l4_protocol; 7125 __be16 sport; 7126 __be16 dport; 7127 7128 union { /* used for MTU check */ 7129 /* input to lookup */ 7130 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7131 7132 /* output: MTU value */ 7133 __u16 mtu_result; 7134 }; 7135 /* input: L3 device index for lookup 7136 * output: device index from FIB lookup 7137 */ 7138 __u32 ifindex; 7139 7140 union { 7141 /* inputs to lookup */ 7142 __u8 tos; /* AF_INET */ 7143 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7144 7145 /* output: metric of fib result (IPv4/IPv6 only) */ 7146 __u32 rt_metric; 7147 }; 7148 7149 /* input: source address to consider for lookup 7150 * output: source address result from lookup 7151 */ 7152 union { 7153 __be32 ipv4_src; 7154 __u32 ipv6_src[4]; /* in6_addr; network order */ 7155 }; 7156 7157 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7158 * network header. output: bpf_fib_lookup sets to gateway address 7159 * if FIB lookup returns gateway route 7160 */ 7161 union { 7162 __be32 ipv4_dst; 7163 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7164 }; 7165 7166 union { 7167 struct { 7168 /* output */ 7169 __be16 h_vlan_proto; 7170 __be16 h_vlan_TCI; 7171 }; 7172 /* input: when accompanied with the 7173 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7174 * specific routing table to use for the fib lookup. 7175 */ 7176 __u32 tbid; 7177 }; 7178 7179 __u8 smac[6]; /* ETH_ALEN */ 7180 __u8 dmac[6]; /* ETH_ALEN */ 7181 }; 7182 7183 struct bpf_redir_neigh { 7184 /* network family for lookup (AF_INET, AF_INET6) */ 7185 __u32 nh_family; 7186 /* network address of nexthop; skips fib lookup to find gateway */ 7187 union { 7188 __be32 ipv4_nh; 7189 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7190 }; 7191 }; 7192 7193 /* bpf_check_mtu flags*/ 7194 enum bpf_check_mtu_flags { 7195 BPF_MTU_CHK_SEGS = (1U << 0), 7196 }; 7197 7198 enum bpf_check_mtu_ret { 7199 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7200 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7201 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7202 }; 7203 7204 enum bpf_task_fd_type { 7205 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7206 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7207 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7208 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7209 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7210 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7211 }; 7212 7213 enum { 7214 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7215 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7216 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7217 }; 7218 7219 struct bpf_flow_keys { 7220 __u16 nhoff; 7221 __u16 thoff; 7222 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7223 __u8 is_frag; 7224 __u8 is_first_frag; 7225 __u8 is_encap; 7226 __u8 ip_proto; 7227 __be16 n_proto; 7228 __be16 sport; 7229 __be16 dport; 7230 union { 7231 struct { 7232 __be32 ipv4_src; 7233 __be32 ipv4_dst; 7234 }; 7235 struct { 7236 __u32 ipv6_src[4]; /* in6_addr; network order */ 7237 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7238 }; 7239 }; 7240 __u32 flags; 7241 __be32 flow_label; 7242 }; 7243 7244 struct bpf_func_info { 7245 __u32 insn_off; 7246 __u32 type_id; 7247 }; 7248 7249 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7250 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7251 7252 struct bpf_line_info { 7253 __u32 insn_off; 7254 __u32 file_name_off; 7255 __u32 line_off; 7256 __u32 line_col; 7257 }; 7258 7259 struct bpf_spin_lock { 7260 __u32 val; 7261 }; 7262 7263 struct bpf_timer { 7264 __u64 __opaque[2]; 7265 } __attribute__((aligned(8))); 7266 7267 struct bpf_dynptr { 7268 __u64 __opaque[2]; 7269 } __attribute__((aligned(8))); 7270 7271 struct bpf_list_head { 7272 __u64 __opaque[2]; 7273 } __attribute__((aligned(8))); 7274 7275 struct bpf_list_node { 7276 __u64 __opaque[3]; 7277 } __attribute__((aligned(8))); 7278 7279 struct bpf_rb_root { 7280 __u64 __opaque[2]; 7281 } __attribute__((aligned(8))); 7282 7283 struct bpf_rb_node { 7284 __u64 __opaque[4]; 7285 } __attribute__((aligned(8))); 7286 7287 struct bpf_refcount { 7288 __u32 __opaque[1]; 7289 } __attribute__((aligned(4))); 7290 7291 struct bpf_sysctl { 7292 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7293 * Allows 1,2,4-byte read, but no write. 7294 */ 7295 __u32 file_pos; /* Sysctl file position to read from, write to. 7296 * Allows 1,2,4-byte read an 4-byte write. 7297 */ 7298 }; 7299 7300 struct bpf_sockopt { 7301 __bpf_md_ptr(struct bpf_sock *, sk); 7302 __bpf_md_ptr(void *, optval); 7303 __bpf_md_ptr(void *, optval_end); 7304 7305 __s32 level; 7306 __s32 optname; 7307 __s32 optlen; 7308 __s32 retval; 7309 }; 7310 7311 struct bpf_pidns_info { 7312 __u32 pid; 7313 __u32 tgid; 7314 }; 7315 7316 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7317 struct bpf_sk_lookup { 7318 union { 7319 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7320 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7321 }; 7322 7323 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7324 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7325 __u32 remote_ip4; /* Network byte order */ 7326 __u32 remote_ip6[4]; /* Network byte order */ 7327 __be16 remote_port; /* Network byte order */ 7328 __u16 :16; /* Zero padding */ 7329 __u32 local_ip4; /* Network byte order */ 7330 __u32 local_ip6[4]; /* Network byte order */ 7331 __u32 local_port; /* Host byte order */ 7332 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7333 }; 7334 7335 /* 7336 * struct btf_ptr is used for typed pointer representation; the 7337 * type id is used to render the pointer data as the appropriate type 7338 * via the bpf_snprintf_btf() helper described above. A flags field - 7339 * potentially to specify additional details about the BTF pointer 7340 * (rather than its mode of display) - is included for future use. 7341 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7342 */ 7343 struct btf_ptr { 7344 void *ptr; 7345 __u32 type_id; 7346 __u32 flags; /* BTF ptr flags; unused at present. */ 7347 }; 7348 7349 /* 7350 * Flags to control bpf_snprintf_btf() behaviour. 7351 * - BTF_F_COMPACT: no formatting around type information 7352 * - BTF_F_NONAME: no struct/union member names/types 7353 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7354 * equivalent to %px. 7355 * - BTF_F_ZERO: show zero-valued struct/union members; they 7356 * are not displayed by default 7357 */ 7358 enum { 7359 BTF_F_COMPACT = (1ULL << 0), 7360 BTF_F_NONAME = (1ULL << 1), 7361 BTF_F_PTR_RAW = (1ULL << 2), 7362 BTF_F_ZERO = (1ULL << 3), 7363 }; 7364 7365 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7366 * has to be adjusted by relocations. It is emitted by llvm and passed to 7367 * libbpf and later to the kernel. 7368 */ 7369 enum bpf_core_relo_kind { 7370 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7371 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7372 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7373 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7374 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7375 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7376 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7377 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7378 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7379 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7380 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7381 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7382 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7383 }; 7384 7385 /* 7386 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7387 * and from libbpf to the kernel. 7388 * 7389 * CO-RE relocation captures the following data: 7390 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7391 * its insn->imm field to be relocated with actual field info; 7392 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7393 * type or field; 7394 * - access_str_off - offset into corresponding .BTF string section. String 7395 * interpretation depends on specific relocation kind: 7396 * - for field-based relocations, string encodes an accessed field using 7397 * a sequence of field and array indices, separated by colon (:). It's 7398 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7399 * arguments for identifying offset to a field. 7400 * - for type-based relocations, strings is expected to be just "0"; 7401 * - for enum value-based relocations, string contains an index of enum 7402 * value within its enum type; 7403 * - kind - one of enum bpf_core_relo_kind; 7404 * 7405 * Example: 7406 * struct sample { 7407 * int a; 7408 * struct { 7409 * int b[10]; 7410 * }; 7411 * }; 7412 * 7413 * struct sample *s = ...; 7414 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7415 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7416 * // b is field #0 inside anon struct, accessing elem #5) 7417 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7418 * 7419 * type_id for all relocs in this example will capture BTF type id of 7420 * `struct sample`. 7421 * 7422 * Such relocation is emitted when using __builtin_preserve_access_index() 7423 * Clang built-in, passing expression that captures field address, e.g.: 7424 * 7425 * bpf_probe_read(&dst, sizeof(dst), 7426 * __builtin_preserve_access_index(&src->a.b.c)); 7427 * 7428 * In this case Clang will emit field relocation recording necessary data to 7429 * be able to find offset of embedded `a.b.c` field within `src` struct. 7430 * 7431 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7432 */ 7433 struct bpf_core_relo { 7434 __u32 insn_off; 7435 __u32 type_id; 7436 __u32 access_str_off; 7437 enum bpf_core_relo_kind kind; 7438 }; 7439 7440 /* 7441 * Flags to control bpf_timer_start() behaviour. 7442 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7443 * relative to current time. 7444 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7445 */ 7446 enum { 7447 BPF_F_TIMER_ABS = (1ULL << 0), 7448 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7449 }; 7450 7451 /* BPF numbers iterator state */ 7452 struct bpf_iter_num { 7453 /* opaque iterator state; having __u64 here allows to preserve correct 7454 * alignment requirements in vmlinux.h, generated from BTF 7455 */ 7456 __u64 __opaque[1]; 7457 } __attribute__((aligned(8))); 7458 7459 #endif /* _UAPI__LINUX_BPF_H__ */ 7460