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