github.com/cilium/cilium@v1.16.2/bpf/include/linux/bpf.h (about) 1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * Copyright Authors of the Linux kernel 4 * 5 * This program is free software; you can redistribute it and/or 6 * modify it under the terms of version 2 of the GNU General Public 7 * License as published by the Free Software Foundation. 8 */ 9 10 #pragma once 11 12 #include <linux/types.h> 13 #include <linux/bpf_common.h> 14 15 /* Extended instruction set based on top of classic BPF */ 16 17 /* instruction classes */ 18 #define BPF_JMP32 0x06 /* jmp mode in word width */ 19 #define BPF_ALU64 0x07 /* alu mode in double word width */ 20 21 /* ld/ldx fields */ 22 #define BPF_DW 0x18 /* double word (64-bit) */ 23 #define BPF_MEMSX 0x80 /* load with sign extension */ 24 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 25 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 26 27 /* alu/jmp fields */ 28 #define BPF_MOV 0xb0 /* mov reg to reg */ 29 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 30 31 /* change endianness of a register */ 32 #define BPF_END 0xd0 /* flags for endianness conversion: */ 33 #define BPF_TO_LE 0x00 /* convert to little-endian */ 34 #define BPF_TO_BE 0x08 /* convert to big-endian */ 35 #define BPF_FROM_LE BPF_TO_LE 36 #define BPF_FROM_BE BPF_TO_BE 37 38 /* jmp encodings */ 39 #define BPF_JNE 0x50 /* jump != */ 40 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 41 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 42 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 43 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 44 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 45 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 46 #define BPF_CALL 0x80 /* function call */ 47 #define BPF_EXIT 0x90 /* function return */ 48 49 /* atomic op type fields (stored in immediate) */ 50 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 51 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 52 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 53 54 /* Register numbers */ 55 enum { 56 BPF_REG_0 = 0, 57 BPF_REG_1, 58 BPF_REG_2, 59 BPF_REG_3, 60 BPF_REG_4, 61 BPF_REG_5, 62 BPF_REG_6, 63 BPF_REG_7, 64 BPF_REG_8, 65 BPF_REG_9, 66 BPF_REG_10, 67 __MAX_BPF_REG, 68 }; 69 70 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 71 #define MAX_BPF_REG __MAX_BPF_REG 72 73 struct bpf_insn { 74 __u8 code; /* opcode */ 75 __u8 dst_reg:4; /* dest register */ 76 __u8 src_reg:4; /* source register */ 77 __s16 off; /* signed offset */ 78 __s32 imm; /* signed immediate constant */ 79 }; 80 81 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 82 struct bpf_lpm_trie_key { 83 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 84 __u8 data[0]; /* Arbitrary size */ 85 }; 86 87 struct bpf_cgroup_storage_key { 88 __u64 cgroup_inode_id; /* cgroup inode id */ 89 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 90 }; 91 92 enum bpf_cgroup_iter_order { 93 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 94 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 95 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 96 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 97 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 98 }; 99 100 union bpf_iter_link_info { 101 struct { 102 __u32 map_fd; 103 } map; 104 struct { 105 enum bpf_cgroup_iter_order order; 106 107 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 108 * both are zero, the walk starts from the default cgroup v2 109 * root. For walking v1 hierarchy, one should always explicitly 110 * specify cgroup_fd. 111 */ 112 __u32 cgroup_fd; 113 __u64 cgroup_id; 114 } cgroup; 115 /* Parameters of task iterators. */ 116 struct { 117 __u32 tid; 118 __u32 pid; 119 __u32 pid_fd; 120 } task; 121 }; 122 123 /* BPF syscall commands, see bpf(2) man-page for more details. */ 124 /** 125 * DOC: eBPF Syscall Preamble 126 * 127 * The operation to be performed by the **bpf**\ () system call is determined 128 * by the *cmd* argument. Each operation takes an accompanying argument, 129 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 130 * below). The size argument is the size of the union pointed to by *attr*. 131 */ 132 /** 133 * DOC: eBPF Syscall Commands 134 * 135 * BPF_MAP_CREATE 136 * Description 137 * Create a map and return a file descriptor that refers to the 138 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 139 * is automatically enabled for the new file descriptor. 140 * 141 * Applying **close**\ (2) to the file descriptor returned by 142 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 143 * 144 * Return 145 * A new file descriptor (a nonnegative integer), or -1 if an 146 * error occurred (in which case, *errno* is set appropriately). 147 * 148 * BPF_MAP_LOOKUP_ELEM 149 * Description 150 * Look up an element with a given *key* in the map referred to 151 * by the file descriptor *map_fd*. 152 * 153 * The *flags* argument may be specified as one of the 154 * following: 155 * 156 * **BPF_F_LOCK** 157 * Look up the value of a spin-locked map without 158 * returning the lock. This must be specified if the 159 * elements contain a spinlock. 160 * 161 * Return 162 * Returns zero on success. On error, -1 is returned and *errno* 163 * is set appropriately. 164 * 165 * BPF_MAP_UPDATE_ELEM 166 * Description 167 * Create or update an element (key/value pair) in a specified map. 168 * 169 * The *flags* argument should be specified as one of the 170 * following: 171 * 172 * **BPF_ANY** 173 * Create a new element or update an existing element. 174 * **BPF_NOEXIST** 175 * Create a new element only if it did not exist. 176 * **BPF_EXIST** 177 * Update an existing element. 178 * **BPF_F_LOCK** 179 * Update a spin_lock-ed map element. 180 * 181 * Return 182 * Returns zero on success. On error, -1 is returned and *errno* 183 * is set appropriately. 184 * 185 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 186 * **E2BIG**, **EEXIST**, or **ENOENT**. 187 * 188 * **E2BIG** 189 * The number of elements in the map reached the 190 * *max_entries* limit specified at map creation time. 191 * **EEXIST** 192 * If *flags* specifies **BPF_NOEXIST** and the element 193 * with *key* already exists in the map. 194 * **ENOENT** 195 * If *flags* specifies **BPF_EXIST** and the element with 196 * *key* does not exist in the map. 197 * 198 * BPF_MAP_DELETE_ELEM 199 * Description 200 * Look up and delete an element by key in a specified map. 201 * 202 * Return 203 * Returns zero on success. On error, -1 is returned and *errno* 204 * is set appropriately. 205 * 206 * BPF_MAP_GET_NEXT_KEY 207 * Description 208 * Look up an element by key in a specified map and return the key 209 * of the next element. Can be used to iterate over all elements 210 * in the map. 211 * 212 * Return 213 * Returns zero on success. On error, -1 is returned and *errno* 214 * is set appropriately. 215 * 216 * The following cases can be used to iterate over all elements of 217 * the map: 218 * 219 * * If *key* is not found, the operation returns zero and sets 220 * the *next_key* pointer to the key of the first element. 221 * * If *key* is found, the operation returns zero and sets the 222 * *next_key* pointer to the key of the next element. 223 * * If *key* is the last element, returns -1 and *errno* is set 224 * to **ENOENT**. 225 * 226 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 227 * **EINVAL** on error. 228 * 229 * BPF_PROG_LOAD 230 * Description 231 * Verify and load an eBPF program, returning a new file 232 * descriptor associated with the program. 233 * 234 * Applying **close**\ (2) to the file descriptor returned by 235 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 236 * 237 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 238 * automatically enabled for the new file descriptor. 239 * 240 * Return 241 * A new file descriptor (a nonnegative integer), or -1 if an 242 * error occurred (in which case, *errno* is set appropriately). 243 * 244 * BPF_OBJ_PIN 245 * Description 246 * Pin an eBPF program or map referred by the specified *bpf_fd* 247 * to the provided *pathname* on the filesystem. 248 * 249 * The *pathname* argument must not contain a dot ("."). 250 * 251 * On success, *pathname* retains a reference to the eBPF object, 252 * preventing deallocation of the object when the original 253 * *bpf_fd* is closed. This allow the eBPF object to live beyond 254 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 255 * process. 256 * 257 * Applying **unlink**\ (2) or similar calls to the *pathname* 258 * unpins the object from the filesystem, removing the reference. 259 * If no other file descriptors or filesystem nodes refer to the 260 * same object, it will be deallocated (see NOTES). 261 * 262 * The filesystem type for the parent directory of *pathname* must 263 * be **BPF_FS_MAGIC**. 264 * 265 * Return 266 * Returns zero on success. On error, -1 is returned and *errno* 267 * is set appropriately. 268 * 269 * BPF_OBJ_GET 270 * Description 271 * Open a file descriptor for the eBPF object pinned to the 272 * specified *pathname*. 273 * 274 * Return 275 * A new file descriptor (a nonnegative integer), or -1 if an 276 * error occurred (in which case, *errno* is set appropriately). 277 * 278 * BPF_PROG_ATTACH 279 * Description 280 * Attach an eBPF program to a *target_fd* at the specified 281 * *attach_type* hook. 282 * 283 * The *attach_type* specifies the eBPF attachment point to 284 * attach the program to, and must be one of *bpf_attach_type* 285 * (see below). 286 * 287 * The *attach_bpf_fd* must be a valid file descriptor for a 288 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 289 * or sock_ops type corresponding to the specified *attach_type*. 290 * 291 * The *target_fd* must be a valid file descriptor for a kernel 292 * object which depends on the attach type of *attach_bpf_fd*: 293 * 294 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 295 * **BPF_PROG_TYPE_CGROUP_SKB**, 296 * **BPF_PROG_TYPE_CGROUP_SOCK**, 297 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 298 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 299 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 300 * **BPF_PROG_TYPE_SOCK_OPS** 301 * 302 * Control Group v2 hierarchy with the eBPF controller 303 * enabled. Requires the kernel to be compiled with 304 * **CONFIG_CGROUP_BPF**. 305 * 306 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 307 * 308 * Network namespace (eg /proc/self/ns/net). 309 * 310 * **BPF_PROG_TYPE_LIRC_MODE2** 311 * 312 * LIRC device path (eg /dev/lircN). Requires the kernel 313 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 314 * 315 * **BPF_PROG_TYPE_SK_SKB**, 316 * **BPF_PROG_TYPE_SK_MSG** 317 * 318 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 319 * 320 * Return 321 * Returns zero on success. On error, -1 is returned and *errno* 322 * is set appropriately. 323 * 324 * BPF_PROG_DETACH 325 * Description 326 * Detach the eBPF program associated with the *target_fd* at the 327 * hook specified by *attach_type*. The program must have been 328 * previously attached using **BPF_PROG_ATTACH**. 329 * 330 * Return 331 * Returns zero on success. On error, -1 is returned and *errno* 332 * is set appropriately. 333 * 334 * BPF_PROG_TEST_RUN 335 * Description 336 * Run the eBPF program associated with the *prog_fd* a *repeat* 337 * number of times against a provided program context *ctx_in* and 338 * data *data_in*, and return the modified program context 339 * *ctx_out*, *data_out* (for example, packet data), result of the 340 * execution *retval*, and *duration* of the test run. 341 * 342 * The sizes of the buffers provided as input and output 343 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 344 * be provided in the corresponding variables *ctx_size_in*, 345 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 346 * of these parameters are not provided (ie set to NULL), the 347 * corresponding size field must be zero. 348 * 349 * Some program types have particular requirements: 350 * 351 * **BPF_PROG_TYPE_SK_LOOKUP** 352 * *data_in* and *data_out* must be NULL. 353 * 354 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 355 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 356 * 357 * *ctx_out*, *data_in* and *data_out* must be NULL. 358 * *repeat* must be zero. 359 * 360 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 361 * 362 * Return 363 * Returns zero on success. On error, -1 is returned and *errno* 364 * is set appropriately. 365 * 366 * **ENOSPC** 367 * Either *data_size_out* or *ctx_size_out* is too small. 368 * **ENOTSUPP** 369 * This command is not supported by the program type of 370 * the program referred to by *prog_fd*. 371 * 372 * BPF_PROG_GET_NEXT_ID 373 * Description 374 * Fetch the next eBPF program currently loaded into the kernel. 375 * 376 * Looks for the eBPF program with an id greater than *start_id* 377 * and updates *next_id* on success. If no other eBPF programs 378 * remain with ids higher than *start_id*, returns -1 and sets 379 * *errno* to **ENOENT**. 380 * 381 * Return 382 * Returns zero on success. On error, or when no id remains, -1 383 * is returned and *errno* is set appropriately. 384 * 385 * BPF_MAP_GET_NEXT_ID 386 * Description 387 * Fetch the next eBPF map currently loaded into the kernel. 388 * 389 * Looks for the eBPF map with an id greater than *start_id* 390 * and updates *next_id* on success. If no other eBPF maps 391 * remain with ids higher than *start_id*, returns -1 and sets 392 * *errno* to **ENOENT**. 393 * 394 * Return 395 * Returns zero on success. On error, or when no id remains, -1 396 * is returned and *errno* is set appropriately. 397 * 398 * BPF_PROG_GET_FD_BY_ID 399 * Description 400 * Open a file descriptor for the eBPF program corresponding to 401 * *prog_id*. 402 * 403 * Return 404 * A new file descriptor (a nonnegative integer), or -1 if an 405 * error occurred (in which case, *errno* is set appropriately). 406 * 407 * BPF_MAP_GET_FD_BY_ID 408 * Description 409 * Open a file descriptor for the eBPF map corresponding to 410 * *map_id*. 411 * 412 * Return 413 * A new file descriptor (a nonnegative integer), or -1 if an 414 * error occurred (in which case, *errno* is set appropriately). 415 * 416 * BPF_OBJ_GET_INFO_BY_FD 417 * Description 418 * Obtain information about the eBPF object corresponding to 419 * *bpf_fd*. 420 * 421 * Populates up to *info_len* bytes of *info*, which will be in 422 * one of the following formats depending on the eBPF object type 423 * of *bpf_fd*: 424 * 425 * * **struct bpf_prog_info** 426 * * **struct bpf_map_info** 427 * * **struct bpf_btf_info** 428 * * **struct bpf_link_info** 429 * 430 * Return 431 * Returns zero on success. On error, -1 is returned and *errno* 432 * is set appropriately. 433 * 434 * BPF_PROG_QUERY 435 * Description 436 * Obtain information about eBPF programs associated with the 437 * specified *attach_type* hook. 438 * 439 * The *target_fd* must be a valid file descriptor for a kernel 440 * object which depends on the attach type of *attach_bpf_fd*: 441 * 442 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 443 * **BPF_PROG_TYPE_CGROUP_SKB**, 444 * **BPF_PROG_TYPE_CGROUP_SOCK**, 445 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 446 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 447 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 448 * **BPF_PROG_TYPE_SOCK_OPS** 449 * 450 * Control Group v2 hierarchy with the eBPF controller 451 * enabled. Requires the kernel to be compiled with 452 * **CONFIG_CGROUP_BPF**. 453 * 454 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 455 * 456 * Network namespace (eg /proc/self/ns/net). 457 * 458 * **BPF_PROG_TYPE_LIRC_MODE2** 459 * 460 * LIRC device path (eg /dev/lircN). Requires the kernel 461 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 462 * 463 * **BPF_PROG_QUERY** always fetches the number of programs 464 * attached and the *attach_flags* which were used to attach those 465 * programs. Additionally, if *prog_ids* is nonzero and the number 466 * of attached programs is less than *prog_cnt*, populates 467 * *prog_ids* with the eBPF program ids of the programs attached 468 * at *target_fd*. 469 * 470 * The following flags may alter the result: 471 * 472 * **BPF_F_QUERY_EFFECTIVE** 473 * Only return information regarding programs which are 474 * currently effective at the specified *target_fd*. 475 * 476 * Return 477 * Returns zero on success. On error, -1 is returned and *errno* 478 * is set appropriately. 479 * 480 * BPF_RAW_TRACEPOINT_OPEN 481 * Description 482 * Attach an eBPF program to a tracepoint *name* to access kernel 483 * internal arguments of the tracepoint in their raw form. 484 * 485 * The *prog_fd* must be a valid file descriptor associated with 486 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 487 * 488 * No ABI guarantees are made about the content of tracepoint 489 * arguments exposed to the corresponding eBPF program. 490 * 491 * Applying **close**\ (2) to the file descriptor returned by 492 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 493 * 494 * Return 495 * A new file descriptor (a nonnegative integer), or -1 if an 496 * error occurred (in which case, *errno* is set appropriately). 497 * 498 * BPF_BTF_LOAD 499 * Description 500 * Verify and load BPF Type Format (BTF) metadata into the kernel, 501 * returning a new file descriptor associated with the metadata. 502 * BTF is described in more detail at 503 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 504 * 505 * The *btf* parameter must point to valid memory providing 506 * *btf_size* bytes of BTF binary metadata. 507 * 508 * The returned file descriptor can be passed to other **bpf**\ () 509 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 510 * associate the BTF with those objects. 511 * 512 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 513 * parameters to specify a *btf_log_buf*, *btf_log_size* and 514 * *btf_log_level* which allow the kernel to return freeform log 515 * output regarding the BTF verification process. 516 * 517 * Return 518 * A new file descriptor (a nonnegative integer), or -1 if an 519 * error occurred (in which case, *errno* is set appropriately). 520 * 521 * BPF_BTF_GET_FD_BY_ID 522 * Description 523 * Open a file descriptor for the BPF Type Format (BTF) 524 * corresponding to *btf_id*. 525 * 526 * Return 527 * A new file descriptor (a nonnegative integer), or -1 if an 528 * error occurred (in which case, *errno* is set appropriately). 529 * 530 * BPF_TASK_FD_QUERY 531 * Description 532 * Obtain information about eBPF programs associated with the 533 * target process identified by *pid* and *fd*. 534 * 535 * If the *pid* and *fd* are associated with a tracepoint, kprobe 536 * or uprobe perf event, then the *prog_id* and *fd_type* will 537 * be populated with the eBPF program id and file descriptor type 538 * of type **bpf_task_fd_type**. If associated with a kprobe or 539 * uprobe, the *probe_offset* and *probe_addr* will also be 540 * populated. Optionally, if *buf* is provided, then up to 541 * *buf_len* bytes of *buf* will be populated with the name of 542 * the tracepoint, kprobe or uprobe. 543 * 544 * The resulting *prog_id* may be introspected in deeper detail 545 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 546 * 547 * Return 548 * Returns zero on success. On error, -1 is returned and *errno* 549 * is set appropriately. 550 * 551 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 552 * Description 553 * Look up an element with the given *key* in the map referred to 554 * by the file descriptor *fd*, and if found, delete the element. 555 * 556 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 557 * types, the *flags* argument needs to be set to 0, but for other 558 * map types, it may be specified as: 559 * 560 * **BPF_F_LOCK** 561 * Look up and delete the value of a spin-locked map 562 * without returning the lock. This must be specified if 563 * the elements contain a spinlock. 564 * 565 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 566 * implement this command as a "pop" operation, deleting the top 567 * element rather than one corresponding to *key*. 568 * The *key* and *key_len* parameters should be zeroed when 569 * issuing this operation for these map types. 570 * 571 * This command is only valid for the following map types: 572 * * **BPF_MAP_TYPE_QUEUE** 573 * * **BPF_MAP_TYPE_STACK** 574 * * **BPF_MAP_TYPE_HASH** 575 * * **BPF_MAP_TYPE_PERCPU_HASH** 576 * * **BPF_MAP_TYPE_LRU_HASH** 577 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 578 * 579 * Return 580 * Returns zero on success. On error, -1 is returned and *errno* 581 * is set appropriately. 582 * 583 * BPF_MAP_FREEZE 584 * Description 585 * Freeze the permissions of the specified map. 586 * 587 * Write permissions may be frozen by passing zero *flags*. 588 * Upon success, no future syscall invocations may alter the 589 * map state of *map_fd*. Write operations from eBPF programs 590 * are still possible for a frozen map. 591 * 592 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 593 * 594 * Return 595 * Returns zero on success. On error, -1 is returned and *errno* 596 * is set appropriately. 597 * 598 * BPF_BTF_GET_NEXT_ID 599 * Description 600 * Fetch the next BPF Type Format (BTF) object currently loaded 601 * into the kernel. 602 * 603 * Looks for the BTF object with an id greater than *start_id* 604 * and updates *next_id* on success. If no other BTF objects 605 * remain with ids higher than *start_id*, returns -1 and sets 606 * *errno* to **ENOENT**. 607 * 608 * Return 609 * Returns zero on success. On error, or when no id remains, -1 610 * is returned and *errno* is set appropriately. 611 * 612 * BPF_MAP_LOOKUP_BATCH 613 * Description 614 * Iterate and fetch multiple elements in a map. 615 * 616 * Two opaque values are used to manage batch operations, 617 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 618 * to NULL to begin the batched operation. After each subsequent 619 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 620 * *out_batch* as the *in_batch* for the next operation to 621 * continue iteration from the current point. 622 * 623 * The *keys* and *values* are output parameters which must point 624 * to memory large enough to hold *count* items based on the key 625 * and value size of the map *map_fd*. The *keys* buffer must be 626 * of *key_size* * *count*. The *values* buffer must be of 627 * *value_size* * *count*. 628 * 629 * The *elem_flags* argument may be specified as one of the 630 * following: 631 * 632 * **BPF_F_LOCK** 633 * Look up the value of a spin-locked map without 634 * returning the lock. This must be specified if the 635 * elements contain a spinlock. 636 * 637 * On success, *count* elements from the map are copied into the 638 * user buffer, with the keys copied into *keys* and the values 639 * copied into the corresponding indices in *values*. 640 * 641 * If an error is returned and *errno* is not **EFAULT**, *count* 642 * is set to the number of successfully processed elements. 643 * 644 * Return 645 * Returns zero on success. On error, -1 is returned and *errno* 646 * is set appropriately. 647 * 648 * May set *errno* to **ENOSPC** to indicate that *keys* or 649 * *values* is too small to dump an entire bucket during 650 * iteration of a hash-based map type. 651 * 652 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 653 * Description 654 * Iterate and delete all elements in a map. 655 * 656 * This operation has the same behavior as 657 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 658 * 659 * * Every element that is successfully returned is also deleted 660 * from the map. This is at least *count* elements. Note that 661 * *count* is both an input and an output parameter. 662 * * Upon returning with *errno* set to **EFAULT**, up to 663 * *count* elements may be deleted without returning the keys 664 * and values of the deleted elements. 665 * 666 * Return 667 * Returns zero on success. On error, -1 is returned and *errno* 668 * is set appropriately. 669 * 670 * BPF_MAP_UPDATE_BATCH 671 * Description 672 * Update multiple elements in a map by *key*. 673 * 674 * The *keys* and *values* are input parameters which must point 675 * to memory large enough to hold *count* items based on the key 676 * and value size of the map *map_fd*. The *keys* buffer must be 677 * of *key_size* * *count*. The *values* buffer must be of 678 * *value_size* * *count*. 679 * 680 * Each element specified in *keys* is sequentially updated to the 681 * value in the corresponding index in *values*. The *in_batch* 682 * and *out_batch* parameters are ignored and should be zeroed. 683 * 684 * The *elem_flags* argument should be specified as one of the 685 * following: 686 * 687 * **BPF_ANY** 688 * Create new elements or update a existing elements. 689 * **BPF_NOEXIST** 690 * Create new elements only if they do not exist. 691 * **BPF_EXIST** 692 * Update existing elements. 693 * **BPF_F_LOCK** 694 * Update spin_lock-ed map elements. This must be 695 * specified if the map value contains a spinlock. 696 * 697 * On success, *count* elements from the map are updated. 698 * 699 * If an error is returned and *errno* is not **EFAULT**, *count* 700 * is set to the number of successfully processed elements. 701 * 702 * Return 703 * Returns zero on success. On error, -1 is returned and *errno* 704 * is set appropriately. 705 * 706 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 707 * **E2BIG**. **E2BIG** indicates that the number of elements in 708 * the map reached the *max_entries* limit specified at map 709 * creation time. 710 * 711 * May set *errno* to one of the following error codes under 712 * specific circumstances: 713 * 714 * **EEXIST** 715 * If *flags* specifies **BPF_NOEXIST** and the element 716 * with *key* already exists in the map. 717 * **ENOENT** 718 * If *flags* specifies **BPF_EXIST** and the element with 719 * *key* does not exist in the map. 720 * 721 * BPF_MAP_DELETE_BATCH 722 * Description 723 * Delete multiple elements in a map by *key*. 724 * 725 * The *keys* parameter is an input parameter which must point 726 * to memory large enough to hold *count* items based on the key 727 * size of the map *map_fd*, that is, *key_size* * *count*. 728 * 729 * Each element specified in *keys* is sequentially deleted. The 730 * *in_batch*, *out_batch*, and *values* parameters are ignored 731 * and should be zeroed. 732 * 733 * The *elem_flags* argument may be specified as one of the 734 * following: 735 * 736 * **BPF_F_LOCK** 737 * Look up the value of a spin-locked map without 738 * returning the lock. This must be specified if the 739 * elements contain a spinlock. 740 * 741 * On success, *count* elements from the map are updated. 742 * 743 * If an error is returned and *errno* is not **EFAULT**, *count* 744 * is set to the number of successfully processed elements. If 745 * *errno* is **EFAULT**, up to *count* elements may be been 746 * deleted. 747 * 748 * Return 749 * Returns zero on success. On error, -1 is returned and *errno* 750 * is set appropriately. 751 * 752 * BPF_LINK_CREATE 753 * Description 754 * Attach an eBPF program to a *target_fd* at the specified 755 * *attach_type* hook and return a file descriptor handle for 756 * managing the link. 757 * 758 * Return 759 * A new file descriptor (a nonnegative integer), or -1 if an 760 * error occurred (in which case, *errno* is set appropriately). 761 * 762 * BPF_LINK_UPDATE 763 * Description 764 * Update the eBPF program in the specified *link_fd* to 765 * *new_prog_fd*. 766 * 767 * Return 768 * Returns zero on success. On error, -1 is returned and *errno* 769 * is set appropriately. 770 * 771 * BPF_LINK_GET_FD_BY_ID 772 * Description 773 * Open a file descriptor for the eBPF Link corresponding to 774 * *link_id*. 775 * 776 * Return 777 * A new file descriptor (a nonnegative integer), or -1 if an 778 * error occurred (in which case, *errno* is set appropriately). 779 * 780 * BPF_LINK_GET_NEXT_ID 781 * Description 782 * Fetch the next eBPF link currently loaded into the kernel. 783 * 784 * Looks for the eBPF link with an id greater than *start_id* 785 * and updates *next_id* on success. If no other eBPF links 786 * remain with ids higher than *start_id*, returns -1 and sets 787 * *errno* to **ENOENT**. 788 * 789 * Return 790 * Returns zero on success. On error, or when no id remains, -1 791 * is returned and *errno* is set appropriately. 792 * 793 * BPF_ENABLE_STATS 794 * Description 795 * Enable eBPF runtime statistics gathering. 796 * 797 * Runtime statistics gathering for the eBPF runtime is disabled 798 * by default to minimize the corresponding performance overhead. 799 * This command enables statistics globally. 800 * 801 * Multiple programs may independently enable statistics. 802 * After gathering the desired statistics, eBPF runtime statistics 803 * may be disabled again by calling **close**\ (2) for the file 804 * descriptor returned by this function. Statistics will only be 805 * disabled system-wide when all outstanding file descriptors 806 * returned by prior calls for this subcommand are closed. 807 * 808 * Return 809 * A new file descriptor (a nonnegative integer), or -1 if an 810 * error occurred (in which case, *errno* is set appropriately). 811 * 812 * BPF_ITER_CREATE 813 * Description 814 * Create an iterator on top of the specified *link_fd* (as 815 * previously created using **BPF_LINK_CREATE**) and return a 816 * file descriptor that can be used to trigger the iteration. 817 * 818 * If the resulting file descriptor is pinned to the filesystem 819 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 820 * for that path will trigger the iterator to read kernel state 821 * using the eBPF program attached to *link_fd*. 822 * 823 * Return 824 * A new file descriptor (a nonnegative integer), or -1 if an 825 * error occurred (in which case, *errno* is set appropriately). 826 * 827 * BPF_LINK_DETACH 828 * Description 829 * Forcefully detach the specified *link_fd* from its 830 * corresponding attachment point. 831 * 832 * Return 833 * Returns zero on success. On error, -1 is returned and *errno* 834 * is set appropriately. 835 * 836 * BPF_PROG_BIND_MAP 837 * Description 838 * Bind a map to the lifetime of an eBPF program. 839 * 840 * The map identified by *map_fd* is bound to the program 841 * identified by *prog_fd* and only released when *prog_fd* is 842 * released. This may be used in cases where metadata should be 843 * associated with a program which otherwise does not contain any 844 * references to the map (for example, embedded in the eBPF 845 * program instructions). 846 * 847 * Return 848 * Returns zero on success. On error, -1 is returned and *errno* 849 * is set appropriately. 850 * 851 * NOTES 852 * eBPF objects (maps and programs) can be shared between processes. 853 * 854 * * After **fork**\ (2), the child inherits file descriptors 855 * referring to the same eBPF objects. 856 * * File descriptors referring to eBPF objects can be transferred over 857 * **unix**\ (7) domain sockets. 858 * * File descriptors referring to eBPF objects can be duplicated in the 859 * usual way, using **dup**\ (2) and similar calls. 860 * * File descriptors referring to eBPF objects can be pinned to the 861 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 862 * 863 * An eBPF object is deallocated only after all file descriptors referring 864 * to the object have been closed and no references remain pinned to the 865 * filesystem or attached (for example, bound to a program or device). 866 */ 867 enum bpf_cmd { 868 BPF_MAP_CREATE, 869 BPF_MAP_LOOKUP_ELEM, 870 BPF_MAP_UPDATE_ELEM, 871 BPF_MAP_DELETE_ELEM, 872 BPF_MAP_GET_NEXT_KEY, 873 BPF_PROG_LOAD, 874 BPF_OBJ_PIN, 875 BPF_OBJ_GET, 876 BPF_PROG_ATTACH, 877 BPF_PROG_DETACH, 878 BPF_PROG_TEST_RUN, 879 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 880 BPF_PROG_GET_NEXT_ID, 881 BPF_MAP_GET_NEXT_ID, 882 BPF_PROG_GET_FD_BY_ID, 883 BPF_MAP_GET_FD_BY_ID, 884 BPF_OBJ_GET_INFO_BY_FD, 885 BPF_PROG_QUERY, 886 BPF_RAW_TRACEPOINT_OPEN, 887 BPF_BTF_LOAD, 888 BPF_BTF_GET_FD_BY_ID, 889 BPF_TASK_FD_QUERY, 890 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 891 BPF_MAP_FREEZE, 892 BPF_BTF_GET_NEXT_ID, 893 BPF_MAP_LOOKUP_BATCH, 894 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 895 BPF_MAP_UPDATE_BATCH, 896 BPF_MAP_DELETE_BATCH, 897 BPF_LINK_CREATE, 898 BPF_LINK_UPDATE, 899 BPF_LINK_GET_FD_BY_ID, 900 BPF_LINK_GET_NEXT_ID, 901 BPF_ENABLE_STATS, 902 BPF_ITER_CREATE, 903 BPF_LINK_DETACH, 904 BPF_PROG_BIND_MAP, 905 }; 906 907 enum bpf_map_type { 908 BPF_MAP_TYPE_UNSPEC, 909 BPF_MAP_TYPE_HASH, 910 BPF_MAP_TYPE_ARRAY, 911 BPF_MAP_TYPE_PROG_ARRAY, 912 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 913 BPF_MAP_TYPE_PERCPU_HASH, 914 BPF_MAP_TYPE_PERCPU_ARRAY, 915 BPF_MAP_TYPE_STACK_TRACE, 916 BPF_MAP_TYPE_CGROUP_ARRAY, 917 BPF_MAP_TYPE_LRU_HASH, 918 BPF_MAP_TYPE_LRU_PERCPU_HASH, 919 BPF_MAP_TYPE_LPM_TRIE, 920 BPF_MAP_TYPE_ARRAY_OF_MAPS, 921 BPF_MAP_TYPE_HASH_OF_MAPS, 922 BPF_MAP_TYPE_DEVMAP, 923 BPF_MAP_TYPE_SOCKMAP, 924 BPF_MAP_TYPE_CPUMAP, 925 BPF_MAP_TYPE_XSKMAP, 926 BPF_MAP_TYPE_SOCKHASH, 927 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 928 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 929 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 930 * both cgroup-attached and other progs and supports all functionality 931 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 932 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 933 */ 934 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 935 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 936 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 937 BPF_MAP_TYPE_QUEUE, 938 BPF_MAP_TYPE_STACK, 939 BPF_MAP_TYPE_SK_STORAGE, 940 BPF_MAP_TYPE_DEVMAP_HASH, 941 BPF_MAP_TYPE_STRUCT_OPS, 942 BPF_MAP_TYPE_RINGBUF, 943 BPF_MAP_TYPE_INODE_STORAGE, 944 BPF_MAP_TYPE_TASK_STORAGE, 945 BPF_MAP_TYPE_BLOOM_FILTER, 946 BPF_MAP_TYPE_USER_RINGBUF, 947 BPF_MAP_TYPE_CGRP_STORAGE, 948 }; 949 950 /* Note that tracing related programs such as 951 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 952 * are not subject to a stable API since kernel internal data 953 * structures can change from release to release and may 954 * therefore break existing tracing BPF programs. Tracing BPF 955 * programs correspond to /a/ specific kernel which is to be 956 * analyzed, and not /a/ specific kernel /and/ all future ones. 957 */ 958 enum bpf_prog_type { 959 BPF_PROG_TYPE_UNSPEC, 960 BPF_PROG_TYPE_SOCKET_FILTER, 961 BPF_PROG_TYPE_KPROBE, 962 BPF_PROG_TYPE_SCHED_CLS, 963 BPF_PROG_TYPE_SCHED_ACT, 964 BPF_PROG_TYPE_TRACEPOINT, 965 BPF_PROG_TYPE_XDP, 966 BPF_PROG_TYPE_PERF_EVENT, 967 BPF_PROG_TYPE_CGROUP_SKB, 968 BPF_PROG_TYPE_CGROUP_SOCK, 969 BPF_PROG_TYPE_LWT_IN, 970 BPF_PROG_TYPE_LWT_OUT, 971 BPF_PROG_TYPE_LWT_XMIT, 972 BPF_PROG_TYPE_SOCK_OPS, 973 BPF_PROG_TYPE_SK_SKB, 974 BPF_PROG_TYPE_CGROUP_DEVICE, 975 BPF_PROG_TYPE_SK_MSG, 976 BPF_PROG_TYPE_RAW_TRACEPOINT, 977 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 978 BPF_PROG_TYPE_LWT_SEG6LOCAL, 979 BPF_PROG_TYPE_LIRC_MODE2, 980 BPF_PROG_TYPE_SK_REUSEPORT, 981 BPF_PROG_TYPE_FLOW_DISSECTOR, 982 BPF_PROG_TYPE_CGROUP_SYSCTL, 983 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 984 BPF_PROG_TYPE_CGROUP_SOCKOPT, 985 BPF_PROG_TYPE_TRACING, 986 BPF_PROG_TYPE_STRUCT_OPS, 987 BPF_PROG_TYPE_EXT, 988 BPF_PROG_TYPE_LSM, 989 BPF_PROG_TYPE_SK_LOOKUP, 990 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 991 BPF_PROG_TYPE_NETFILTER, 992 }; 993 994 enum bpf_attach_type { 995 BPF_CGROUP_INET_INGRESS, 996 BPF_CGROUP_INET_EGRESS, 997 BPF_CGROUP_INET_SOCK_CREATE, 998 BPF_CGROUP_SOCK_OPS, 999 BPF_SK_SKB_STREAM_PARSER, 1000 BPF_SK_SKB_STREAM_VERDICT, 1001 BPF_CGROUP_DEVICE, 1002 BPF_SK_MSG_VERDICT, 1003 BPF_CGROUP_INET4_BIND, 1004 BPF_CGROUP_INET6_BIND, 1005 BPF_CGROUP_INET4_CONNECT, 1006 BPF_CGROUP_INET6_CONNECT, 1007 BPF_CGROUP_INET4_POST_BIND, 1008 BPF_CGROUP_INET6_POST_BIND, 1009 BPF_CGROUP_UDP4_SENDMSG, 1010 BPF_CGROUP_UDP6_SENDMSG, 1011 BPF_LIRC_MODE2, 1012 BPF_FLOW_DISSECTOR, 1013 BPF_CGROUP_SYSCTL, 1014 BPF_CGROUP_UDP4_RECVMSG, 1015 BPF_CGROUP_UDP6_RECVMSG, 1016 BPF_CGROUP_GETSOCKOPT, 1017 BPF_CGROUP_SETSOCKOPT, 1018 BPF_TRACE_RAW_TP, 1019 BPF_TRACE_FENTRY, 1020 BPF_TRACE_FEXIT, 1021 BPF_MODIFY_RETURN, 1022 BPF_LSM_MAC, 1023 BPF_TRACE_ITER, 1024 BPF_CGROUP_INET4_GETPEERNAME, 1025 BPF_CGROUP_INET6_GETPEERNAME, 1026 BPF_CGROUP_INET4_GETSOCKNAME, 1027 BPF_CGROUP_INET6_GETSOCKNAME, 1028 BPF_XDP_DEVMAP, 1029 BPF_CGROUP_INET_SOCK_RELEASE, 1030 BPF_XDP_CPUMAP, 1031 BPF_SK_LOOKUP, 1032 BPF_XDP, 1033 BPF_SK_SKB_VERDICT, 1034 BPF_SK_REUSEPORT_SELECT, 1035 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1036 BPF_PERF_EVENT, 1037 BPF_TRACE_KPROBE_MULTI, 1038 BPF_LSM_CGROUP, 1039 BPF_STRUCT_OPS, 1040 BPF_NETFILTER, 1041 BPF_TCX_INGRESS, 1042 BPF_TCX_EGRESS, 1043 BPF_TRACE_UPROBE_MULTI, 1044 __MAX_BPF_ATTACH_TYPE 1045 }; 1046 1047 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1048 1049 enum bpf_link_type { 1050 BPF_LINK_TYPE_UNSPEC = 0, 1051 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1052 BPF_LINK_TYPE_TRACING = 2, 1053 BPF_LINK_TYPE_CGROUP = 3, 1054 BPF_LINK_TYPE_ITER = 4, 1055 BPF_LINK_TYPE_NETNS = 5, 1056 BPF_LINK_TYPE_XDP = 6, 1057 BPF_LINK_TYPE_PERF_EVENT = 7, 1058 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1059 BPF_LINK_TYPE_STRUCT_OPS = 9, 1060 BPF_LINK_TYPE_NETFILTER = 10, 1061 BPF_LINK_TYPE_TCX = 11, 1062 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1063 MAX_BPF_LINK_TYPE, 1064 }; 1065 1066 enum bpf_perf_event_type { 1067 BPF_PERF_EVENT_UNSPEC = 0, 1068 BPF_PERF_EVENT_UPROBE = 1, 1069 BPF_PERF_EVENT_URETPROBE = 2, 1070 BPF_PERF_EVENT_KPROBE = 3, 1071 BPF_PERF_EVENT_KRETPROBE = 4, 1072 BPF_PERF_EVENT_TRACEPOINT = 5, 1073 BPF_PERF_EVENT_EVENT = 6, 1074 }; 1075 1076 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1077 * 1078 * NONE(default): No further bpf programs allowed in the subtree. 1079 * 1080 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1081 * the program in this cgroup yields to sub-cgroup program. 1082 * 1083 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1084 * that cgroup program gets run in addition to the program in this cgroup. 1085 * 1086 * Only one program is allowed to be attached to a cgroup with 1087 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1088 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1089 * release old program and attach the new one. Attach flags has to match. 1090 * 1091 * Multiple programs are allowed to be attached to a cgroup with 1092 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1093 * (those that were attached first, run first) 1094 * The programs of sub-cgroup are executed first, then programs of 1095 * this cgroup and then programs of parent cgroup. 1096 * When children program makes decision (like picking TCP CA or sock bind) 1097 * parent program has a chance to override it. 1098 * 1099 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1100 * programs for a cgroup. Though it's possible to replace an old program at 1101 * any position by also specifying BPF_F_REPLACE flag and position itself in 1102 * replace_bpf_fd attribute. Old program at this position will be released. 1103 * 1104 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1105 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1106 * Ex1: 1107 * cgrp1 (MULTI progs A, B) -> 1108 * cgrp2 (OVERRIDE prog C) -> 1109 * cgrp3 (MULTI prog D) -> 1110 * cgrp4 (OVERRIDE prog E) -> 1111 * cgrp5 (NONE prog F) 1112 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1113 * if prog F is detached, the execution is E,D,A,B 1114 * if prog F and D are detached, the execution is E,A,B 1115 * if prog F, E and D are detached, the execution is C,A,B 1116 * 1117 * All eligible programs are executed regardless of return code from 1118 * earlier programs. 1119 */ 1120 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1121 #define BPF_F_ALLOW_MULTI (1U << 1) 1122 /* Generic attachment flags. */ 1123 #define BPF_F_REPLACE (1U << 2) 1124 #define BPF_F_BEFORE (1U << 3) 1125 #define BPF_F_AFTER (1U << 4) 1126 #define BPF_F_ID (1U << 5) 1127 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1128 1129 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1130 * verifier will perform strict alignment checking as if the kernel 1131 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1132 * and NET_IP_ALIGN defined to 2. 1133 */ 1134 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1135 1136 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1137 * verifier will allow any alignment whatsoever. On platforms 1138 * with strict alignment requirements for loads ands stores (such 1139 * as sparc and mips) the verifier validates that all loads and 1140 * stores provably follow this requirement. This flag turns that 1141 * checking and enforcement off. 1142 * 1143 * It is mostly used for testing when we want to validate the 1144 * context and memory access aspects of the verifier, but because 1145 * of an unaligned access the alignment check would trigger before 1146 * the one we are interested in. 1147 */ 1148 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1149 1150 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1151 * Verifier does sub-register def/use analysis and identifies instructions whose 1152 * def only matters for low 32-bit, high 32-bit is never referenced later 1153 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1154 * that it is safe to ignore clearing high 32-bit for these instructions. This 1155 * saves some back-ends a lot of code-gen. However such optimization is not 1156 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1157 * hence hasn't used verifier's analysis result. But, we really want to have a 1158 * way to be able to verify the correctness of the described optimization on 1159 * x86_64 on which testsuites are frequently exercised. 1160 * 1161 * So, this flag is introduced. Once it is set, verifier will randomize high 1162 * 32-bit for those instructions who has been identified as safe to ignore them. 1163 * Then, if verifier is not doing correct analysis, such randomization will 1164 * regress tests to expose bugs. 1165 */ 1166 #define BPF_F_TEST_RND_HI32 (1U << 2) 1167 1168 /* The verifier internal test flag. Behavior is undefined */ 1169 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1170 1171 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1172 * restrict map and helper usage for such programs. Sleepable BPF programs can 1173 * only be attached to hooks where kernel execution context allows sleeping. 1174 * Such programs are allowed to use helpers that may sleep like 1175 * bpf_copy_from_user(). 1176 */ 1177 #define BPF_F_SLEEPABLE (1U << 4) 1178 1179 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1180 * fully support xdp frags. 1181 */ 1182 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1183 1184 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1185 * program becomes device-bound but can access XDP metadata. 1186 */ 1187 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1188 1189 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1190 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1191 */ 1192 enum { 1193 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1194 }; 1195 1196 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1197 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1198 */ 1199 enum { 1200 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1201 }; 1202 1203 /* link_create.netfilter.flags used in LINK_CREATE command for 1204 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1205 */ 1206 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1207 1208 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1209 * the following extensions: 1210 * 1211 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1212 * insn[0].imm: map fd or fd_idx 1213 * insn[1].imm: 0 1214 * insn[0].off: 0 1215 * insn[1].off: 0 1216 * ldimm64 rewrite: address of map 1217 * verifier type: CONST_PTR_TO_MAP 1218 */ 1219 #define BPF_PSEUDO_MAP_FD 1 1220 #define BPF_PSEUDO_MAP_IDX 5 1221 1222 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1223 * insn[0].imm: map fd or fd_idx 1224 * insn[1].imm: offset into value 1225 * insn[0].off: 0 1226 * insn[1].off: 0 1227 * ldimm64 rewrite: address of map[0]+offset 1228 * verifier type: PTR_TO_MAP_VALUE 1229 */ 1230 #define BPF_PSEUDO_MAP_VALUE 2 1231 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1232 1233 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1234 * insn[0].imm: kernel btd id of VAR 1235 * insn[1].imm: 0 1236 * insn[0].off: 0 1237 * insn[1].off: 0 1238 * ldimm64 rewrite: address of the kernel variable 1239 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1240 * is struct/union. 1241 */ 1242 #define BPF_PSEUDO_BTF_ID 3 1243 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1244 * insn[0].imm: insn offset to the func 1245 * insn[1].imm: 0 1246 * insn[0].off: 0 1247 * insn[1].off: 0 1248 * ldimm64 rewrite: address of the function 1249 * verifier type: PTR_TO_FUNC. 1250 */ 1251 #define BPF_PSEUDO_FUNC 4 1252 1253 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1254 * offset to another bpf function 1255 */ 1256 #define BPF_PSEUDO_CALL 1 1257 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1258 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1259 */ 1260 #define BPF_PSEUDO_KFUNC_CALL 2 1261 1262 /* flags for BPF_MAP_UPDATE_ELEM command */ 1263 enum { 1264 BPF_ANY = 0, /* create new element or update existing */ 1265 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1266 BPF_EXIST = 2, /* update existing element */ 1267 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1268 }; 1269 1270 /* flags for BPF_MAP_CREATE command */ 1271 enum { 1272 BPF_F_NO_PREALLOC = (1U << 0), 1273 /* Instead of having one common LRU list in the 1274 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1275 * which can scale and perform better. 1276 * Note, the LRU nodes (including free nodes) cannot be moved 1277 * across different LRU lists. 1278 */ 1279 BPF_F_NO_COMMON_LRU = (1U << 1), 1280 /* Specify numa node during map creation */ 1281 BPF_F_NUMA_NODE = (1U << 2), 1282 1283 /* Flags for accessing BPF object from syscall side. */ 1284 BPF_F_RDONLY = (1U << 3), 1285 BPF_F_WRONLY = (1U << 4), 1286 1287 /* Flag for stack_map, store build_id+offset instead of pointer */ 1288 BPF_F_STACK_BUILD_ID = (1U << 5), 1289 1290 /* Zero-initialize hash function seed. This should only be used for testing. */ 1291 BPF_F_ZERO_SEED = (1U << 6), 1292 1293 /* Flags for accessing BPF object from program side. */ 1294 BPF_F_RDONLY_PROG = (1U << 7), 1295 BPF_F_WRONLY_PROG = (1U << 8), 1296 1297 /* Clone map from listener for newly accepted socket */ 1298 BPF_F_CLONE = (1U << 9), 1299 1300 /* Enable memory-mapping BPF map */ 1301 BPF_F_MMAPABLE = (1U << 10), 1302 1303 /* Share perf_event among processes */ 1304 BPF_F_PRESERVE_ELEMS = (1U << 11), 1305 1306 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1307 BPF_F_INNER_MAP = (1U << 12), 1308 1309 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1310 BPF_F_LINK = (1U << 13), 1311 1312 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1313 BPF_F_PATH_FD = (1U << 14), 1314 }; 1315 1316 /* Flags for BPF_PROG_QUERY. */ 1317 1318 /* Query effective (directly attached + inherited from ancestor cgroups) 1319 * programs that will be executed for events within a cgroup. 1320 * attach_flags with this flag are always returned 0. 1321 */ 1322 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1323 1324 /* Flags for BPF_PROG_TEST_RUN */ 1325 1326 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1327 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1328 /* If set, XDP frames will be transmitted after processing */ 1329 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1330 1331 /* type for BPF_ENABLE_STATS */ 1332 enum bpf_stats_type { 1333 /* enabled run_time_ns and run_cnt */ 1334 BPF_STATS_RUN_TIME = 0, 1335 }; 1336 1337 enum bpf_stack_build_id_status { 1338 /* user space need an empty entry to identify end of a trace */ 1339 BPF_STACK_BUILD_ID_EMPTY = 0, 1340 /* with valid build_id and offset */ 1341 BPF_STACK_BUILD_ID_VALID = 1, 1342 /* couldn't get build_id, fallback to ip */ 1343 BPF_STACK_BUILD_ID_IP = 2, 1344 }; 1345 1346 #define BPF_BUILD_ID_SIZE 20 1347 struct bpf_stack_build_id { 1348 __s32 status; 1349 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1350 union { 1351 __u64 offset; 1352 __u64 ip; 1353 }; 1354 }; 1355 1356 #define BPF_OBJ_NAME_LEN 16U 1357 1358 union bpf_attr { 1359 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1360 __u32 map_type; /* one of enum bpf_map_type */ 1361 __u32 key_size; /* size of key in bytes */ 1362 __u32 value_size; /* size of value in bytes */ 1363 __u32 max_entries; /* max number of entries in a map */ 1364 __u32 map_flags; /* BPF_MAP_CREATE related 1365 * flags defined above. 1366 */ 1367 __u32 inner_map_fd; /* fd pointing to the inner map */ 1368 __u32 numa_node; /* numa node (effective only if 1369 * BPF_F_NUMA_NODE is set). 1370 */ 1371 char map_name[BPF_OBJ_NAME_LEN]; 1372 __u32 map_ifindex; /* ifindex of netdev to create on */ 1373 __u32 btf_fd; /* fd pointing to a BTF type data */ 1374 __u32 btf_key_type_id; /* BTF type_id of the key */ 1375 __u32 btf_value_type_id; /* BTF type_id of the value */ 1376 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1377 * struct stored as the 1378 * map value 1379 */ 1380 /* Any per-map-type extra fields 1381 * 1382 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1383 * number of hash functions (if 0, the bloom filter will default 1384 * to using 5 hash functions). 1385 */ 1386 __u64 map_extra; 1387 }; 1388 1389 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1390 __u32 map_fd; 1391 __aligned_u64 key; 1392 union { 1393 __aligned_u64 value; 1394 __aligned_u64 next_key; 1395 }; 1396 __u64 flags; 1397 }; 1398 1399 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1400 __aligned_u64 in_batch; /* start batch, 1401 * NULL to start from beginning 1402 */ 1403 __aligned_u64 out_batch; /* output: next start batch */ 1404 __aligned_u64 keys; 1405 __aligned_u64 values; 1406 __u32 count; /* input/output: 1407 * input: # of key/value 1408 * elements 1409 * output: # of filled elements 1410 */ 1411 __u32 map_fd; 1412 __u64 elem_flags; 1413 __u64 flags; 1414 } batch; 1415 1416 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1417 __u32 prog_type; /* one of enum bpf_prog_type */ 1418 __u32 insn_cnt; 1419 __aligned_u64 insns; 1420 __aligned_u64 license; 1421 __u32 log_level; /* verbosity level of verifier */ 1422 __u32 log_size; /* size of user buffer */ 1423 __aligned_u64 log_buf; /* user supplied buffer */ 1424 __u32 kern_version; /* not used */ 1425 __u32 prog_flags; 1426 char prog_name[BPF_OBJ_NAME_LEN]; 1427 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1428 /* For some prog types expected attach type must be known at 1429 * load time to verify attach type specific parts of prog 1430 * (context accesses, allowed helpers, etc). 1431 */ 1432 __u32 expected_attach_type; 1433 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1434 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1435 __aligned_u64 func_info; /* func info */ 1436 __u32 func_info_cnt; /* number of bpf_func_info records */ 1437 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1438 __aligned_u64 line_info; /* line info */ 1439 __u32 line_info_cnt; /* number of bpf_line_info records */ 1440 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1441 union { 1442 /* valid prog_fd to attach to bpf prog */ 1443 __u32 attach_prog_fd; 1444 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1445 __u32 attach_btf_obj_fd; 1446 }; 1447 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1448 __aligned_u64 fd_array; /* array of FDs */ 1449 __aligned_u64 core_relos; 1450 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1451 /* output: actual total log contents size (including termintaing zero). 1452 * It could be both larger than original log_size (if log was 1453 * truncated), or smaller (if log buffer wasn't filled completely). 1454 */ 1455 __u32 log_true_size; 1456 }; 1457 1458 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1459 __aligned_u64 pathname; 1460 __u32 bpf_fd; 1461 __u32 file_flags; 1462 /* Same as dirfd in openat() syscall; see openat(2) 1463 * manpage for details of path FD and pathname semantics; 1464 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1465 * file_flags field, otherwise it should be set to zero; 1466 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1467 */ 1468 __s32 path_fd; 1469 }; 1470 1471 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1472 union { 1473 __u32 target_fd; /* target object to attach to or ... */ 1474 __u32 target_ifindex; /* target ifindex */ 1475 }; 1476 __u32 attach_bpf_fd; 1477 __u32 attach_type; 1478 __u32 attach_flags; 1479 __u32 replace_bpf_fd; 1480 union { 1481 __u32 relative_fd; 1482 __u32 relative_id; 1483 }; 1484 __u64 expected_revision; 1485 }; 1486 1487 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1488 __u32 prog_fd; 1489 __u32 retval; 1490 __u32 data_size_in; /* input: len of data_in */ 1491 __u32 data_size_out; /* input/output: len of data_out 1492 * returns ENOSPC if data_out 1493 * is too small. 1494 */ 1495 __aligned_u64 data_in; 1496 __aligned_u64 data_out; 1497 __u32 repeat; 1498 __u32 duration; 1499 __u32 ctx_size_in; /* input: len of ctx_in */ 1500 __u32 ctx_size_out; /* input/output: len of ctx_out 1501 * returns ENOSPC if ctx_out 1502 * is too small. 1503 */ 1504 __aligned_u64 ctx_in; 1505 __aligned_u64 ctx_out; 1506 __u32 flags; 1507 __u32 cpu; 1508 __u32 batch_size; 1509 } test; 1510 1511 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1512 union { 1513 __u32 start_id; 1514 __u32 prog_id; 1515 __u32 map_id; 1516 __u32 btf_id; 1517 __u32 link_id; 1518 }; 1519 __u32 next_id; 1520 __u32 open_flags; 1521 }; 1522 1523 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1524 __u32 bpf_fd; 1525 __u32 info_len; 1526 __aligned_u64 info; 1527 } info; 1528 1529 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1530 union { 1531 __u32 target_fd; /* target object to query or ... */ 1532 __u32 target_ifindex; /* target ifindex */ 1533 }; 1534 __u32 attach_type; 1535 __u32 query_flags; 1536 __u32 attach_flags; 1537 __aligned_u64 prog_ids; 1538 union { 1539 __u32 prog_cnt; 1540 __u32 count; 1541 }; 1542 __u32 :32; 1543 /* output: per-program attach_flags. 1544 * not allowed to be set during effective query. 1545 */ 1546 __aligned_u64 prog_attach_flags; 1547 __aligned_u64 link_ids; 1548 __aligned_u64 link_attach_flags; 1549 __u64 revision; 1550 } query; 1551 1552 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1553 __u64 name; 1554 __u32 prog_fd; 1555 } raw_tracepoint; 1556 1557 struct { /* anonymous struct for BPF_BTF_LOAD */ 1558 __aligned_u64 btf; 1559 __aligned_u64 btf_log_buf; 1560 __u32 btf_size; 1561 __u32 btf_log_size; 1562 __u32 btf_log_level; 1563 /* output: actual total log contents size (including termintaing zero). 1564 * It could be both larger than original log_size (if log was 1565 * truncated), or smaller (if log buffer wasn't filled completely). 1566 */ 1567 __u32 btf_log_true_size; 1568 }; 1569 1570 struct { 1571 __u32 pid; /* input: pid */ 1572 __u32 fd; /* input: fd */ 1573 __u32 flags; /* input: flags */ 1574 __u32 buf_len; /* input/output: buf len */ 1575 __aligned_u64 buf; /* input/output: 1576 * tp_name for tracepoint 1577 * symbol for kprobe 1578 * filename for uprobe 1579 */ 1580 __u32 prog_id; /* output: prod_id */ 1581 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1582 __u64 probe_offset; /* output: probe_offset */ 1583 __u64 probe_addr; /* output: probe_addr */ 1584 } task_fd_query; 1585 1586 struct { /* struct used by BPF_LINK_CREATE command */ 1587 union { 1588 __u32 prog_fd; /* eBPF program to attach */ 1589 __u32 map_fd; /* struct_ops to attach */ 1590 }; 1591 union { 1592 __u32 target_fd; /* target object to attach to or ... */ 1593 __u32 target_ifindex; /* target ifindex */ 1594 }; 1595 __u32 attach_type; /* attach type */ 1596 __u32 flags; /* extra flags */ 1597 union { 1598 __u32 target_btf_id; /* btf_id of target to attach to */ 1599 struct { 1600 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1601 __u32 iter_info_len; /* iter_info length */ 1602 }; 1603 struct { 1604 /* black box user-provided value passed through 1605 * to BPF program at the execution time and 1606 * accessible through bpf_get_attach_cookie() BPF helper 1607 */ 1608 __u64 bpf_cookie; 1609 } perf_event; 1610 struct { 1611 __u32 flags; 1612 __u32 cnt; 1613 __aligned_u64 syms; 1614 __aligned_u64 addrs; 1615 __aligned_u64 cookies; 1616 } kprobe_multi; 1617 struct { 1618 /* this is overlaid with the target_btf_id above. */ 1619 __u32 target_btf_id; 1620 /* black box user-provided value passed through 1621 * to BPF program at the execution time and 1622 * accessible through bpf_get_attach_cookie() BPF helper 1623 */ 1624 __u64 cookie; 1625 } tracing; 1626 struct { 1627 __u32 pf; 1628 __u32 hooknum; 1629 __s32 priority; 1630 __u32 flags; 1631 } netfilter; 1632 struct { 1633 union { 1634 __u32 relative_fd; 1635 __u32 relative_id; 1636 }; 1637 __u64 expected_revision; 1638 } tcx; 1639 struct { 1640 __aligned_u64 path; 1641 __aligned_u64 offsets; 1642 __aligned_u64 ref_ctr_offsets; 1643 __aligned_u64 cookies; 1644 __u32 cnt; 1645 __u32 flags; 1646 __u32 pid; 1647 } uprobe_multi; 1648 }; 1649 } link_create; 1650 1651 struct { /* struct used by BPF_LINK_UPDATE command */ 1652 __u32 link_fd; /* link fd */ 1653 union { 1654 /* new program fd to update link with */ 1655 __u32 new_prog_fd; 1656 /* new struct_ops map fd to update link with */ 1657 __u32 new_map_fd; 1658 }; 1659 __u32 flags; /* extra flags */ 1660 union { 1661 /* expected link's program fd; is specified only if 1662 * BPF_F_REPLACE flag is set in flags. 1663 */ 1664 __u32 old_prog_fd; 1665 /* expected link's map fd; is specified only 1666 * if BPF_F_REPLACE flag is set. 1667 */ 1668 __u32 old_map_fd; 1669 }; 1670 } link_update; 1671 1672 struct { 1673 __u32 link_fd; 1674 } link_detach; 1675 1676 struct { /* struct used by BPF_ENABLE_STATS command */ 1677 __u32 type; 1678 } enable_stats; 1679 1680 struct { /* struct used by BPF_ITER_CREATE command */ 1681 __u32 link_fd; 1682 __u32 flags; 1683 } iter_create; 1684 1685 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1686 __u32 prog_fd; 1687 __u32 map_fd; 1688 __u32 flags; /* extra flags */ 1689 } prog_bind_map; 1690 1691 } __attribute__((aligned(8))); 1692 1693 /* The description below is an attempt at providing documentation to eBPF 1694 * developers about the multiple available eBPF helper functions. It can be 1695 * parsed and used to produce a manual page. The workflow is the following, 1696 * and requires the rst2man utility: 1697 * 1698 * $ ./scripts/bpf_doc.py \ 1699 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1700 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1701 * $ man /tmp/bpf-helpers.7 1702 * 1703 * Note that in order to produce this external documentation, some RST 1704 * formatting is used in the descriptions to get "bold" and "italics" in 1705 * manual pages. Also note that the few trailing white spaces are 1706 * intentional, removing them would break paragraphs for rst2man. 1707 * 1708 * Start of BPF helper function descriptions: 1709 * 1710 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1711 * Description 1712 * Perform a lookup in *map* for an entry associated to *key*. 1713 * Return 1714 * Map value associated to *key*, or **NULL** if no entry was 1715 * found. 1716 * 1717 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1718 * Description 1719 * Add or update the value of the entry associated to *key* in 1720 * *map* with *value*. *flags* is one of: 1721 * 1722 * **BPF_NOEXIST** 1723 * The entry for *key* must not exist in the map. 1724 * **BPF_EXIST** 1725 * The entry for *key* must already exist in the map. 1726 * **BPF_ANY** 1727 * No condition on the existence of the entry for *key*. 1728 * 1729 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1730 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1731 * elements always exist), the helper would return an error. 1732 * Return 1733 * 0 on success, or a negative error in case of failure. 1734 * 1735 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1736 * Description 1737 * Delete entry with *key* from *map*. 1738 * Return 1739 * 0 on success, or a negative error in case of failure. 1740 * 1741 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1742 * Description 1743 * For tracing programs, safely attempt to read *size* bytes from 1744 * kernel space address *unsafe_ptr* and store the data in *dst*. 1745 * 1746 * Generally, use **bpf_probe_read_user**\ () or 1747 * **bpf_probe_read_kernel**\ () instead. 1748 * Return 1749 * 0 on success, or a negative error in case of failure. 1750 * 1751 * u64 bpf_ktime_get_ns(void) 1752 * Description 1753 * Return the time elapsed since system boot, in nanoseconds. 1754 * Does not include time the system was suspended. 1755 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1756 * Return 1757 * Current *ktime*. 1758 * 1759 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1760 * Description 1761 * This helper is a "printk()-like" facility for debugging. It 1762 * prints a message defined by format *fmt* (of size *fmt_size*) 1763 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1764 * available. It can take up to three additional **u64** 1765 * arguments (as an eBPF helpers, the total number of arguments is 1766 * limited to five). 1767 * 1768 * Each time the helper is called, it appends a line to the trace. 1769 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1770 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1771 * The format of the trace is customizable, and the exact output 1772 * one will get depends on the options set in 1773 * *\/sys/kernel/tracing/trace_options* (see also the 1774 * *README* file under the same directory). However, it usually 1775 * defaults to something like: 1776 * 1777 * :: 1778 * 1779 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1780 * 1781 * In the above: 1782 * 1783 * * ``telnet`` is the name of the current task. 1784 * * ``470`` is the PID of the current task. 1785 * * ``001`` is the CPU number on which the task is 1786 * running. 1787 * * In ``.N..``, each character refers to a set of 1788 * options (whether irqs are enabled, scheduling 1789 * options, whether hard/softirqs are running, level of 1790 * preempt_disabled respectively). **N** means that 1791 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1792 * are set. 1793 * * ``419421.045894`` is a timestamp. 1794 * * ``0x00000001`` is a fake value used by BPF for the 1795 * instruction pointer register. 1796 * * ``<formatted msg>`` is the message formatted with 1797 * *fmt*. 1798 * 1799 * The conversion specifiers supported by *fmt* are similar, but 1800 * more limited than for printk(). They are **%d**, **%i**, 1801 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1802 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1803 * of field, padding with zeroes, etc.) is available, and the 1804 * helper will return **-EINVAL** (but print nothing) if it 1805 * encounters an unknown specifier. 1806 * 1807 * Also, note that **bpf_trace_printk**\ () is slow, and should 1808 * only be used for debugging purposes. For this reason, a notice 1809 * block (spanning several lines) is printed to kernel logs and 1810 * states that the helper should not be used "for production use" 1811 * the first time this helper is used (or more precisely, when 1812 * **trace_printk**\ () buffers are allocated). For passing values 1813 * to user space, perf events should be preferred. 1814 * Return 1815 * The number of bytes written to the buffer, or a negative error 1816 * in case of failure. 1817 * 1818 * u32 bpf_get_prandom_u32(void) 1819 * Description 1820 * Get a pseudo-random number. 1821 * 1822 * From a security point of view, this helper uses its own 1823 * pseudo-random internal state, and cannot be used to infer the 1824 * seed of other random functions in the kernel. However, it is 1825 * essential to note that the generator used by the helper is not 1826 * cryptographically secure. 1827 * Return 1828 * A random 32-bit unsigned value. 1829 * 1830 * u32 bpf_get_smp_processor_id(void) 1831 * Description 1832 * Get the SMP (symmetric multiprocessing) processor id. Note that 1833 * all programs run with migration disabled, which means that the 1834 * SMP processor id is stable during all the execution of the 1835 * program. 1836 * Return 1837 * The SMP id of the processor running the program. 1838 * 1839 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1840 * Description 1841 * Store *len* bytes from address *from* into the packet 1842 * associated to *skb*, at *offset*. *flags* are a combination of 1843 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1844 * checksum for the packet after storing the bytes) and 1845 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1846 * **->swhash** and *skb*\ **->l4hash** to 0). 1847 * 1848 * A call to this helper is susceptible to change the underlying 1849 * packet buffer. Therefore, at load time, all checks on pointers 1850 * previously done by the verifier are invalidated and must be 1851 * performed again, if the helper is used in combination with 1852 * direct packet access. 1853 * Return 1854 * 0 on success, or a negative error in case of failure. 1855 * 1856 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1857 * Description 1858 * Recompute the layer 3 (e.g. IP) checksum for the packet 1859 * associated to *skb*. Computation is incremental, so the helper 1860 * must know the former value of the header field that was 1861 * modified (*from*), the new value of this field (*to*), and the 1862 * number of bytes (2 or 4) for this field, stored in *size*. 1863 * Alternatively, it is possible to store the difference between 1864 * the previous and the new values of the header field in *to*, by 1865 * setting *from* and *size* to 0. For both methods, *offset* 1866 * indicates the location of the IP checksum within the packet. 1867 * 1868 * This helper works in combination with **bpf_csum_diff**\ (), 1869 * which does not update the checksum in-place, but offers more 1870 * flexibility and can handle sizes larger than 2 or 4 for the 1871 * checksum to update. 1872 * 1873 * A call to this helper is susceptible to change the underlying 1874 * packet buffer. Therefore, at load time, all checks on pointers 1875 * previously done by the verifier are invalidated and must be 1876 * performed again, if the helper is used in combination with 1877 * direct packet access. 1878 * Return 1879 * 0 on success, or a negative error in case of failure. 1880 * 1881 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1882 * Description 1883 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1884 * packet associated to *skb*. Computation is incremental, so the 1885 * helper must know the former value of the header field that was 1886 * modified (*from*), the new value of this field (*to*), and the 1887 * number of bytes (2 or 4) for this field, stored on the lowest 1888 * four bits of *flags*. Alternatively, it is possible to store 1889 * the difference between the previous and the new values of the 1890 * header field in *to*, by setting *from* and the four lowest 1891 * bits of *flags* to 0. For both methods, *offset* indicates the 1892 * location of the IP checksum within the packet. In addition to 1893 * the size of the field, *flags* can be added (bitwise OR) actual 1894 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1895 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1896 * for updates resulting in a null checksum the value is set to 1897 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1898 * the checksum is to be computed against a pseudo-header. 1899 * 1900 * This helper works in combination with **bpf_csum_diff**\ (), 1901 * which does not update the checksum in-place, but offers more 1902 * flexibility and can handle sizes larger than 2 or 4 for the 1903 * checksum to update. 1904 * 1905 * A call to this helper is susceptible to change the underlying 1906 * packet buffer. Therefore, at load time, all checks on pointers 1907 * previously done by the verifier are invalidated and must be 1908 * performed again, if the helper is used in combination with 1909 * direct packet access. 1910 * Return 1911 * 0 on success, or a negative error in case of failure. 1912 * 1913 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1914 * Description 1915 * This special helper is used to trigger a "tail call", or in 1916 * other words, to jump into another eBPF program. The same stack 1917 * frame is used (but values on stack and in registers for the 1918 * caller are not accessible to the callee). This mechanism allows 1919 * for program chaining, either for raising the maximum number of 1920 * available eBPF instructions, or to execute given programs in 1921 * conditional blocks. For security reasons, there is an upper 1922 * limit to the number of successive tail calls that can be 1923 * performed. 1924 * 1925 * Upon call of this helper, the program attempts to jump into a 1926 * program referenced at index *index* in *prog_array_map*, a 1927 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1928 * *ctx*, a pointer to the context. 1929 * 1930 * If the call succeeds, the kernel immediately runs the first 1931 * instruction of the new program. This is not a function call, 1932 * and it never returns to the previous program. If the call 1933 * fails, then the helper has no effect, and the caller continues 1934 * to run its subsequent instructions. A call can fail if the 1935 * destination program for the jump does not exist (i.e. *index* 1936 * is superior to the number of entries in *prog_array_map*), or 1937 * if the maximum number of tail calls has been reached for this 1938 * chain of programs. This limit is defined in the kernel by the 1939 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1940 * which is currently set to 33. 1941 * Return 1942 * 0 on success, or a negative error in case of failure. 1943 * 1944 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1945 * Description 1946 * Clone and redirect the packet associated to *skb* to another 1947 * net device of index *ifindex*. Both ingress and egress 1948 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1949 * value in *flags* is used to make the distinction (ingress path 1950 * is selected if the flag is present, egress path otherwise). 1951 * This is the only flag supported for now. 1952 * 1953 * In comparison with **bpf_redirect**\ () helper, 1954 * **bpf_clone_redirect**\ () has the associated cost of 1955 * duplicating the packet buffer, but this can be executed out of 1956 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1957 * efficient, but it is handled through an action code where the 1958 * redirection happens only after the eBPF program has returned. 1959 * 1960 * A call to this helper is susceptible to change the underlying 1961 * packet buffer. Therefore, at load time, all checks on pointers 1962 * previously done by the verifier are invalidated and must be 1963 * performed again, if the helper is used in combination with 1964 * direct packet access. 1965 * Return 1966 * 0 on success, or a negative error in case of failure. Positive 1967 * error indicates a potential drop or congestion in the target 1968 * device. The particular positive error codes are not defined. 1969 * 1970 * u64 bpf_get_current_pid_tgid(void) 1971 * Description 1972 * Get the current pid and tgid. 1973 * Return 1974 * A 64-bit integer containing the current tgid and pid, and 1975 * created as such: 1976 * *current_task*\ **->tgid << 32 \|** 1977 * *current_task*\ **->pid**. 1978 * 1979 * u64 bpf_get_current_uid_gid(void) 1980 * Description 1981 * Get the current uid and gid. 1982 * Return 1983 * A 64-bit integer containing the current GID and UID, and 1984 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1985 * 1986 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1987 * Description 1988 * Copy the **comm** attribute of the current task into *buf* of 1989 * *size_of_buf*. The **comm** attribute contains the name of 1990 * the executable (excluding the path) for the current task. The 1991 * *size_of_buf* must be strictly positive. On success, the 1992 * helper makes sure that the *buf* is NUL-terminated. On failure, 1993 * it is filled with zeroes. 1994 * Return 1995 * 0 on success, or a negative error in case of failure. 1996 * 1997 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1998 * Description 1999 * Retrieve the classid for the current task, i.e. for the net_cls 2000 * cgroup to which *skb* belongs. 2001 * 2002 * This helper can be used on TC egress path, but not on ingress. 2003 * 2004 * The net_cls cgroup provides an interface to tag network packets 2005 * based on a user-provided identifier for all traffic coming from 2006 * the tasks belonging to the related cgroup. See also the related 2007 * kernel documentation, available from the Linux sources in file 2008 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2009 * 2010 * The Linux kernel has two versions for cgroups: there are 2011 * cgroups v1 and cgroups v2. Both are available to users, who can 2012 * use a mixture of them, but note that the net_cls cgroup is for 2013 * cgroup v1 only. This makes it incompatible with BPF programs 2014 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2015 * only hold data for one version of cgroups at a time). 2016 * 2017 * This helper is only available is the kernel was compiled with 2018 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2019 * "**y**" or to "**m**". 2020 * Return 2021 * The classid, or 0 for the default unconfigured classid. 2022 * 2023 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2024 * Description 2025 * Push a *vlan_tci* (VLAN tag control information) of protocol 2026 * *vlan_proto* to the packet associated to *skb*, then update 2027 * the checksum. Note that if *vlan_proto* is different from 2028 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2029 * be **ETH_P_8021Q**. 2030 * 2031 * A call to this helper is susceptible to change the underlying 2032 * packet buffer. Therefore, at load time, all checks on pointers 2033 * previously done by the verifier are invalidated and must be 2034 * performed again, if the helper is used in combination with 2035 * direct packet access. 2036 * Return 2037 * 0 on success, or a negative error in case of failure. 2038 * 2039 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2040 * Description 2041 * Pop a VLAN header from the packet associated to *skb*. 2042 * 2043 * A call to this helper is susceptible to change the underlying 2044 * packet buffer. Therefore, at load time, all checks on pointers 2045 * previously done by the verifier are invalidated and must be 2046 * performed again, if the helper is used in combination with 2047 * direct packet access. 2048 * Return 2049 * 0 on success, or a negative error in case of failure. 2050 * 2051 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2052 * Description 2053 * Get tunnel metadata. This helper takes a pointer *key* to an 2054 * empty **struct bpf_tunnel_key** of **size**, that will be 2055 * filled with tunnel metadata for the packet associated to *skb*. 2056 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2057 * indicates that the tunnel is based on IPv6 protocol instead of 2058 * IPv4. 2059 * 2060 * The **struct bpf_tunnel_key** is an object that generalizes the 2061 * principal parameters used by various tunneling protocols into a 2062 * single struct. This way, it can be used to easily make a 2063 * decision based on the contents of the encapsulation header, 2064 * "summarized" in this struct. In particular, it holds the IP 2065 * address of the remote end (IPv4 or IPv6, depending on the case) 2066 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2067 * this struct exposes the *key*\ **->tunnel_id**, which is 2068 * generally mapped to a VNI (Virtual Network Identifier), making 2069 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2070 * () helper. 2071 * 2072 * Let's imagine that the following code is part of a program 2073 * attached to the TC ingress interface, on one end of a GRE 2074 * tunnel, and is supposed to filter out all messages coming from 2075 * remote ends with IPv4 address other than 10.0.0.1: 2076 * 2077 * :: 2078 * 2079 * int ret; 2080 * struct bpf_tunnel_key key = {}; 2081 * 2082 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2083 * if (ret < 0) 2084 * return TC_ACT_SHOT; // drop packet 2085 * 2086 * if (key.remote_ipv4 != 0x0a000001) 2087 * return TC_ACT_SHOT; // drop packet 2088 * 2089 * return TC_ACT_OK; // accept packet 2090 * 2091 * This interface can also be used with all encapsulation devices 2092 * that can operate in "collect metadata" mode: instead of having 2093 * one network device per specific configuration, the "collect 2094 * metadata" mode only requires a single device where the 2095 * configuration can be extracted from this helper. 2096 * 2097 * This can be used together with various tunnels such as VXLan, 2098 * Geneve, GRE or IP in IP (IPIP). 2099 * Return 2100 * 0 on success, or a negative error in case of failure. 2101 * 2102 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2103 * Description 2104 * Populate tunnel metadata for packet associated to *skb.* The 2105 * tunnel metadata is set to the contents of *key*, of *size*. The 2106 * *flags* can be set to a combination of the following values: 2107 * 2108 * **BPF_F_TUNINFO_IPV6** 2109 * Indicate that the tunnel is based on IPv6 protocol 2110 * instead of IPv4. 2111 * **BPF_F_ZERO_CSUM_TX** 2112 * For IPv4 packets, add a flag to tunnel metadata 2113 * indicating that checksum computation should be skipped 2114 * and checksum set to zeroes. 2115 * **BPF_F_DONT_FRAGMENT** 2116 * Add a flag to tunnel metadata indicating that the 2117 * packet should not be fragmented. 2118 * **BPF_F_SEQ_NUMBER** 2119 * Add a flag to tunnel metadata indicating that a 2120 * sequence number should be added to tunnel header before 2121 * sending the packet. This flag was added for GRE 2122 * encapsulation, but might be used with other protocols 2123 * as well in the future. 2124 * **BPF_F_NO_TUNNEL_KEY** 2125 * Add a flag to tunnel metadata indicating that no tunnel 2126 * key should be set in the resulting tunnel header. 2127 * 2128 * Here is a typical usage on the transmit path: 2129 * 2130 * :: 2131 * 2132 * struct bpf_tunnel_key key; 2133 * populate key ... 2134 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2135 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2136 * 2137 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2138 * helper for additional information. 2139 * Return 2140 * 0 on success, or a negative error in case of failure. 2141 * 2142 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2143 * Description 2144 * Read the value of a perf event counter. This helper relies on a 2145 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2146 * the perf event counter is selected when *map* is updated with 2147 * perf event file descriptors. The *map* is an array whose size 2148 * is the number of available CPUs, and each cell contains a value 2149 * relative to one CPU. The value to retrieve is indicated by 2150 * *flags*, that contains the index of the CPU to look up, masked 2151 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2152 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2153 * current CPU should be retrieved. 2154 * 2155 * Note that before Linux 4.13, only hardware perf event can be 2156 * retrieved. 2157 * 2158 * Also, be aware that the newer helper 2159 * **bpf_perf_event_read_value**\ () is recommended over 2160 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2161 * quirks where error and counter value are used as a return code 2162 * (which is wrong to do since ranges may overlap). This issue is 2163 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2164 * time provides more features over the **bpf_perf_event_read**\ 2165 * () interface. Please refer to the description of 2166 * **bpf_perf_event_read_value**\ () for details. 2167 * Return 2168 * The value of the perf event counter read from the map, or a 2169 * negative error code in case of failure. 2170 * 2171 * long bpf_redirect(u32 ifindex, u64 flags) 2172 * Description 2173 * Redirect the packet to another net device of index *ifindex*. 2174 * This helper is somewhat similar to **bpf_clone_redirect**\ 2175 * (), except that the packet is not cloned, which provides 2176 * increased performance. 2177 * 2178 * Except for XDP, both ingress and egress interfaces can be used 2179 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2180 * to make the distinction (ingress path is selected if the flag 2181 * is present, egress path otherwise). Currently, XDP only 2182 * supports redirection to the egress interface, and accepts no 2183 * flag at all. 2184 * 2185 * The same effect can also be attained with the more generic 2186 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2187 * redirect target instead of providing it directly to the helper. 2188 * Return 2189 * For XDP, the helper returns **XDP_REDIRECT** on success or 2190 * **XDP_ABORTED** on error. For other program types, the values 2191 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2192 * error. 2193 * 2194 * u32 bpf_get_route_realm(struct sk_buff *skb) 2195 * Description 2196 * Retrieve the realm or the route, that is to say the 2197 * **tclassid** field of the destination for the *skb*. The 2198 * identifier retrieved is a user-provided tag, similar to the 2199 * one used with the net_cls cgroup (see description for 2200 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2201 * held by a route (a destination entry), not by a task. 2202 * 2203 * Retrieving this identifier works with the clsact TC egress hook 2204 * (see also **tc-bpf(8)**), or alternatively on conventional 2205 * classful egress qdiscs, but not on TC ingress path. In case of 2206 * clsact TC egress hook, this has the advantage that, internally, 2207 * the destination entry has not been dropped yet in the transmit 2208 * path. Therefore, the destination entry does not need to be 2209 * artificially held via **netif_keep_dst**\ () for a classful 2210 * qdisc until the *skb* is freed. 2211 * 2212 * This helper is available only if the kernel was compiled with 2213 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2214 * Return 2215 * The realm of the route for the packet associated to *skb*, or 0 2216 * if none was found. 2217 * 2218 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2219 * Description 2220 * Write raw *data* blob into a special BPF perf event held by 2221 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2222 * event must have the following attributes: **PERF_SAMPLE_RAW** 2223 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2224 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2225 * 2226 * The *flags* are used to indicate the index in *map* for which 2227 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2228 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2229 * to indicate that the index of the current CPU core should be 2230 * used. 2231 * 2232 * The value to write, of *size*, is passed through eBPF stack and 2233 * pointed by *data*. 2234 * 2235 * The context of the program *ctx* needs also be passed to the 2236 * helper. 2237 * 2238 * On user space, a program willing to read the values needs to 2239 * call **perf_event_open**\ () on the perf event (either for 2240 * one or for all CPUs) and to store the file descriptor into the 2241 * *map*. This must be done before the eBPF program can send data 2242 * into it. An example is available in file 2243 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2244 * tree (the eBPF program counterpart is in 2245 * *samples/bpf/trace_output_kern.c*). 2246 * 2247 * **bpf_perf_event_output**\ () achieves better performance 2248 * than **bpf_trace_printk**\ () for sharing data with user 2249 * space, and is much better suitable for streaming data from eBPF 2250 * programs. 2251 * 2252 * Note that this helper is not restricted to tracing use cases 2253 * and can be used with programs attached to TC or XDP as well, 2254 * where it allows for passing data to user space listeners. Data 2255 * can be: 2256 * 2257 * * Only custom structs, 2258 * * Only the packet payload, or 2259 * * A combination of both. 2260 * Return 2261 * 0 on success, or a negative error in case of failure. 2262 * 2263 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2264 * Description 2265 * This helper was provided as an easy way to load data from a 2266 * packet. It can be used to load *len* bytes from *offset* from 2267 * the packet associated to *skb*, into the buffer pointed by 2268 * *to*. 2269 * 2270 * Since Linux 4.7, usage of this helper has mostly been replaced 2271 * by "direct packet access", enabling packet data to be 2272 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2273 * pointing respectively to the first byte of packet data and to 2274 * the byte after the last byte of packet data. However, it 2275 * remains useful if one wishes to read large quantities of data 2276 * at once from a packet into the eBPF stack. 2277 * Return 2278 * 0 on success, or a negative error in case of failure. 2279 * 2280 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2281 * Description 2282 * Walk a user or a kernel stack and return its id. To achieve 2283 * this, the helper needs *ctx*, which is a pointer to the context 2284 * on which the tracing program is executed, and a pointer to a 2285 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2286 * 2287 * The last argument, *flags*, holds the number of stack frames to 2288 * skip (from 0 to 255), masked with 2289 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2290 * a combination of the following flags: 2291 * 2292 * **BPF_F_USER_STACK** 2293 * Collect a user space stack instead of a kernel stack. 2294 * **BPF_F_FAST_STACK_CMP** 2295 * Compare stacks by hash only. 2296 * **BPF_F_REUSE_STACKID** 2297 * If two different stacks hash into the same *stackid*, 2298 * discard the old one. 2299 * 2300 * The stack id retrieved is a 32 bit long integer handle which 2301 * can be further combined with other data (including other stack 2302 * ids) and used as a key into maps. This can be useful for 2303 * generating a variety of graphs (such as flame graphs or off-cpu 2304 * graphs). 2305 * 2306 * For walking a stack, this helper is an improvement over 2307 * **bpf_probe_read**\ (), which can be used with unrolled loops 2308 * but is not efficient and consumes a lot of eBPF instructions. 2309 * Instead, **bpf_get_stackid**\ () can collect up to 2310 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2311 * this limit can be controlled with the **sysctl** program, and 2312 * that it should be manually increased in order to profile long 2313 * user stacks (such as stacks for Java programs). To do so, use: 2314 * 2315 * :: 2316 * 2317 * # sysctl kernel.perf_event_max_stack=<new value> 2318 * Return 2319 * The positive or null stack id on success, or a negative error 2320 * in case of failure. 2321 * 2322 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2323 * Description 2324 * Compute a checksum difference, from the raw buffer pointed by 2325 * *from*, of length *from_size* (that must be a multiple of 4), 2326 * towards the raw buffer pointed by *to*, of size *to_size* 2327 * (same remark). An optional *seed* can be added to the value 2328 * (this can be cascaded, the seed may come from a previous call 2329 * to the helper). 2330 * 2331 * This is flexible enough to be used in several ways: 2332 * 2333 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2334 * checksum, it can be used when pushing new data. 2335 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2336 * checksum, it can be used when removing data from a packet. 2337 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2338 * can be used to compute a diff. Note that *from_size* and 2339 * *to_size* do not need to be equal. 2340 * 2341 * This helper can be used in combination with 2342 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2343 * which one can feed in the difference computed with 2344 * **bpf_csum_diff**\ (). 2345 * Return 2346 * The checksum result, or a negative error code in case of 2347 * failure. 2348 * 2349 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2350 * Description 2351 * Retrieve tunnel options metadata for the packet associated to 2352 * *skb*, and store the raw tunnel option data to the buffer *opt* 2353 * of *size*. 2354 * 2355 * This helper can be used with encapsulation devices that can 2356 * operate in "collect metadata" mode (please refer to the related 2357 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2358 * more details). A particular example where this can be used is 2359 * in combination with the Geneve encapsulation protocol, where it 2360 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2361 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2362 * the eBPF program. This allows for full customization of these 2363 * headers. 2364 * Return 2365 * The size of the option data retrieved. 2366 * 2367 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2368 * Description 2369 * Set tunnel options metadata for the packet associated to *skb* 2370 * to the option data contained in the raw buffer *opt* of *size*. 2371 * 2372 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2373 * helper for additional information. 2374 * Return 2375 * 0 on success, or a negative error in case of failure. 2376 * 2377 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2378 * Description 2379 * Change the protocol of the *skb* to *proto*. Currently 2380 * supported are transition from IPv4 to IPv6, and from IPv6 to 2381 * IPv4. The helper takes care of the groundwork for the 2382 * transition, including resizing the socket buffer. The eBPF 2383 * program is expected to fill the new headers, if any, via 2384 * **skb_store_bytes**\ () and to recompute the checksums with 2385 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2386 * (). The main case for this helper is to perform NAT64 2387 * operations out of an eBPF program. 2388 * 2389 * Internally, the GSO type is marked as dodgy so that headers are 2390 * checked and segments are recalculated by the GSO/GRO engine. 2391 * The size for GSO target is adapted as well. 2392 * 2393 * All values for *flags* are reserved for future usage, and must 2394 * be left at zero. 2395 * 2396 * A call to this helper is susceptible to change the underlying 2397 * packet buffer. Therefore, at load time, all checks on pointers 2398 * previously done by the verifier are invalidated and must be 2399 * performed again, if the helper is used in combination with 2400 * direct packet access. 2401 * Return 2402 * 0 on success, or a negative error in case of failure. 2403 * 2404 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2405 * Description 2406 * Change the packet type for the packet associated to *skb*. This 2407 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2408 * the eBPF program does not have a write access to *skb*\ 2409 * **->pkt_type** beside this helper. Using a helper here allows 2410 * for graceful handling of errors. 2411 * 2412 * The major use case is to change incoming *skb*s to 2413 * **PACKET_HOST** in a programmatic way instead of having to 2414 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2415 * example. 2416 * 2417 * Note that *type* only allows certain values. At this time, they 2418 * are: 2419 * 2420 * **PACKET_HOST** 2421 * Packet is for us. 2422 * **PACKET_BROADCAST** 2423 * Send packet to all. 2424 * **PACKET_MULTICAST** 2425 * Send packet to group. 2426 * **PACKET_OTHERHOST** 2427 * Send packet to someone else. 2428 * Return 2429 * 0 on success, or a negative error in case of failure. 2430 * 2431 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2432 * Description 2433 * Check whether *skb* is a descendant of the cgroup2 held by 2434 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2435 * Return 2436 * The return value depends on the result of the test, and can be: 2437 * 2438 * * 0, if the *skb* failed the cgroup2 descendant test. 2439 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2440 * * A negative error code, if an error occurred. 2441 * 2442 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2443 * Description 2444 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2445 * not set, in particular if the hash was cleared due to mangling, 2446 * recompute this hash. Later accesses to the hash can be done 2447 * directly with *skb*\ **->hash**. 2448 * 2449 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2450 * prototype with **bpf_skb_change_proto**\ (), or calling 2451 * **bpf_skb_store_bytes**\ () with the 2452 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2453 * the hash and to trigger a new computation for the next call to 2454 * **bpf_get_hash_recalc**\ (). 2455 * Return 2456 * The 32-bit hash. 2457 * 2458 * u64 bpf_get_current_task(void) 2459 * Description 2460 * Get the current task. 2461 * Return 2462 * A pointer to the current task struct. 2463 * 2464 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2465 * Description 2466 * Attempt in a safe way to write *len* bytes from the buffer 2467 * *src* to *dst* in memory. It only works for threads that are in 2468 * user context, and *dst* must be a valid user space address. 2469 * 2470 * This helper should not be used to implement any kind of 2471 * security mechanism because of TOC-TOU attacks, but rather to 2472 * debug, divert, and manipulate execution of semi-cooperative 2473 * processes. 2474 * 2475 * Keep in mind that this feature is meant for experiments, and it 2476 * has a risk of crashing the system and running programs. 2477 * Therefore, when an eBPF program using this helper is attached, 2478 * a warning including PID and process name is printed to kernel 2479 * logs. 2480 * Return 2481 * 0 on success, or a negative error in case of failure. 2482 * 2483 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2484 * Description 2485 * Check whether the probe is being run is the context of a given 2486 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2487 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2488 * Return 2489 * The return value depends on the result of the test, and can be: 2490 * 2491 * * 1, if current task belongs to the cgroup2. 2492 * * 0, if current task does not belong to the cgroup2. 2493 * * A negative error code, if an error occurred. 2494 * 2495 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2496 * Description 2497 * Resize (trim or grow) the packet associated to *skb* to the 2498 * new *len*. The *flags* are reserved for future usage, and must 2499 * be left at zero. 2500 * 2501 * The basic idea is that the helper performs the needed work to 2502 * change the size of the packet, then the eBPF program rewrites 2503 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2504 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2505 * and others. This helper is a slow path utility intended for 2506 * replies with control messages. And because it is targeted for 2507 * slow path, the helper itself can afford to be slow: it 2508 * implicitly linearizes, unclones and drops offloads from the 2509 * *skb*. 2510 * 2511 * A call to this helper is susceptible to change the underlying 2512 * packet buffer. Therefore, at load time, all checks on pointers 2513 * previously done by the verifier are invalidated and must be 2514 * performed again, if the helper is used in combination with 2515 * direct packet access. 2516 * Return 2517 * 0 on success, or a negative error in case of failure. 2518 * 2519 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2520 * Description 2521 * Pull in non-linear data in case the *skb* is non-linear and not 2522 * all of *len* are part of the linear section. Make *len* bytes 2523 * from *skb* readable and writable. If a zero value is passed for 2524 * *len*, then all bytes in the linear part of *skb* will be made 2525 * readable and writable. 2526 * 2527 * This helper is only needed for reading and writing with direct 2528 * packet access. 2529 * 2530 * For direct packet access, testing that offsets to access 2531 * are within packet boundaries (test on *skb*\ **->data_end**) is 2532 * susceptible to fail if offsets are invalid, or if the requested 2533 * data is in non-linear parts of the *skb*. On failure the 2534 * program can just bail out, or in the case of a non-linear 2535 * buffer, use a helper to make the data available. The 2536 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2537 * the data. Another one consists in using **bpf_skb_pull_data** 2538 * to pull in once the non-linear parts, then retesting and 2539 * eventually access the data. 2540 * 2541 * At the same time, this also makes sure the *skb* is uncloned, 2542 * which is a necessary condition for direct write. As this needs 2543 * to be an invariant for the write part only, the verifier 2544 * detects writes and adds a prologue that is calling 2545 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2546 * the very beginning in case it is indeed cloned. 2547 * 2548 * A call to this helper is susceptible to change the underlying 2549 * packet buffer. Therefore, at load time, all checks on pointers 2550 * previously done by the verifier are invalidated and must be 2551 * performed again, if the helper is used in combination with 2552 * direct packet access. 2553 * Return 2554 * 0 on success, or a negative error in case of failure. 2555 * 2556 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2557 * Description 2558 * Add the checksum *csum* into *skb*\ **->csum** in case the 2559 * driver has supplied a checksum for the entire packet into that 2560 * field. Return an error otherwise. This helper is intended to be 2561 * used in combination with **bpf_csum_diff**\ (), in particular 2562 * when the checksum needs to be updated after data has been 2563 * written into the packet through direct packet access. 2564 * Return 2565 * The checksum on success, or a negative error code in case of 2566 * failure. 2567 * 2568 * void bpf_set_hash_invalid(struct sk_buff *skb) 2569 * Description 2570 * Invalidate the current *skb*\ **->hash**. It can be used after 2571 * mangling on headers through direct packet access, in order to 2572 * indicate that the hash is outdated and to trigger a 2573 * recalculation the next time the kernel tries to access this 2574 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2575 * Return 2576 * void. 2577 * 2578 * long bpf_get_numa_node_id(void) 2579 * Description 2580 * Return the id of the current NUMA node. The primary use case 2581 * for this helper is the selection of sockets for the local NUMA 2582 * node, when the program is attached to sockets using the 2583 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2584 * but the helper is also available to other eBPF program types, 2585 * similarly to **bpf_get_smp_processor_id**\ (). 2586 * Return 2587 * The id of current NUMA node. 2588 * 2589 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2590 * Description 2591 * Grows headroom of packet associated to *skb* and adjusts the 2592 * offset of the MAC header accordingly, adding *len* bytes of 2593 * space. It automatically extends and reallocates memory as 2594 * required. 2595 * 2596 * This helper can be used on a layer 3 *skb* to push a MAC header 2597 * for redirection into a layer 2 device. 2598 * 2599 * All values for *flags* are reserved for future usage, and must 2600 * be left at zero. 2601 * 2602 * A call to this helper is susceptible to change the underlying 2603 * packet buffer. Therefore, at load time, all checks on pointers 2604 * previously done by the verifier are invalidated and must be 2605 * performed again, if the helper is used in combination with 2606 * direct packet access. 2607 * Return 2608 * 0 on success, or a negative error in case of failure. 2609 * 2610 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2611 * Description 2612 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2613 * it is possible to use a negative value for *delta*. This helper 2614 * can be used to prepare the packet for pushing or popping 2615 * headers. 2616 * 2617 * A call to this helper is susceptible to change the underlying 2618 * packet buffer. Therefore, at load time, all checks on pointers 2619 * previously done by the verifier are invalidated and must be 2620 * performed again, if the helper is used in combination with 2621 * direct packet access. 2622 * Return 2623 * 0 on success, or a negative error in case of failure. 2624 * 2625 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2626 * Description 2627 * Copy a NUL terminated string from an unsafe kernel address 2628 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2629 * more details. 2630 * 2631 * Generally, use **bpf_probe_read_user_str**\ () or 2632 * **bpf_probe_read_kernel_str**\ () instead. 2633 * Return 2634 * On success, the strictly positive length of the string, 2635 * including the trailing NUL character. On error, a negative 2636 * value. 2637 * 2638 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2639 * Description 2640 * If the **struct sk_buff** pointed by *skb* has a known socket, 2641 * retrieve the cookie (generated by the kernel) of this socket. 2642 * If no cookie has been set yet, generate a new cookie. Once 2643 * generated, the socket cookie remains stable for the life of the 2644 * socket. This helper can be useful for monitoring per socket 2645 * networking traffic statistics as it provides a global socket 2646 * identifier that can be assumed unique. 2647 * Return 2648 * A 8-byte long unique number on success, or 0 if the socket 2649 * field is missing inside *skb*. 2650 * 2651 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2652 * Description 2653 * Equivalent to bpf_get_socket_cookie() helper that accepts 2654 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2655 * Return 2656 * A 8-byte long unique number. 2657 * 2658 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2659 * Description 2660 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2661 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2662 * Return 2663 * A 8-byte long unique number. 2664 * 2665 * u64 bpf_get_socket_cookie(struct sock *sk) 2666 * Description 2667 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2668 * *sk*, but gets socket from a BTF **struct sock**. This helper 2669 * also works for sleepable programs. 2670 * Return 2671 * A 8-byte long unique number or 0 if *sk* is NULL. 2672 * 2673 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2674 * Description 2675 * Get the owner UID of the socked associated to *skb*. 2676 * Return 2677 * The owner UID of the socket associated to *skb*. If the socket 2678 * is **NULL**, or if it is not a full socket (i.e. if it is a 2679 * time-wait or a request socket instead), **overflowuid** value 2680 * is returned (note that **overflowuid** might also be the actual 2681 * UID value for the socket). 2682 * 2683 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2684 * Description 2685 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2686 * to value *hash*. 2687 * Return 2688 * 0 2689 * 2690 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2691 * Description 2692 * Emulate a call to **setsockopt()** on the socket associated to 2693 * *bpf_socket*, which must be a full socket. The *level* at 2694 * which the option resides and the name *optname* of the option 2695 * must be specified, see **setsockopt(2)** for more information. 2696 * The option value of length *optlen* is pointed by *optval*. 2697 * 2698 * *bpf_socket* should be one of the following: 2699 * 2700 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2701 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2702 * and **BPF_CGROUP_INET6_CONNECT**. 2703 * 2704 * This helper actually implements a subset of **setsockopt()**. 2705 * It supports the following *level*\ s: 2706 * 2707 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2708 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2709 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2710 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2711 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2712 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2713 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2714 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2715 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2716 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2717 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2718 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2719 * **TCP_BPF_RTO_MIN**. 2720 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2721 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2722 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2723 * Return 2724 * 0 on success, or a negative error in case of failure. 2725 * 2726 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2727 * Description 2728 * Grow or shrink the room for data in the packet associated to 2729 * *skb* by *len_diff*, and according to the selected *mode*. 2730 * 2731 * By default, the helper will reset any offloaded checksum 2732 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2733 * by the following flag: 2734 * 2735 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2736 * checksum data of the skb to CHECKSUM_NONE. 2737 * 2738 * There are two supported modes at this time: 2739 * 2740 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2741 * (room space is added or removed between the layer 2 and 2742 * layer 3 headers). 2743 * 2744 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2745 * (room space is added or removed between the layer 3 and 2746 * layer 4 headers). 2747 * 2748 * The following flags are supported at this time: 2749 * 2750 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2751 * Adjusting mss in this way is not allowed for datagrams. 2752 * 2753 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2754 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2755 * Any new space is reserved to hold a tunnel header. 2756 * Configure skb offsets and other fields accordingly. 2757 * 2758 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2759 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2760 * Use with ENCAP_L3 flags to further specify the tunnel type. 2761 * 2762 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2763 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2764 * type; *len* is the length of the inner MAC header. 2765 * 2766 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2767 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2768 * L2 type as Ethernet. 2769 * 2770 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2771 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2772 * Indicate the new IP header version after decapsulating the outer 2773 * IP header. Used when the inner and outer IP versions are different. 2774 * 2775 * A call to this helper is susceptible to change the underlying 2776 * packet buffer. Therefore, at load time, all checks on pointers 2777 * previously done by the verifier are invalidated and must be 2778 * performed again, if the helper is used in combination with 2779 * direct packet access. 2780 * Return 2781 * 0 on success, or a negative error in case of failure. 2782 * 2783 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2784 * Description 2785 * Redirect the packet to the endpoint referenced by *map* at 2786 * index *key*. Depending on its type, this *map* can contain 2787 * references to net devices (for forwarding packets through other 2788 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2789 * but this is only implemented for native XDP (with driver 2790 * support) as of this writing). 2791 * 2792 * The lower two bits of *flags* are used as the return code if 2793 * the map lookup fails. This is so that the return value can be 2794 * one of the XDP program return codes up to **XDP_TX**, as chosen 2795 * by the caller. The higher bits of *flags* can be set to 2796 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2797 * 2798 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2799 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2800 * interface will be excluded when do broadcasting. 2801 * 2802 * See also **bpf_redirect**\ (), which only supports redirecting 2803 * to an ifindex, but doesn't require a map to do so. 2804 * Return 2805 * **XDP_REDIRECT** on success, or the value of the two lower bits 2806 * of the *flags* argument on error. 2807 * 2808 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2809 * Description 2810 * Redirect the packet to the socket referenced by *map* (of type 2811 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2812 * egress interfaces can be used for redirection. The 2813 * **BPF_F_INGRESS** value in *flags* is used to make the 2814 * distinction (ingress path is selected if the flag is present, 2815 * egress path otherwise). This is the only flag supported for now. 2816 * Return 2817 * **SK_PASS** on success, or **SK_DROP** on error. 2818 * 2819 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2820 * Description 2821 * Add an entry to, or update a *map* referencing sockets. The 2822 * *skops* is used as a new value for the entry associated to 2823 * *key*. *flags* is one of: 2824 * 2825 * **BPF_NOEXIST** 2826 * The entry for *key* must not exist in the map. 2827 * **BPF_EXIST** 2828 * The entry for *key* must already exist in the map. 2829 * **BPF_ANY** 2830 * No condition on the existence of the entry for *key*. 2831 * 2832 * If the *map* has eBPF programs (parser and verdict), those will 2833 * be inherited by the socket being added. If the socket is 2834 * already attached to eBPF programs, this results in an error. 2835 * Return 2836 * 0 on success, or a negative error in case of failure. 2837 * 2838 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2839 * Description 2840 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2841 * *delta* (which can be positive or negative). Note that this 2842 * operation modifies the address stored in *xdp_md*\ **->data**, 2843 * so the latter must be loaded only after the helper has been 2844 * called. 2845 * 2846 * The use of *xdp_md*\ **->data_meta** is optional and programs 2847 * are not required to use it. The rationale is that when the 2848 * packet is processed with XDP (e.g. as DoS filter), it is 2849 * possible to push further meta data along with it before passing 2850 * to the stack, and to give the guarantee that an ingress eBPF 2851 * program attached as a TC classifier on the same device can pick 2852 * this up for further post-processing. Since TC works with socket 2853 * buffers, it remains possible to set from XDP the **mark** or 2854 * **priority** pointers, or other pointers for the socket buffer. 2855 * Having this scratch space generic and programmable allows for 2856 * more flexibility as the user is free to store whatever meta 2857 * data they need. 2858 * 2859 * A call to this helper is susceptible to change the underlying 2860 * packet buffer. Therefore, at load time, all checks on pointers 2861 * previously done by the verifier are invalidated and must be 2862 * performed again, if the helper is used in combination with 2863 * direct packet access. 2864 * Return 2865 * 0 on success, or a negative error in case of failure. 2866 * 2867 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2868 * Description 2869 * Read the value of a perf event counter, and store it into *buf* 2870 * of size *buf_size*. This helper relies on a *map* of type 2871 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2872 * counter is selected when *map* is updated with perf event file 2873 * descriptors. The *map* is an array whose size is the number of 2874 * available CPUs, and each cell contains a value relative to one 2875 * CPU. The value to retrieve is indicated by *flags*, that 2876 * contains the index of the CPU to look up, masked with 2877 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2878 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2879 * current CPU should be retrieved. 2880 * 2881 * This helper behaves in a way close to 2882 * **bpf_perf_event_read**\ () helper, save that instead of 2883 * just returning the value observed, it fills the *buf* 2884 * structure. This allows for additional data to be retrieved: in 2885 * particular, the enabled and running times (in *buf*\ 2886 * **->enabled** and *buf*\ **->running**, respectively) are 2887 * copied. In general, **bpf_perf_event_read_value**\ () is 2888 * recommended over **bpf_perf_event_read**\ (), which has some 2889 * ABI issues and provides fewer functionalities. 2890 * 2891 * These values are interesting, because hardware PMU (Performance 2892 * Monitoring Unit) counters are limited resources. When there are 2893 * more PMU based perf events opened than available counters, 2894 * kernel will multiplex these events so each event gets certain 2895 * percentage (but not all) of the PMU time. In case that 2896 * multiplexing happens, the number of samples or counter value 2897 * will not reflect the case compared to when no multiplexing 2898 * occurs. This makes comparison between different runs difficult. 2899 * Typically, the counter value should be normalized before 2900 * comparing to other experiments. The usual normalization is done 2901 * as follows. 2902 * 2903 * :: 2904 * 2905 * normalized_counter = counter * t_enabled / t_running 2906 * 2907 * Where t_enabled is the time enabled for event and t_running is 2908 * the time running for event since last normalization. The 2909 * enabled and running times are accumulated since the perf event 2910 * open. To achieve scaling factor between two invocations of an 2911 * eBPF program, users can use CPU id as the key (which is 2912 * typical for perf array usage model) to remember the previous 2913 * value and do the calculation inside the eBPF program. 2914 * Return 2915 * 0 on success, or a negative error in case of failure. 2916 * 2917 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2918 * Description 2919 * For an eBPF program attached to a perf event, retrieve the 2920 * value of the event counter associated to *ctx* and store it in 2921 * the structure pointed by *buf* and of size *buf_size*. Enabled 2922 * and running times are also stored in the structure (see 2923 * description of helper **bpf_perf_event_read_value**\ () for 2924 * more details). 2925 * Return 2926 * 0 on success, or a negative error in case of failure. 2927 * 2928 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2929 * Description 2930 * Emulate a call to **getsockopt()** on the socket associated to 2931 * *bpf_socket*, which must be a full socket. The *level* at 2932 * which the option resides and the name *optname* of the option 2933 * must be specified, see **getsockopt(2)** for more information. 2934 * The retrieved value is stored in the structure pointed by 2935 * *opval* and of length *optlen*. 2936 * 2937 * *bpf_socket* should be one of the following: 2938 * 2939 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2940 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2941 * and **BPF_CGROUP_INET6_CONNECT**. 2942 * 2943 * This helper actually implements a subset of **getsockopt()**. 2944 * It supports the same set of *optname*\ s that is supported by 2945 * the **bpf_setsockopt**\ () helper. The exceptions are 2946 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 2947 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 2948 * Return 2949 * 0 on success, or a negative error in case of failure. 2950 * 2951 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2952 * Description 2953 * Used for error injection, this helper uses kprobes to override 2954 * the return value of the probed function, and to set it to *rc*. 2955 * The first argument is the context *regs* on which the kprobe 2956 * works. 2957 * 2958 * This helper works by setting the PC (program counter) 2959 * to an override function which is run in place of the original 2960 * probed function. This means the probed function is not run at 2961 * all. The replacement function just returns with the required 2962 * value. 2963 * 2964 * This helper has security implications, and thus is subject to 2965 * restrictions. It is only available if the kernel was compiled 2966 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2967 * option, and in this case it only works on functions tagged with 2968 * **ALLOW_ERROR_INJECTION** in the kernel code. 2969 * 2970 * Also, the helper is only available for the architectures having 2971 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2972 * x86 architecture is the only one to support this feature. 2973 * Return 2974 * 0 2975 * 2976 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2977 * Description 2978 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2979 * for the full TCP socket associated to *bpf_sock_ops* to 2980 * *argval*. 2981 * 2982 * The primary use of this field is to determine if there should 2983 * be calls to eBPF programs of type 2984 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2985 * code. A program of the same type can change its value, per 2986 * connection and as necessary, when the connection is 2987 * established. This field is directly accessible for reading, but 2988 * this helper must be used for updates in order to return an 2989 * error if an eBPF program tries to set a callback that is not 2990 * supported in the current kernel. 2991 * 2992 * *argval* is a flag array which can combine these flags: 2993 * 2994 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2995 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2996 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2997 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2998 * 2999 * Therefore, this function can be used to clear a callback flag by 3000 * setting the appropriate bit to zero. e.g. to disable the RTO 3001 * callback: 3002 * 3003 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3004 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3005 * 3006 * Here are some examples of where one could call such eBPF 3007 * program: 3008 * 3009 * * When RTO fires. 3010 * * When a packet is retransmitted. 3011 * * When the connection terminates. 3012 * * When a packet is sent. 3013 * * When a packet is received. 3014 * Return 3015 * Code **-EINVAL** if the socket is not a full TCP socket; 3016 * otherwise, a positive number containing the bits that could not 3017 * be set is returned (which comes down to 0 if all bits were set 3018 * as required). 3019 * 3020 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3021 * Description 3022 * This helper is used in programs implementing policies at the 3023 * socket level. If the message *msg* is allowed to pass (i.e. if 3024 * the verdict eBPF program returns **SK_PASS**), redirect it to 3025 * the socket referenced by *map* (of type 3026 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3027 * egress interfaces can be used for redirection. The 3028 * **BPF_F_INGRESS** value in *flags* is used to make the 3029 * distinction (ingress path is selected if the flag is present, 3030 * egress path otherwise). This is the only flag supported for now. 3031 * Return 3032 * **SK_PASS** on success, or **SK_DROP** on error. 3033 * 3034 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3035 * Description 3036 * For socket policies, apply the verdict of the eBPF program to 3037 * the next *bytes* (number of bytes) of message *msg*. 3038 * 3039 * For example, this helper can be used in the following cases: 3040 * 3041 * * A single **sendmsg**\ () or **sendfile**\ () system call 3042 * contains multiple logical messages that the eBPF program is 3043 * supposed to read and for which it should apply a verdict. 3044 * * An eBPF program only cares to read the first *bytes* of a 3045 * *msg*. If the message has a large payload, then setting up 3046 * and calling the eBPF program repeatedly for all bytes, even 3047 * though the verdict is already known, would create unnecessary 3048 * overhead. 3049 * 3050 * When called from within an eBPF program, the helper sets a 3051 * counter internal to the BPF infrastructure, that is used to 3052 * apply the last verdict to the next *bytes*. If *bytes* is 3053 * smaller than the current data being processed from a 3054 * **sendmsg**\ () or **sendfile**\ () system call, the first 3055 * *bytes* will be sent and the eBPF program will be re-run with 3056 * the pointer for start of data pointing to byte number *bytes* 3057 * **+ 1**. If *bytes* is larger than the current data being 3058 * processed, then the eBPF verdict will be applied to multiple 3059 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3060 * consumed. 3061 * 3062 * Note that if a socket closes with the internal counter holding 3063 * a non-zero value, this is not a problem because data is not 3064 * being buffered for *bytes* and is sent as it is received. 3065 * Return 3066 * 0 3067 * 3068 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3069 * Description 3070 * For socket policies, prevent the execution of the verdict eBPF 3071 * program for message *msg* until *bytes* (byte number) have been 3072 * accumulated. 3073 * 3074 * This can be used when one needs a specific number of bytes 3075 * before a verdict can be assigned, even if the data spans 3076 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3077 * case would be a user calling **sendmsg**\ () repeatedly with 3078 * 1-byte long message segments. Obviously, this is bad for 3079 * performance, but it is still valid. If the eBPF program needs 3080 * *bytes* bytes to validate a header, this helper can be used to 3081 * prevent the eBPF program to be called again until *bytes* have 3082 * been accumulated. 3083 * Return 3084 * 0 3085 * 3086 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3087 * Description 3088 * For socket policies, pull in non-linear data from user space 3089 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3090 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3091 * respectively. 3092 * 3093 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3094 * *msg* it can only parse data that the (**data**, **data_end**) 3095 * pointers have already consumed. For **sendmsg**\ () hooks this 3096 * is likely the first scatterlist element. But for calls relying 3097 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3098 * be the range (**0**, **0**) because the data is shared with 3099 * user space and by default the objective is to avoid allowing 3100 * user space to modify data while (or after) eBPF verdict is 3101 * being decided. This helper can be used to pull in data and to 3102 * set the start and end pointer to given values. Data will be 3103 * copied if necessary (i.e. if data was not linear and if start 3104 * and end pointers do not point to the same chunk). 3105 * 3106 * A call to this helper is susceptible to change the underlying 3107 * packet buffer. Therefore, at load time, all checks on pointers 3108 * previously done by the verifier are invalidated and must be 3109 * performed again, if the helper is used in combination with 3110 * direct packet access. 3111 * 3112 * All values for *flags* are reserved for future usage, and must 3113 * be left at zero. 3114 * Return 3115 * 0 on success, or a negative error in case of failure. 3116 * 3117 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3118 * Description 3119 * Bind the socket associated to *ctx* to the address pointed by 3120 * *addr*, of length *addr_len*. This allows for making outgoing 3121 * connection from the desired IP address, which can be useful for 3122 * example when all processes inside a cgroup should use one 3123 * single IP address on a host that has multiple IP configured. 3124 * 3125 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3126 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3127 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3128 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3129 * behavior and lets the kernel efficiently pick up an unused 3130 * port as long as 4-tuple is unique. Passing non-zero port might 3131 * lead to degraded performance. 3132 * Return 3133 * 0 on success, or a negative error in case of failure. 3134 * 3135 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3136 * Description 3137 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3138 * possible to both shrink and grow the packet tail. 3139 * Shrink done via *delta* being a negative integer. 3140 * 3141 * A call to this helper is susceptible to change the underlying 3142 * packet buffer. Therefore, at load time, all checks on pointers 3143 * previously done by the verifier are invalidated and must be 3144 * performed again, if the helper is used in combination with 3145 * direct packet access. 3146 * Return 3147 * 0 on success, or a negative error in case of failure. 3148 * 3149 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3150 * Description 3151 * Retrieve the XFRM state (IP transform framework, see also 3152 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3153 * 3154 * The retrieved value is stored in the **struct bpf_xfrm_state** 3155 * pointed by *xfrm_state* and of length *size*. 3156 * 3157 * All values for *flags* are reserved for future usage, and must 3158 * be left at zero. 3159 * 3160 * This helper is available only if the kernel was compiled with 3161 * **CONFIG_XFRM** configuration option. 3162 * Return 3163 * 0 on success, or a negative error in case of failure. 3164 * 3165 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3166 * Description 3167 * Return a user or a kernel stack in bpf program provided buffer. 3168 * To achieve this, the helper needs *ctx*, which is a pointer 3169 * to the context on which the tracing program is executed. 3170 * To store the stacktrace, the bpf program provides *buf* with 3171 * a nonnegative *size*. 3172 * 3173 * The last argument, *flags*, holds the number of stack frames to 3174 * skip (from 0 to 255), masked with 3175 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3176 * the following flags: 3177 * 3178 * **BPF_F_USER_STACK** 3179 * Collect a user space stack instead of a kernel stack. 3180 * **BPF_F_USER_BUILD_ID** 3181 * Collect (build_id, file_offset) instead of ips for user 3182 * stack, only valid if **BPF_F_USER_STACK** is also 3183 * specified. 3184 * 3185 * *file_offset* is an offset relative to the beginning 3186 * of the executable or shared object file backing the vma 3187 * which the *ip* falls in. It is *not* an offset relative 3188 * to that object's base address. Accordingly, it must be 3189 * adjusted by adding (sh_addr - sh_offset), where 3190 * sh_{addr,offset} correspond to the executable section 3191 * containing *file_offset* in the object, for comparisons 3192 * to symbols' st_value to be valid. 3193 * 3194 * **bpf_get_stack**\ () can collect up to 3195 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3196 * to sufficient large buffer size. Note that 3197 * this limit can be controlled with the **sysctl** program, and 3198 * that it should be manually increased in order to profile long 3199 * user stacks (such as stacks for Java programs). To do so, use: 3200 * 3201 * :: 3202 * 3203 * # sysctl kernel.perf_event_max_stack=<new value> 3204 * Return 3205 * The non-negative copied *buf* length equal to or less than 3206 * *size* on success, or a negative error in case of failure. 3207 * 3208 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3209 * Description 3210 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3211 * it provides an easy way to load *len* bytes from *offset* 3212 * from the packet associated to *skb*, into the buffer pointed 3213 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3214 * a fifth argument *start_header* exists in order to select a 3215 * base offset to start from. *start_header* can be one of: 3216 * 3217 * **BPF_HDR_START_MAC** 3218 * Base offset to load data from is *skb*'s mac header. 3219 * **BPF_HDR_START_NET** 3220 * Base offset to load data from is *skb*'s network header. 3221 * 3222 * In general, "direct packet access" is the preferred method to 3223 * access packet data, however, this helper is in particular useful 3224 * in socket filters where *skb*\ **->data** does not always point 3225 * to the start of the mac header and where "direct packet access" 3226 * is not available. 3227 * Return 3228 * 0 on success, or a negative error in case of failure. 3229 * 3230 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3231 * Description 3232 * Do FIB lookup in kernel tables using parameters in *params*. 3233 * If lookup is successful and result shows packet is to be 3234 * forwarded, the neighbor tables are searched for the nexthop. 3235 * If successful (ie., FIB lookup shows forwarding and nexthop 3236 * is resolved), the nexthop address is returned in ipv4_dst 3237 * or ipv6_dst based on family, smac is set to mac address of 3238 * egress device, dmac is set to nexthop mac address, rt_metric 3239 * is set to metric from route (IPv4/IPv6 only), and ifindex 3240 * is set to the device index of the nexthop from the FIB lookup. 3241 * 3242 * *plen* argument is the size of the passed in struct. 3243 * *flags* argument can be a combination of one or more of the 3244 * following values: 3245 * 3246 * **BPF_FIB_LOOKUP_DIRECT** 3247 * Do a direct table lookup vs full lookup using FIB 3248 * rules. 3249 * **BPF_FIB_LOOKUP_TBID** 3250 * Used with BPF_FIB_LOOKUP_DIRECT. 3251 * Use the routing table ID present in *params*->tbid 3252 * for the fib lookup. 3253 * **BPF_FIB_LOOKUP_OUTPUT** 3254 * Perform lookup from an egress perspective (default is 3255 * ingress). 3256 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3257 * Skip the neighbour table lookup. *params*->dmac 3258 * and *params*->smac will not be set as output. A common 3259 * use case is to call **bpf_redirect_neigh**\ () after 3260 * doing **bpf_fib_lookup**\ (). 3261 * 3262 * *ctx* is either **struct xdp_md** for XDP programs or 3263 * **struct sk_buff** tc cls_act programs. 3264 * Return 3265 * * < 0 if any input argument is invalid 3266 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3267 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3268 * packet is not forwarded or needs assist from full stack 3269 * 3270 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3271 * was exceeded and output params->mtu_result contains the MTU. 3272 * 3273 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3274 * Description 3275 * Add an entry to, or update a sockhash *map* referencing sockets. 3276 * The *skops* is used as a new value for the entry associated to 3277 * *key*. *flags* is one of: 3278 * 3279 * **BPF_NOEXIST** 3280 * The entry for *key* must not exist in the map. 3281 * **BPF_EXIST** 3282 * The entry for *key* must already exist in the map. 3283 * **BPF_ANY** 3284 * No condition on the existence of the entry for *key*. 3285 * 3286 * If the *map* has eBPF programs (parser and verdict), those will 3287 * be inherited by the socket being added. If the socket is 3288 * already attached to eBPF programs, this results in an error. 3289 * Return 3290 * 0 on success, or a negative error in case of failure. 3291 * 3292 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3293 * Description 3294 * This helper is used in programs implementing policies at the 3295 * socket level. If the message *msg* is allowed to pass (i.e. if 3296 * the verdict eBPF program returns **SK_PASS**), redirect it to 3297 * the socket referenced by *map* (of type 3298 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3299 * egress interfaces can be used for redirection. The 3300 * **BPF_F_INGRESS** value in *flags* is used to make the 3301 * distinction (ingress path is selected if the flag is present, 3302 * egress path otherwise). This is the only flag supported for now. 3303 * Return 3304 * **SK_PASS** on success, or **SK_DROP** on error. 3305 * 3306 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3307 * Description 3308 * This helper is used in programs implementing policies at the 3309 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3310 * if the verdict eBPF program returns **SK_PASS**), redirect it 3311 * to the socket referenced by *map* (of type 3312 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3313 * egress interfaces can be used for redirection. The 3314 * **BPF_F_INGRESS** value in *flags* is used to make the 3315 * distinction (ingress path is selected if the flag is present, 3316 * egress otherwise). This is the only flag supported for now. 3317 * Return 3318 * **SK_PASS** on success, or **SK_DROP** on error. 3319 * 3320 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3321 * Description 3322 * Encapsulate the packet associated to *skb* within a Layer 3 3323 * protocol header. This header is provided in the buffer at 3324 * address *hdr*, with *len* its size in bytes. *type* indicates 3325 * the protocol of the header and can be one of: 3326 * 3327 * **BPF_LWT_ENCAP_SEG6** 3328 * IPv6 encapsulation with Segment Routing Header 3329 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3330 * the IPv6 header is computed by the kernel. 3331 * **BPF_LWT_ENCAP_SEG6_INLINE** 3332 * Only works if *skb* contains an IPv6 packet. Insert a 3333 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3334 * the IPv6 header. 3335 * **BPF_LWT_ENCAP_IP** 3336 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3337 * must be IPv4 or IPv6, followed by zero or more 3338 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3339 * total bytes in all prepended headers. Please note that 3340 * if **skb_is_gso**\ (*skb*) is true, no more than two 3341 * headers can be prepended, and the inner header, if 3342 * present, should be either GRE or UDP/GUE. 3343 * 3344 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3345 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3346 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3347 * **BPF_PROG_TYPE_LWT_XMIT**. 3348 * 3349 * A call to this helper is susceptible to change the underlying 3350 * packet buffer. Therefore, at load time, all checks on pointers 3351 * previously done by the verifier are invalidated and must be 3352 * performed again, if the helper is used in combination with 3353 * direct packet access. 3354 * Return 3355 * 0 on success, or a negative error in case of failure. 3356 * 3357 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3358 * Description 3359 * Store *len* bytes from address *from* into the packet 3360 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3361 * inside the outermost IPv6 Segment Routing Header can be 3362 * modified through this helper. 3363 * 3364 * A call to this helper is susceptible to change the underlying 3365 * packet buffer. Therefore, at load time, all checks on pointers 3366 * previously done by the verifier are invalidated and must be 3367 * performed again, if the helper is used in combination with 3368 * direct packet access. 3369 * Return 3370 * 0 on success, or a negative error in case of failure. 3371 * 3372 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3373 * Description 3374 * Adjust the size allocated to TLVs in the outermost IPv6 3375 * Segment Routing Header contained in the packet associated to 3376 * *skb*, at position *offset* by *delta* bytes. Only offsets 3377 * after the segments are accepted. *delta* can be as well 3378 * positive (growing) as negative (shrinking). 3379 * 3380 * A call to this helper is susceptible to change the underlying 3381 * packet buffer. Therefore, at load time, all checks on pointers 3382 * previously done by the verifier are invalidated and must be 3383 * performed again, if the helper is used in combination with 3384 * direct packet access. 3385 * Return 3386 * 0 on success, or a negative error in case of failure. 3387 * 3388 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3389 * Description 3390 * Apply an IPv6 Segment Routing action of type *action* to the 3391 * packet associated to *skb*. Each action takes a parameter 3392 * contained at address *param*, and of length *param_len* bytes. 3393 * *action* can be one of: 3394 * 3395 * **SEG6_LOCAL_ACTION_END_X** 3396 * End.X action: Endpoint with Layer-3 cross-connect. 3397 * Type of *param*: **struct in6_addr**. 3398 * **SEG6_LOCAL_ACTION_END_T** 3399 * End.T action: Endpoint with specific IPv6 table lookup. 3400 * Type of *param*: **int**. 3401 * **SEG6_LOCAL_ACTION_END_B6** 3402 * End.B6 action: Endpoint bound to an SRv6 policy. 3403 * Type of *param*: **struct ipv6_sr_hdr**. 3404 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3405 * End.B6.Encap action: Endpoint bound to an SRv6 3406 * encapsulation policy. 3407 * Type of *param*: **struct ipv6_sr_hdr**. 3408 * 3409 * A call to this helper is susceptible to change the underlying 3410 * packet buffer. Therefore, at load time, all checks on pointers 3411 * previously done by the verifier are invalidated and must be 3412 * performed again, if the helper is used in combination with 3413 * direct packet access. 3414 * Return 3415 * 0 on success, or a negative error in case of failure. 3416 * 3417 * long bpf_rc_repeat(void *ctx) 3418 * Description 3419 * This helper is used in programs implementing IR decoding, to 3420 * report a successfully decoded repeat key message. This delays 3421 * the generation of a key up event for previously generated 3422 * key down event. 3423 * 3424 * Some IR protocols like NEC have a special IR message for 3425 * repeating last button, for when a button is held down. 3426 * 3427 * The *ctx* should point to the lirc sample as passed into 3428 * the program. 3429 * 3430 * This helper is only available is the kernel was compiled with 3431 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3432 * "**y**". 3433 * Return 3434 * 0 3435 * 3436 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3437 * Description 3438 * This helper is used in programs implementing IR decoding, to 3439 * report a successfully decoded key press with *scancode*, 3440 * *toggle* value in the given *protocol*. The scancode will be 3441 * translated to a keycode using the rc keymap, and reported as 3442 * an input key down event. After a period a key up event is 3443 * generated. This period can be extended by calling either 3444 * **bpf_rc_keydown**\ () again with the same values, or calling 3445 * **bpf_rc_repeat**\ (). 3446 * 3447 * Some protocols include a toggle bit, in case the button was 3448 * released and pressed again between consecutive scancodes. 3449 * 3450 * The *ctx* should point to the lirc sample as passed into 3451 * the program. 3452 * 3453 * The *protocol* is the decoded protocol number (see 3454 * **enum rc_proto** for some predefined values). 3455 * 3456 * This helper is only available is the kernel was compiled with 3457 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3458 * "**y**". 3459 * Return 3460 * 0 3461 * 3462 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3463 * Description 3464 * Return the cgroup v2 id of the socket associated with the *skb*. 3465 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3466 * helper for cgroup v1 by providing a tag resp. identifier that 3467 * can be matched on or used for map lookups e.g. to implement 3468 * policy. The cgroup v2 id of a given path in the hierarchy is 3469 * exposed in user space through the f_handle API in order to get 3470 * to the same 64-bit id. 3471 * 3472 * This helper can be used on TC egress path, but not on ingress, 3473 * and is available only if the kernel was compiled with the 3474 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3475 * Return 3476 * The id is returned or 0 in case the id could not be retrieved. 3477 * 3478 * u64 bpf_get_current_cgroup_id(void) 3479 * Description 3480 * Get the current cgroup id based on the cgroup within which 3481 * the current task is running. 3482 * Return 3483 * A 64-bit integer containing the current cgroup id based 3484 * on the cgroup within which the current task is running. 3485 * 3486 * void *bpf_get_local_storage(void *map, u64 flags) 3487 * Description 3488 * Get the pointer to the local storage area. 3489 * The type and the size of the local storage is defined 3490 * by the *map* argument. 3491 * The *flags* meaning is specific for each map type, 3492 * and has to be 0 for cgroup local storage. 3493 * 3494 * Depending on the BPF program type, a local storage area 3495 * can be shared between multiple instances of the BPF program, 3496 * running simultaneously. 3497 * 3498 * A user should care about the synchronization by himself. 3499 * For example, by using the **BPF_ATOMIC** instructions to alter 3500 * the shared data. 3501 * Return 3502 * A pointer to the local storage area. 3503 * 3504 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3505 * Description 3506 * Select a **SO_REUSEPORT** socket from a 3507 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3508 * It checks the selected socket is matching the incoming 3509 * request in the socket buffer. 3510 * Return 3511 * 0 on success, or a negative error in case of failure. 3512 * 3513 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3514 * Description 3515 * Return id of cgroup v2 that is ancestor of cgroup associated 3516 * with the *skb* at the *ancestor_level*. The root cgroup is at 3517 * *ancestor_level* zero and each step down the hierarchy 3518 * increments the level. If *ancestor_level* == level of cgroup 3519 * associated with *skb*, then return value will be same as that 3520 * of **bpf_skb_cgroup_id**\ (). 3521 * 3522 * The helper is useful to implement policies based on cgroups 3523 * that are upper in hierarchy than immediate cgroup associated 3524 * with *skb*. 3525 * 3526 * The format of returned id and helper limitations are same as in 3527 * **bpf_skb_cgroup_id**\ (). 3528 * Return 3529 * The id is returned or 0 in case the id could not be retrieved. 3530 * 3531 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3532 * Description 3533 * Look for TCP socket matching *tuple*, optionally in a child 3534 * network namespace *netns*. The return value must be checked, 3535 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3536 * 3537 * The *ctx* should point to the context of the program, such as 3538 * the skb or socket (depending on the hook in use). This is used 3539 * to determine the base network namespace for the lookup. 3540 * 3541 * *tuple_size* must be one of: 3542 * 3543 * **sizeof**\ (*tuple*\ **->ipv4**) 3544 * Look for an IPv4 socket. 3545 * **sizeof**\ (*tuple*\ **->ipv6**) 3546 * Look for an IPv6 socket. 3547 * 3548 * If the *netns* is a negative signed 32-bit integer, then the 3549 * socket lookup table in the netns associated with the *ctx* 3550 * will be used. For the TC hooks, this is the netns of the device 3551 * in the skb. For socket hooks, this is the netns of the socket. 3552 * If *netns* is any other signed 32-bit value greater than or 3553 * equal to zero then it specifies the ID of the netns relative to 3554 * the netns associated with the *ctx*. *netns* values beyond the 3555 * range of 32-bit integers are reserved for future use. 3556 * 3557 * All values for *flags* are reserved for future usage, and must 3558 * be left at zero. 3559 * 3560 * This helper is available only if the kernel was compiled with 3561 * **CONFIG_NET** configuration option. 3562 * Return 3563 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3564 * For sockets with reuseport option, the **struct bpf_sock** 3565 * result is from *reuse*\ **->socks**\ [] using the hash of the 3566 * tuple. 3567 * 3568 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3569 * Description 3570 * Look for UDP socket matching *tuple*, optionally in a child 3571 * network namespace *netns*. The return value must be checked, 3572 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3573 * 3574 * The *ctx* should point to the context of the program, such as 3575 * the skb or socket (depending on the hook in use). This is used 3576 * to determine the base network namespace for the lookup. 3577 * 3578 * *tuple_size* must be one of: 3579 * 3580 * **sizeof**\ (*tuple*\ **->ipv4**) 3581 * Look for an IPv4 socket. 3582 * **sizeof**\ (*tuple*\ **->ipv6**) 3583 * Look for an IPv6 socket. 3584 * 3585 * If the *netns* is a negative signed 32-bit integer, then the 3586 * socket lookup table in the netns associated with the *ctx* 3587 * will be used. For the TC hooks, this is the netns of the device 3588 * in the skb. For socket hooks, this is the netns of the socket. 3589 * If *netns* is any other signed 32-bit value greater than or 3590 * equal to zero then it specifies the ID of the netns relative to 3591 * the netns associated with the *ctx*. *netns* values beyond the 3592 * range of 32-bit integers are reserved for future use. 3593 * 3594 * All values for *flags* are reserved for future usage, and must 3595 * be left at zero. 3596 * 3597 * This helper is available only if the kernel was compiled with 3598 * **CONFIG_NET** configuration option. 3599 * Return 3600 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3601 * For sockets with reuseport option, the **struct bpf_sock** 3602 * result is from *reuse*\ **->socks**\ [] using the hash of the 3603 * tuple. 3604 * 3605 * long bpf_sk_release(void *sock) 3606 * Description 3607 * Release the reference held by *sock*. *sock* must be a 3608 * non-**NULL** pointer that was returned from 3609 * **bpf_sk_lookup_xxx**\ (). 3610 * Return 3611 * 0 on success, or a negative error in case of failure. 3612 * 3613 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3614 * Description 3615 * Push an element *value* in *map*. *flags* is one of: 3616 * 3617 * **BPF_EXIST** 3618 * If the queue/stack is full, the oldest element is 3619 * removed to make room for this. 3620 * Return 3621 * 0 on success, or a negative error in case of failure. 3622 * 3623 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3624 * Description 3625 * Pop an element from *map*. 3626 * Return 3627 * 0 on success, or a negative error in case of failure. 3628 * 3629 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3630 * Description 3631 * Get an element from *map* without removing it. 3632 * Return 3633 * 0 on success, or a negative error in case of failure. 3634 * 3635 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3636 * Description 3637 * For socket policies, insert *len* bytes into *msg* at offset 3638 * *start*. 3639 * 3640 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3641 * *msg* it may want to insert metadata or options into the *msg*. 3642 * This can later be read and used by any of the lower layer BPF 3643 * hooks. 3644 * 3645 * This helper may fail if under memory pressure (a malloc 3646 * fails) in these cases BPF programs will get an appropriate 3647 * error and BPF programs will need to handle them. 3648 * Return 3649 * 0 on success, or a negative error in case of failure. 3650 * 3651 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3652 * Description 3653 * Will remove *len* bytes from a *msg* starting at byte *start*. 3654 * This may result in **ENOMEM** errors under certain situations if 3655 * an allocation and copy are required due to a full ring buffer. 3656 * However, the helper will try to avoid doing the allocation 3657 * if possible. Other errors can occur if input parameters are 3658 * invalid either due to *start* byte not being valid part of *msg* 3659 * payload and/or *pop* value being to large. 3660 * Return 3661 * 0 on success, or a negative error in case of failure. 3662 * 3663 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3664 * Description 3665 * This helper is used in programs implementing IR decoding, to 3666 * report a successfully decoded pointer movement. 3667 * 3668 * The *ctx* should point to the lirc sample as passed into 3669 * the program. 3670 * 3671 * This helper is only available is the kernel was compiled with 3672 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3673 * "**y**". 3674 * Return 3675 * 0 3676 * 3677 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3678 * Description 3679 * Acquire a spinlock represented by the pointer *lock*, which is 3680 * stored as part of a value of a map. Taking the lock allows to 3681 * safely update the rest of the fields in that value. The 3682 * spinlock can (and must) later be released with a call to 3683 * **bpf_spin_unlock**\ (\ *lock*\ ). 3684 * 3685 * Spinlocks in BPF programs come with a number of restrictions 3686 * and constraints: 3687 * 3688 * * **bpf_spin_lock** objects are only allowed inside maps of 3689 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3690 * list could be extended in the future). 3691 * * BTF description of the map is mandatory. 3692 * * The BPF program can take ONE lock at a time, since taking two 3693 * or more could cause dead locks. 3694 * * Only one **struct bpf_spin_lock** is allowed per map element. 3695 * * When the lock is taken, calls (either BPF to BPF or helpers) 3696 * are not allowed. 3697 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3698 * allowed inside a spinlock-ed region. 3699 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3700 * the lock, on all execution paths, before it returns. 3701 * * The BPF program can access **struct bpf_spin_lock** only via 3702 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3703 * helpers. Loading or storing data into the **struct 3704 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3705 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3706 * of the map value must be a struct and have **struct 3707 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3708 * Nested lock inside another struct is not allowed. 3709 * * The **struct bpf_spin_lock** *lock* field in a map value must 3710 * be aligned on a multiple of 4 bytes in that value. 3711 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3712 * the **bpf_spin_lock** field to user space. 3713 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3714 * a BPF program, do not update the **bpf_spin_lock** field. 3715 * * **bpf_spin_lock** cannot be on the stack or inside a 3716 * networking packet (it can only be inside of a map values). 3717 * * **bpf_spin_lock** is available to root only. 3718 * * Tracing programs and socket filter programs cannot use 3719 * **bpf_spin_lock**\ () due to insufficient preemption checks 3720 * (but this may change in the future). 3721 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3722 * Return 3723 * 0 3724 * 3725 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3726 * Description 3727 * Release the *lock* previously locked by a call to 3728 * **bpf_spin_lock**\ (\ *lock*\ ). 3729 * Return 3730 * 0 3731 * 3732 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3733 * Description 3734 * This helper gets a **struct bpf_sock** pointer such 3735 * that all the fields in this **bpf_sock** can be accessed. 3736 * Return 3737 * A **struct bpf_sock** pointer on success, or **NULL** in 3738 * case of failure. 3739 * 3740 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3741 * Description 3742 * This helper gets a **struct bpf_tcp_sock** pointer from a 3743 * **struct bpf_sock** pointer. 3744 * Return 3745 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3746 * case of failure. 3747 * 3748 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3749 * Description 3750 * Set ECN (Explicit Congestion Notification) field of IP header 3751 * to **CE** (Congestion Encountered) if current value is **ECT** 3752 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3753 * and IPv4. 3754 * Return 3755 * 1 if the **CE** flag is set (either by the current helper call 3756 * or because it was already present), 0 if it is not set. 3757 * 3758 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3759 * Description 3760 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3761 * **bpf_sk_release**\ () is unnecessary and not allowed. 3762 * Return 3763 * A **struct bpf_sock** pointer on success, or **NULL** in 3764 * case of failure. 3765 * 3766 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3767 * Description 3768 * Look for TCP socket matching *tuple*, optionally in a child 3769 * network namespace *netns*. The return value must be checked, 3770 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3771 * 3772 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3773 * that it also returns timewait or request sockets. Use 3774 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3775 * full structure. 3776 * 3777 * This helper is available only if the kernel was compiled with 3778 * **CONFIG_NET** configuration option. 3779 * Return 3780 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3781 * For sockets with reuseport option, the **struct bpf_sock** 3782 * result is from *reuse*\ **->socks**\ [] using the hash of the 3783 * tuple. 3784 * 3785 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3786 * Description 3787 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3788 * the listening socket in *sk*. 3789 * 3790 * *iph* points to the start of the IPv4 or IPv6 header, while 3791 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3792 * **sizeof**\ (**struct ipv6hdr**). 3793 * 3794 * *th* points to the start of the TCP header, while *th_len* 3795 * contains the length of the TCP header (at least 3796 * **sizeof**\ (**struct tcphdr**)). 3797 * Return 3798 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3799 * error otherwise. 3800 * 3801 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3802 * Description 3803 * Get name of sysctl in /proc/sys/ and copy it into provided by 3804 * program buffer *buf* of size *buf_len*. 3805 * 3806 * The buffer is always NUL terminated, unless it's zero-sized. 3807 * 3808 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3809 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3810 * only (e.g. "tcp_mem"). 3811 * Return 3812 * Number of character copied (not including the trailing NUL). 3813 * 3814 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3815 * truncated name in this case). 3816 * 3817 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3818 * Description 3819 * Get current value of sysctl as it is presented in /proc/sys 3820 * (incl. newline, etc), and copy it as a string into provided 3821 * by program buffer *buf* of size *buf_len*. 3822 * 3823 * The whole value is copied, no matter what file position user 3824 * space issued e.g. sys_read at. 3825 * 3826 * The buffer is always NUL terminated, unless it's zero-sized. 3827 * Return 3828 * Number of character copied (not including the trailing NUL). 3829 * 3830 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3831 * truncated name in this case). 3832 * 3833 * **-EINVAL** if current value was unavailable, e.g. because 3834 * sysctl is uninitialized and read returns -EIO for it. 3835 * 3836 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3837 * Description 3838 * Get new value being written by user space to sysctl (before 3839 * the actual write happens) and copy it as a string into 3840 * provided by program buffer *buf* of size *buf_len*. 3841 * 3842 * User space may write new value at file position > 0. 3843 * 3844 * The buffer is always NUL terminated, unless it's zero-sized. 3845 * Return 3846 * Number of character copied (not including the trailing NUL). 3847 * 3848 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3849 * truncated name in this case). 3850 * 3851 * **-EINVAL** if sysctl is being read. 3852 * 3853 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3854 * Description 3855 * Override new value being written by user space to sysctl with 3856 * value provided by program in buffer *buf* of size *buf_len*. 3857 * 3858 * *buf* should contain a string in same form as provided by user 3859 * space on sysctl write. 3860 * 3861 * User space may write new value at file position > 0. To override 3862 * the whole sysctl value file position should be set to zero. 3863 * Return 3864 * 0 on success. 3865 * 3866 * **-E2BIG** if the *buf_len* is too big. 3867 * 3868 * **-EINVAL** if sysctl is being read. 3869 * 3870 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3871 * Description 3872 * Convert the initial part of the string from buffer *buf* of 3873 * size *buf_len* to a long integer according to the given base 3874 * and save the result in *res*. 3875 * 3876 * The string may begin with an arbitrary amount of white space 3877 * (as determined by **isspace**\ (3)) followed by a single 3878 * optional '**-**' sign. 3879 * 3880 * Five least significant bits of *flags* encode base, other bits 3881 * are currently unused. 3882 * 3883 * Base must be either 8, 10, 16 or 0 to detect it automatically 3884 * similar to user space **strtol**\ (3). 3885 * Return 3886 * Number of characters consumed on success. Must be positive but 3887 * no more than *buf_len*. 3888 * 3889 * **-EINVAL** if no valid digits were found or unsupported base 3890 * was provided. 3891 * 3892 * **-ERANGE** if resulting value was out of range. 3893 * 3894 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3895 * Description 3896 * Convert the initial part of the string from buffer *buf* of 3897 * size *buf_len* to an unsigned long integer according to the 3898 * given base and save the result in *res*. 3899 * 3900 * The string may begin with an arbitrary amount of white space 3901 * (as determined by **isspace**\ (3)). 3902 * 3903 * Five least significant bits of *flags* encode base, other bits 3904 * are currently unused. 3905 * 3906 * Base must be either 8, 10, 16 or 0 to detect it automatically 3907 * similar to user space **strtoul**\ (3). 3908 * Return 3909 * Number of characters consumed on success. Must be positive but 3910 * no more than *buf_len*. 3911 * 3912 * **-EINVAL** if no valid digits were found or unsupported base 3913 * was provided. 3914 * 3915 * **-ERANGE** if resulting value was out of range. 3916 * 3917 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3918 * Description 3919 * Get a bpf-local-storage from a *sk*. 3920 * 3921 * Logically, it could be thought of getting the value from 3922 * a *map* with *sk* as the **key**. From this 3923 * perspective, the usage is not much different from 3924 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3925 * helper enforces the key must be a full socket and the map must 3926 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3927 * 3928 * Underneath, the value is stored locally at *sk* instead of 3929 * the *map*. The *map* is used as the bpf-local-storage 3930 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3931 * searched against all bpf-local-storages residing at *sk*. 3932 * 3933 * *sk* is a kernel **struct sock** pointer for LSM program. 3934 * *sk* is a **struct bpf_sock** pointer for other program types. 3935 * 3936 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3937 * used such that a new bpf-local-storage will be 3938 * created if one does not exist. *value* can be used 3939 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3940 * the initial value of a bpf-local-storage. If *value* is 3941 * **NULL**, the new bpf-local-storage will be zero initialized. 3942 * Return 3943 * A bpf-local-storage pointer is returned on success. 3944 * 3945 * **NULL** if not found or there was an error in adding 3946 * a new bpf-local-storage. 3947 * 3948 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3949 * Description 3950 * Delete a bpf-local-storage from a *sk*. 3951 * Return 3952 * 0 on success. 3953 * 3954 * **-ENOENT** if the bpf-local-storage cannot be found. 3955 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3956 * 3957 * long bpf_send_signal(u32 sig) 3958 * Description 3959 * Send signal *sig* to the process of the current task. 3960 * The signal may be delivered to any of this process's threads. 3961 * Return 3962 * 0 on success or successfully queued. 3963 * 3964 * **-EBUSY** if work queue under nmi is full. 3965 * 3966 * **-EINVAL** if *sig* is invalid. 3967 * 3968 * **-EPERM** if no permission to send the *sig*. 3969 * 3970 * **-EAGAIN** if bpf program can try again. 3971 * 3972 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3973 * Description 3974 * Try to issue a SYN cookie for the packet with corresponding 3975 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3976 * 3977 * *iph* points to the start of the IPv4 or IPv6 header, while 3978 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3979 * **sizeof**\ (**struct ipv6hdr**). 3980 * 3981 * *th* points to the start of the TCP header, while *th_len* 3982 * contains the length of the TCP header with options (at least 3983 * **sizeof**\ (**struct tcphdr**)). 3984 * Return 3985 * On success, lower 32 bits hold the generated SYN cookie in 3986 * followed by 16 bits which hold the MSS value for that cookie, 3987 * and the top 16 bits are unused. 3988 * 3989 * On failure, the returned value is one of the following: 3990 * 3991 * **-EINVAL** SYN cookie cannot be issued due to error 3992 * 3993 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3994 * 3995 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3996 * 3997 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3998 * 3999 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4000 * Description 4001 * Write raw *data* blob into a special BPF perf event held by 4002 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4003 * event must have the following attributes: **PERF_SAMPLE_RAW** 4004 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4005 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4006 * 4007 * The *flags* are used to indicate the index in *map* for which 4008 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4009 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4010 * to indicate that the index of the current CPU core should be 4011 * used. 4012 * 4013 * The value to write, of *size*, is passed through eBPF stack and 4014 * pointed by *data*. 4015 * 4016 * *ctx* is a pointer to in-kernel struct sk_buff. 4017 * 4018 * This helper is similar to **bpf_perf_event_output**\ () but 4019 * restricted to raw_tracepoint bpf programs. 4020 * Return 4021 * 0 on success, or a negative error in case of failure. 4022 * 4023 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4024 * Description 4025 * Safely attempt to read *size* bytes from user space address 4026 * *unsafe_ptr* and store the data in *dst*. 4027 * Return 4028 * 0 on success, or a negative error in case of failure. 4029 * 4030 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4031 * Description 4032 * Safely attempt to read *size* bytes from kernel space address 4033 * *unsafe_ptr* and store the data in *dst*. 4034 * Return 4035 * 0 on success, or a negative error in case of failure. 4036 * 4037 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4038 * Description 4039 * Copy a NUL terminated string from an unsafe user address 4040 * *unsafe_ptr* to *dst*. The *size* should include the 4041 * terminating NUL byte. In case the string length is smaller than 4042 * *size*, the target is not padded with further NUL bytes. If the 4043 * string length is larger than *size*, just *size*-1 bytes are 4044 * copied and the last byte is set to NUL. 4045 * 4046 * On success, returns the number of bytes that were written, 4047 * including the terminal NUL. This makes this helper useful in 4048 * tracing programs for reading strings, and more importantly to 4049 * get its length at runtime. See the following snippet: 4050 * 4051 * :: 4052 * 4053 * SEC("kprobe/sys_open") 4054 * void bpf_sys_open(struct pt_regs *ctx) 4055 * { 4056 * char buf[PATHLEN]; // PATHLEN is defined to 256 4057 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4058 * ctx->di); 4059 * 4060 * // Consume buf, for example push it to 4061 * // userspace via bpf_perf_event_output(); we 4062 * // can use res (the string length) as event 4063 * // size, after checking its boundaries. 4064 * } 4065 * 4066 * In comparison, using **bpf_probe_read_user**\ () helper here 4067 * instead to read the string would require to estimate the length 4068 * at compile time, and would often result in copying more memory 4069 * than necessary. 4070 * 4071 * Another useful use case is when parsing individual process 4072 * arguments or individual environment variables navigating 4073 * *current*\ **->mm->arg_start** and *current*\ 4074 * **->mm->env_start**: using this helper and the return value, 4075 * one can quickly iterate at the right offset of the memory area. 4076 * Return 4077 * On success, the strictly positive length of the output string, 4078 * including the trailing NUL character. On error, a negative 4079 * value. 4080 * 4081 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4082 * Description 4083 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4084 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4085 * Return 4086 * On success, the strictly positive length of the string, including 4087 * the trailing NUL character. On error, a negative value. 4088 * 4089 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4090 * Description 4091 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4092 * *rcv_nxt* is the ack_seq to be sent out. 4093 * Return 4094 * 0 on success, or a negative error in case of failure. 4095 * 4096 * long bpf_send_signal_thread(u32 sig) 4097 * Description 4098 * Send signal *sig* to the thread corresponding to the current task. 4099 * Return 4100 * 0 on success or successfully queued. 4101 * 4102 * **-EBUSY** if work queue under nmi is full. 4103 * 4104 * **-EINVAL** if *sig* is invalid. 4105 * 4106 * **-EPERM** if no permission to send the *sig*. 4107 * 4108 * **-EAGAIN** if bpf program can try again. 4109 * 4110 * u64 bpf_jiffies64(void) 4111 * Description 4112 * Obtain the 64bit jiffies 4113 * Return 4114 * The 64 bit jiffies 4115 * 4116 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4117 * Description 4118 * For an eBPF program attached to a perf event, retrieve the 4119 * branch records (**struct perf_branch_entry**) associated to *ctx* 4120 * and store it in the buffer pointed by *buf* up to size 4121 * *size* bytes. 4122 * Return 4123 * On success, number of bytes written to *buf*. On error, a 4124 * negative value. 4125 * 4126 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4127 * instead return the number of bytes required to store all the 4128 * branch entries. If this flag is set, *buf* may be NULL. 4129 * 4130 * **-EINVAL** if arguments invalid or **size** not a multiple 4131 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4132 * 4133 * **-ENOENT** if architecture does not support branch records. 4134 * 4135 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4136 * Description 4137 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4138 * *namespace* will be returned in *nsdata*. 4139 * Return 4140 * 0 on success, or one of the following in case of failure: 4141 * 4142 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4143 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4144 * 4145 * **-ENOENT** if pidns does not exists for the current task. 4146 * 4147 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4148 * Description 4149 * Write raw *data* blob into a special BPF perf event held by 4150 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4151 * event must have the following attributes: **PERF_SAMPLE_RAW** 4152 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4153 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4154 * 4155 * The *flags* are used to indicate the index in *map* for which 4156 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4157 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4158 * to indicate that the index of the current CPU core should be 4159 * used. 4160 * 4161 * The value to write, of *size*, is passed through eBPF stack and 4162 * pointed by *data*. 4163 * 4164 * *ctx* is a pointer to in-kernel struct xdp_buff. 4165 * 4166 * This helper is similar to **bpf_perf_eventoutput**\ () but 4167 * restricted to raw_tracepoint bpf programs. 4168 * Return 4169 * 0 on success, or a negative error in case of failure. 4170 * 4171 * u64 bpf_get_netns_cookie(void *ctx) 4172 * Description 4173 * Retrieve the cookie (generated by the kernel) of the network 4174 * namespace the input *ctx* is associated with. The network 4175 * namespace cookie remains stable for its lifetime and provides 4176 * a global identifier that can be assumed unique. If *ctx* is 4177 * NULL, then the helper returns the cookie for the initial 4178 * network namespace. The cookie itself is very similar to that 4179 * of **bpf_get_socket_cookie**\ () helper, but for network 4180 * namespaces instead of sockets. 4181 * Return 4182 * A 8-byte long opaque number. 4183 * 4184 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4185 * Description 4186 * Return id of cgroup v2 that is ancestor of the cgroup associated 4187 * with the current task at the *ancestor_level*. The root cgroup 4188 * is at *ancestor_level* zero and each step down the hierarchy 4189 * increments the level. If *ancestor_level* == level of cgroup 4190 * associated with the current task, then return value will be the 4191 * same as that of **bpf_get_current_cgroup_id**\ (). 4192 * 4193 * The helper is useful to implement policies based on cgroups 4194 * that are upper in hierarchy than immediate cgroup associated 4195 * with the current task. 4196 * 4197 * The format of returned id and helper limitations are same as in 4198 * **bpf_get_current_cgroup_id**\ (). 4199 * Return 4200 * The id is returned or 0 in case the id could not be retrieved. 4201 * 4202 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4203 * Description 4204 * Helper is overloaded depending on BPF program type. This 4205 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4206 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4207 * 4208 * Assign the *sk* to the *skb*. When combined with appropriate 4209 * routing configuration to receive the packet towards the socket, 4210 * will cause *skb* to be delivered to the specified socket. 4211 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4212 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4213 * interfere with successful delivery to the socket. 4214 * 4215 * This operation is only valid from TC ingress path. 4216 * 4217 * The *flags* argument must be zero. 4218 * Return 4219 * 0 on success, or a negative error in case of failure: 4220 * 4221 * **-EINVAL** if specified *flags* are not supported. 4222 * 4223 * **-ENOENT** if the socket is unavailable for assignment. 4224 * 4225 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4226 * 4227 * **-EOPNOTSUPP** if the operation is not supported, for example 4228 * a call from outside of TC ingress. 4229 * 4230 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4231 * Description 4232 * Helper is overloaded depending on BPF program type. This 4233 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4234 * 4235 * Select the *sk* as a result of a socket lookup. 4236 * 4237 * For the operation to succeed passed socket must be compatible 4238 * with the packet description provided by the *ctx* object. 4239 * 4240 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4241 * be an exact match. While IP family (**AF_INET** or 4242 * **AF_INET6**) must be compatible, that is IPv6 sockets 4243 * that are not v6-only can be selected for IPv4 packets. 4244 * 4245 * Only TCP listeners and UDP unconnected sockets can be 4246 * selected. *sk* can also be NULL to reset any previous 4247 * selection. 4248 * 4249 * *flags* argument can combination of following values: 4250 * 4251 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4252 * socket selection, potentially done by a BPF program 4253 * that ran before us. 4254 * 4255 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4256 * load-balancing within reuseport group for the socket 4257 * being selected. 4258 * 4259 * On success *ctx->sk* will point to the selected socket. 4260 * 4261 * Return 4262 * 0 on success, or a negative errno in case of failure. 4263 * 4264 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4265 * not compatible with packet family (*ctx->family*). 4266 * 4267 * * **-EEXIST** if socket has been already selected, 4268 * potentially by another program, and 4269 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4270 * 4271 * * **-EINVAL** if unsupported flags were specified. 4272 * 4273 * * **-EPROTOTYPE** if socket L4 protocol 4274 * (*sk->protocol*) doesn't match packet protocol 4275 * (*ctx->protocol*). 4276 * 4277 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4278 * state (TCP listening or UDP unconnected). 4279 * 4280 * u64 bpf_ktime_get_boot_ns(void) 4281 * Description 4282 * Return the time elapsed since system boot, in nanoseconds. 4283 * Does include the time the system was suspended. 4284 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4285 * Return 4286 * Current *ktime*. 4287 * 4288 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4289 * Description 4290 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4291 * out the format string. 4292 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4293 * the format string itself. The *data* and *data_len* are format string 4294 * arguments. The *data* are a **u64** array and corresponding format string 4295 * values are stored in the array. For strings and pointers where pointees 4296 * are accessed, only the pointer values are stored in the *data* array. 4297 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4298 * 4299 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4300 * Reading kernel memory may fail due to either invalid address or 4301 * valid address but requiring a major memory fault. If reading kernel memory 4302 * fails, the string for **%s** will be an empty string, and the ip 4303 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4304 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4305 * Return 4306 * 0 on success, or a negative error in case of failure: 4307 * 4308 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4309 * by returning 1 from bpf program. 4310 * 4311 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4312 * 4313 * **-E2BIG** if *fmt* contains too many format specifiers. 4314 * 4315 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4316 * 4317 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4318 * Description 4319 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4320 * The *m* represents the seq_file. The *data* and *len* represent the 4321 * data to write in bytes. 4322 * Return 4323 * 0 on success, or a negative error in case of failure: 4324 * 4325 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4326 * 4327 * u64 bpf_sk_cgroup_id(void *sk) 4328 * Description 4329 * Return the cgroup v2 id of the socket *sk*. 4330 * 4331 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4332 * returned from **bpf_sk_lookup_xxx**\ (), 4333 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4334 * same as in **bpf_skb_cgroup_id**\ (). 4335 * 4336 * This helper is available only if the kernel was compiled with 4337 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4338 * Return 4339 * The id is returned or 0 in case the id could not be retrieved. 4340 * 4341 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4342 * Description 4343 * Return id of cgroup v2 that is ancestor of cgroup associated 4344 * with the *sk* at the *ancestor_level*. The root cgroup is at 4345 * *ancestor_level* zero and each step down the hierarchy 4346 * increments the level. If *ancestor_level* == level of cgroup 4347 * associated with *sk*, then return value will be same as that 4348 * of **bpf_sk_cgroup_id**\ (). 4349 * 4350 * The helper is useful to implement policies based on cgroups 4351 * that are upper in hierarchy than immediate cgroup associated 4352 * with *sk*. 4353 * 4354 * The format of returned id and helper limitations are same as in 4355 * **bpf_sk_cgroup_id**\ (). 4356 * Return 4357 * The id is returned or 0 in case the id could not be retrieved. 4358 * 4359 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4360 * Description 4361 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4362 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4363 * of new data availability is sent. 4364 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4365 * of new data availability is sent unconditionally. 4366 * If **0** is specified in *flags*, an adaptive notification 4367 * of new data availability is sent. 4368 * 4369 * An adaptive notification is a notification sent whenever the user-space 4370 * process has caught up and consumed all available payloads. In case the user-space 4371 * process is still processing a previous payload, then no notification is needed 4372 * as it will process the newly added payload automatically. 4373 * Return 4374 * 0 on success, or a negative error in case of failure. 4375 * 4376 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4377 * Description 4378 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4379 * *flags* must be 0. 4380 * Return 4381 * Valid pointer with *size* bytes of memory available; NULL, 4382 * otherwise. 4383 * 4384 * void bpf_ringbuf_submit(void *data, u64 flags) 4385 * Description 4386 * Submit reserved ring buffer sample, pointed to by *data*. 4387 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4388 * of new data availability is sent. 4389 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4390 * of new data availability is sent unconditionally. 4391 * If **0** is specified in *flags*, an adaptive notification 4392 * of new data availability is sent. 4393 * 4394 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4395 * Return 4396 * Nothing. Always succeeds. 4397 * 4398 * void bpf_ringbuf_discard(void *data, u64 flags) 4399 * Description 4400 * Discard reserved ring buffer sample, pointed to by *data*. 4401 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4402 * of new data availability is sent. 4403 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4404 * of new data availability is sent unconditionally. 4405 * If **0** is specified in *flags*, an adaptive notification 4406 * of new data availability is sent. 4407 * 4408 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4409 * Return 4410 * Nothing. Always succeeds. 4411 * 4412 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4413 * Description 4414 * Query various characteristics of provided ring buffer. What 4415 * exactly is queries is determined by *flags*: 4416 * 4417 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4418 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4419 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4420 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4421 * 4422 * Data returned is just a momentary snapshot of actual values 4423 * and could be inaccurate, so this facility should be used to 4424 * power heuristics and for reporting, not to make 100% correct 4425 * calculation. 4426 * Return 4427 * Requested value, or 0, if *flags* are not recognized. 4428 * 4429 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4430 * Description 4431 * Change the skbs checksum level by one layer up or down, or 4432 * reset it entirely to none in order to have the stack perform 4433 * checksum validation. The level is applicable to the following 4434 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4435 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4436 * through **bpf_skb_adjust_room**\ () helper with passing in 4437 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4438 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4439 * the UDP header is removed. Similarly, an encap of the latter 4440 * into the former could be accompanied by a helper call to 4441 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4442 * skb is still intended to be processed in higher layers of the 4443 * stack instead of just egressing at tc. 4444 * 4445 * There are three supported level settings at this time: 4446 * 4447 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4448 * with CHECKSUM_UNNECESSARY. 4449 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4450 * with CHECKSUM_UNNECESSARY. 4451 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4452 * sets CHECKSUM_NONE to force checksum validation by the stack. 4453 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4454 * skb->csum_level. 4455 * Return 4456 * 0 on success, or a negative error in case of failure. In the 4457 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4458 * is returned or the error code -EACCES in case the skb is not 4459 * subject to CHECKSUM_UNNECESSARY. 4460 * 4461 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4462 * Description 4463 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4464 * Return 4465 * *sk* if casting is valid, or **NULL** otherwise. 4466 * 4467 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4468 * Description 4469 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4470 * Return 4471 * *sk* if casting is valid, or **NULL** otherwise. 4472 * 4473 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4474 * Description 4475 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4476 * Return 4477 * *sk* if casting is valid, or **NULL** otherwise. 4478 * 4479 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4480 * Description 4481 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4482 * Return 4483 * *sk* if casting is valid, or **NULL** otherwise. 4484 * 4485 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4486 * Description 4487 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4488 * Return 4489 * *sk* if casting is valid, or **NULL** otherwise. 4490 * 4491 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4492 * Description 4493 * Return a user or a kernel stack in bpf program provided buffer. 4494 * To achieve this, the helper needs *task*, which is a valid 4495 * pointer to **struct task_struct**. To store the stacktrace, the 4496 * bpf program provides *buf* with a nonnegative *size*. 4497 * 4498 * The last argument, *flags*, holds the number of stack frames to 4499 * skip (from 0 to 255), masked with 4500 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4501 * the following flags: 4502 * 4503 * **BPF_F_USER_STACK** 4504 * Collect a user space stack instead of a kernel stack. 4505 * **BPF_F_USER_BUILD_ID** 4506 * Collect buildid+offset instead of ips for user stack, 4507 * only valid if **BPF_F_USER_STACK** is also specified. 4508 * 4509 * **bpf_get_task_stack**\ () can collect up to 4510 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4511 * to sufficient large buffer size. Note that 4512 * this limit can be controlled with the **sysctl** program, and 4513 * that it should be manually increased in order to profile long 4514 * user stacks (such as stacks for Java programs). To do so, use: 4515 * 4516 * :: 4517 * 4518 * # sysctl kernel.perf_event_max_stack=<new value> 4519 * Return 4520 * The non-negative copied *buf* length equal to or less than 4521 * *size* on success, or a negative error in case of failure. 4522 * 4523 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4524 * Description 4525 * Load header option. Support reading a particular TCP header 4526 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4527 * 4528 * If *flags* is 0, it will search the option from the 4529 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4530 * has details on what skb_data contains under different 4531 * *skops*\ **->op**. 4532 * 4533 * The first byte of the *searchby_res* specifies the 4534 * kind that it wants to search. 4535 * 4536 * If the searching kind is an experimental kind 4537 * (i.e. 253 or 254 according to RFC6994). It also 4538 * needs to specify the "magic" which is either 4539 * 2 bytes or 4 bytes. It then also needs to 4540 * specify the size of the magic by using 4541 * the 2nd byte which is "kind-length" of a TCP 4542 * header option and the "kind-length" also 4543 * includes the first 2 bytes "kind" and "kind-length" 4544 * itself as a normal TCP header option also does. 4545 * 4546 * For example, to search experimental kind 254 with 4547 * 2 byte magic 0xeB9F, the searchby_res should be 4548 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4549 * 4550 * To search for the standard window scale option (3), 4551 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4552 * Note, kind-length must be 0 for regular option. 4553 * 4554 * Searching for No-Op (0) and End-of-Option-List (1) are 4555 * not supported. 4556 * 4557 * *len* must be at least 2 bytes which is the minimal size 4558 * of a header option. 4559 * 4560 * Supported flags: 4561 * 4562 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4563 * saved_syn packet or the just-received syn packet. 4564 * 4565 * Return 4566 * > 0 when found, the header option is copied to *searchby_res*. 4567 * The return value is the total length copied. On failure, a 4568 * negative error code is returned: 4569 * 4570 * **-EINVAL** if a parameter is invalid. 4571 * 4572 * **-ENOMSG** if the option is not found. 4573 * 4574 * **-ENOENT** if no syn packet is available when 4575 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4576 * 4577 * **-ENOSPC** if there is not enough space. Only *len* number of 4578 * bytes are copied. 4579 * 4580 * **-EFAULT** on failure to parse the header options in the 4581 * packet. 4582 * 4583 * **-EPERM** if the helper cannot be used under the current 4584 * *skops*\ **->op**. 4585 * 4586 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4587 * Description 4588 * Store header option. The data will be copied 4589 * from buffer *from* with length *len* to the TCP header. 4590 * 4591 * The buffer *from* should have the whole option that 4592 * includes the kind, kind-length, and the actual 4593 * option data. The *len* must be at least kind-length 4594 * long. The kind-length does not have to be 4 byte 4595 * aligned. The kernel will take care of the padding 4596 * and setting the 4 bytes aligned value to th->doff. 4597 * 4598 * This helper will check for duplicated option 4599 * by searching the same option in the outgoing skb. 4600 * 4601 * This helper can only be called during 4602 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4603 * 4604 * Return 4605 * 0 on success, or negative error in case of failure: 4606 * 4607 * **-EINVAL** If param is invalid. 4608 * 4609 * **-ENOSPC** if there is not enough space in the header. 4610 * Nothing has been written 4611 * 4612 * **-EEXIST** if the option already exists. 4613 * 4614 * **-EFAULT** on failure to parse the existing header options. 4615 * 4616 * **-EPERM** if the helper cannot be used under the current 4617 * *skops*\ **->op**. 4618 * 4619 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4620 * Description 4621 * Reserve *len* bytes for the bpf header option. The 4622 * space will be used by **bpf_store_hdr_opt**\ () later in 4623 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4624 * 4625 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4626 * the total number of bytes will be reserved. 4627 * 4628 * This helper can only be called during 4629 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4630 * 4631 * Return 4632 * 0 on success, or negative error in case of failure: 4633 * 4634 * **-EINVAL** if a parameter is invalid. 4635 * 4636 * **-ENOSPC** if there is not enough space in the header. 4637 * 4638 * **-EPERM** if the helper cannot be used under the current 4639 * *skops*\ **->op**. 4640 * 4641 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4642 * Description 4643 * Get a bpf_local_storage from an *inode*. 4644 * 4645 * Logically, it could be thought of as getting the value from 4646 * a *map* with *inode* as the **key**. From this 4647 * perspective, the usage is not much different from 4648 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4649 * helper enforces the key must be an inode and the map must also 4650 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4651 * 4652 * Underneath, the value is stored locally at *inode* instead of 4653 * the *map*. The *map* is used as the bpf-local-storage 4654 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4655 * searched against all bpf_local_storage residing at *inode*. 4656 * 4657 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4658 * used such that a new bpf_local_storage will be 4659 * created if one does not exist. *value* can be used 4660 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4661 * the initial value of a bpf_local_storage. If *value* is 4662 * **NULL**, the new bpf_local_storage will be zero initialized. 4663 * Return 4664 * A bpf_local_storage pointer is returned on success. 4665 * 4666 * **NULL** if not found or there was an error in adding 4667 * a new bpf_local_storage. 4668 * 4669 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4670 * Description 4671 * Delete a bpf_local_storage from an *inode*. 4672 * Return 4673 * 0 on success. 4674 * 4675 * **-ENOENT** if the bpf_local_storage cannot be found. 4676 * 4677 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4678 * Description 4679 * Return full path for given **struct path** object, which 4680 * needs to be the kernel BTF *path* object. The path is 4681 * returned in the provided buffer *buf* of size *sz* and 4682 * is zero terminated. 4683 * 4684 * Return 4685 * On success, the strictly positive length of the string, 4686 * including the trailing NUL character. On error, a negative 4687 * value. 4688 * 4689 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4690 * Description 4691 * Read *size* bytes from user space address *user_ptr* and store 4692 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4693 * Return 4694 * 0 on success, or a negative error in case of failure. 4695 * 4696 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4697 * Description 4698 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4699 * using *ptr*->type_id. This value should specify the type 4700 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4701 * can be used to look up vmlinux BTF type ids. Traversing the 4702 * data structure using BTF, the type information and values are 4703 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4704 * the pointer data is carried out to avoid kernel crashes during 4705 * operation. Smaller types can use string space on the stack; 4706 * larger programs can use map data to store the string 4707 * representation. 4708 * 4709 * The string can be subsequently shared with userspace via 4710 * bpf_perf_event_output() or ring buffer interfaces. 4711 * bpf_trace_printk() is to be avoided as it places too small 4712 * a limit on string size to be useful. 4713 * 4714 * *flags* is a combination of 4715 * 4716 * **BTF_F_COMPACT** 4717 * no formatting around type information 4718 * **BTF_F_NONAME** 4719 * no struct/union member names/types 4720 * **BTF_F_PTR_RAW** 4721 * show raw (unobfuscated) pointer values; 4722 * equivalent to printk specifier %px. 4723 * **BTF_F_ZERO** 4724 * show zero-valued struct/union members; they 4725 * are not displayed by default 4726 * 4727 * Return 4728 * The number of bytes that were written (or would have been 4729 * written if output had to be truncated due to string size), 4730 * or a negative error in cases of failure. 4731 * 4732 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4733 * Description 4734 * Use BTF to write to seq_write a string representation of 4735 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4736 * *flags* are identical to those used for bpf_snprintf_btf. 4737 * Return 4738 * 0 on success or a negative error in case of failure. 4739 * 4740 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4741 * Description 4742 * See **bpf_get_cgroup_classid**\ () for the main description. 4743 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4744 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4745 * associated socket instead of the current process. 4746 * Return 4747 * The id is returned or 0 in case the id could not be retrieved. 4748 * 4749 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4750 * Description 4751 * Redirect the packet to another net device of index *ifindex* 4752 * and fill in L2 addresses from neighboring subsystem. This helper 4753 * is somewhat similar to **bpf_redirect**\ (), except that it 4754 * populates L2 addresses as well, meaning, internally, the helper 4755 * relies on the neighbor lookup for the L2 address of the nexthop. 4756 * 4757 * The helper will perform a FIB lookup based on the skb's 4758 * networking header to get the address of the next hop, unless 4759 * this is supplied by the caller in the *params* argument. The 4760 * *plen* argument indicates the len of *params* and should be set 4761 * to 0 if *params* is NULL. 4762 * 4763 * The *flags* argument is reserved and must be 0. The helper is 4764 * currently only supported for tc BPF program types, and enabled 4765 * for IPv4 and IPv6 protocols. 4766 * Return 4767 * The helper returns **TC_ACT_REDIRECT** on success or 4768 * **TC_ACT_SHOT** on error. 4769 * 4770 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4771 * Description 4772 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4773 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4774 * extern variable decorated with '__ksym'. For ksym, there is a 4775 * global var (either static or global) defined of the same name 4776 * in the kernel. The ksym is percpu if the global var is percpu. 4777 * The returned pointer points to the global percpu var on *cpu*. 4778 * 4779 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4780 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4781 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4782 * bpf_per_cpu_ptr() must check the returned value. 4783 * Return 4784 * A pointer pointing to the kernel percpu variable on *cpu*, or 4785 * NULL, if *cpu* is invalid. 4786 * 4787 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4788 * Description 4789 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4790 * pointer to the percpu kernel variable on this cpu. See the 4791 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4792 * 4793 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4794 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4795 * never return NULL. 4796 * Return 4797 * A pointer pointing to the kernel percpu variable on this cpu. 4798 * 4799 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4800 * Description 4801 * Redirect the packet to another net device of index *ifindex*. 4802 * This helper is somewhat similar to **bpf_redirect**\ (), except 4803 * that the redirection happens to the *ifindex*' peer device and 4804 * the netns switch takes place from ingress to ingress without 4805 * going through the CPU's backlog queue. 4806 * 4807 * The *flags* argument is reserved and must be 0. The helper is 4808 * currently only supported for tc BPF program types at the ingress 4809 * hook and for veth device types. The peer device must reside in a 4810 * different network namespace. 4811 * Return 4812 * The helper returns **TC_ACT_REDIRECT** on success or 4813 * **TC_ACT_SHOT** on error. 4814 * 4815 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4816 * Description 4817 * Get a bpf_local_storage from the *task*. 4818 * 4819 * Logically, it could be thought of as getting the value from 4820 * a *map* with *task* as the **key**. From this 4821 * perspective, the usage is not much different from 4822 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4823 * helper enforces the key must be a task_struct and the map must also 4824 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4825 * 4826 * Underneath, the value is stored locally at *task* instead of 4827 * the *map*. The *map* is used as the bpf-local-storage 4828 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4829 * searched against all bpf_local_storage residing at *task*. 4830 * 4831 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4832 * used such that a new bpf_local_storage will be 4833 * created if one does not exist. *value* can be used 4834 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4835 * the initial value of a bpf_local_storage. If *value* is 4836 * **NULL**, the new bpf_local_storage will be zero initialized. 4837 * Return 4838 * A bpf_local_storage pointer is returned on success. 4839 * 4840 * **NULL** if not found or there was an error in adding 4841 * a new bpf_local_storage. 4842 * 4843 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4844 * Description 4845 * Delete a bpf_local_storage from a *task*. 4846 * Return 4847 * 0 on success. 4848 * 4849 * **-ENOENT** if the bpf_local_storage cannot be found. 4850 * 4851 * struct task_struct *bpf_get_current_task_btf(void) 4852 * Description 4853 * Return a BTF pointer to the "current" task. 4854 * This pointer can also be used in helpers that accept an 4855 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4856 * Return 4857 * Pointer to the current task. 4858 * 4859 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4860 * Description 4861 * Set or clear certain options on *bprm*: 4862 * 4863 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4864 * which sets the **AT_SECURE** auxv for glibc. The bit 4865 * is cleared if the flag is not specified. 4866 * Return 4867 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4868 * 4869 * u64 bpf_ktime_get_coarse_ns(void) 4870 * Description 4871 * Return a coarse-grained version of the time elapsed since 4872 * system boot, in nanoseconds. Does not include time the system 4873 * was suspended. 4874 * 4875 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4876 * Return 4877 * Current *ktime*. 4878 * 4879 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4880 * Description 4881 * Returns the stored IMA hash of the *inode* (if it's available). 4882 * If the hash is larger than *size*, then only *size* 4883 * bytes will be copied to *dst* 4884 * Return 4885 * The **hash_algo** is returned on success, 4886 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4887 * invalid arguments are passed. 4888 * 4889 * struct socket *bpf_sock_from_file(struct file *file) 4890 * Description 4891 * If the given file represents a socket, returns the associated 4892 * socket. 4893 * Return 4894 * A pointer to a struct socket on success or NULL if the file is 4895 * not a socket. 4896 * 4897 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4898 * Description 4899 * Check packet size against exceeding MTU of net device (based 4900 * on *ifindex*). This helper will likely be used in combination 4901 * with helpers that adjust/change the packet size. 4902 * 4903 * The argument *len_diff* can be used for querying with a planned 4904 * size change. This allows to check MTU prior to changing packet 4905 * ctx. Providing a *len_diff* adjustment that is larger than the 4906 * actual packet size (resulting in negative packet size) will in 4907 * principle not exceed the MTU, which is why it is not considered 4908 * a failure. Other BPF helpers are needed for performing the 4909 * planned size change; therefore the responsibility for catching 4910 * a negative packet size belongs in those helpers. 4911 * 4912 * Specifying *ifindex* zero means the MTU check is performed 4913 * against the current net device. This is practical if this isn't 4914 * used prior to redirect. 4915 * 4916 * On input *mtu_len* must be a valid pointer, else verifier will 4917 * reject BPF program. If the value *mtu_len* is initialized to 4918 * zero then the ctx packet size is use. When value *mtu_len* is 4919 * provided as input this specify the L3 length that the MTU check 4920 * is done against. Remember XDP and TC length operate at L2, but 4921 * this value is L3 as this correlate to MTU and IP-header tot_len 4922 * values which are L3 (similar behavior as bpf_fib_lookup). 4923 * 4924 * The Linux kernel route table can configure MTUs on a more 4925 * specific per route level, which is not provided by this helper. 4926 * For route level MTU checks use the **bpf_fib_lookup**\ () 4927 * helper. 4928 * 4929 * *ctx* is either **struct xdp_md** for XDP programs or 4930 * **struct sk_buff** for tc cls_act programs. 4931 * 4932 * The *flags* argument can be a combination of one or more of the 4933 * following values: 4934 * 4935 * **BPF_MTU_CHK_SEGS** 4936 * This flag will only works for *ctx* **struct sk_buff**. 4937 * If packet context contains extra packet segment buffers 4938 * (often knows as GSO skb), then MTU check is harder to 4939 * check at this point, because in transmit path it is 4940 * possible for the skb packet to get re-segmented 4941 * (depending on net device features). This could still be 4942 * a MTU violation, so this flag enables performing MTU 4943 * check against segments, with a different violation 4944 * return code to tell it apart. Check cannot use len_diff. 4945 * 4946 * On return *mtu_len* pointer contains the MTU value of the net 4947 * device. Remember the net device configured MTU is the L3 size, 4948 * which is returned here and XDP and TC length operate at L2. 4949 * Helper take this into account for you, but remember when using 4950 * MTU value in your BPF-code. 4951 * 4952 * Return 4953 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4954 * 4955 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4956 * 4957 * MTU violations return positive values, but also populate MTU 4958 * value in *mtu_len* pointer, as this can be needed for 4959 * implementing PMTU handing: 4960 * 4961 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4962 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4963 * 4964 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4965 * Description 4966 * For each element in **map**, call **callback_fn** function with 4967 * **map**, **callback_ctx** and other map-specific parameters. 4968 * The **callback_fn** should be a static function and 4969 * the **callback_ctx** should be a pointer to the stack. 4970 * The **flags** is used to control certain aspects of the helper. 4971 * Currently, the **flags** must be 0. 4972 * 4973 * The following are a list of supported map types and their 4974 * respective expected callback signatures: 4975 * 4976 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4977 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4978 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4979 * 4980 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4981 * 4982 * For per_cpu maps, the map_value is the value on the cpu where the 4983 * bpf_prog is running. 4984 * 4985 * If **callback_fn** return 0, the helper will continue to the next 4986 * element. If return value is 1, the helper will skip the rest of 4987 * elements and return. Other return values are not used now. 4988 * 4989 * Return 4990 * The number of traversed map elements for success, **-EINVAL** for 4991 * invalid **flags**. 4992 * 4993 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4994 * Description 4995 * Outputs a string into the **str** buffer of size **str_size** 4996 * based on a format string stored in a read-only map pointed by 4997 * **fmt**. 4998 * 4999 * Each format specifier in **fmt** corresponds to one u64 element 5000 * in the **data** array. For strings and pointers where pointees 5001 * are accessed, only the pointer values are stored in the *data* 5002 * array. The *data_len* is the size of *data* in bytes - must be 5003 * a multiple of 8. 5004 * 5005 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5006 * memory. Reading kernel memory may fail due to either invalid 5007 * address or valid address but requiring a major memory fault. If 5008 * reading kernel memory fails, the string for **%s** will be an 5009 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5010 * Not returning error to bpf program is consistent with what 5011 * **bpf_trace_printk**\ () does for now. 5012 * 5013 * Return 5014 * The strictly positive length of the formatted string, including 5015 * the trailing zero character. If the return value is greater than 5016 * **str_size**, **str** contains a truncated string, guaranteed to 5017 * be zero-terminated except when **str_size** is 0. 5018 * 5019 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5020 * 5021 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5022 * Description 5023 * Execute bpf syscall with given arguments. 5024 * Return 5025 * A syscall result. 5026 * 5027 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5028 * Description 5029 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5030 * Return 5031 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5032 * 5033 * long bpf_sys_close(u32 fd) 5034 * Description 5035 * Execute close syscall for given FD. 5036 * Return 5037 * A syscall result. 5038 * 5039 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5040 * Description 5041 * Initialize the timer. 5042 * First 4 bits of *flags* specify clockid. 5043 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5044 * All other bits of *flags* are reserved. 5045 * The verifier will reject the program if *timer* is not from 5046 * the same *map*. 5047 * Return 5048 * 0 on success. 5049 * **-EBUSY** if *timer* is already initialized. 5050 * **-EINVAL** if invalid *flags* are passed. 5051 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5052 * The user space should either hold a file descriptor to a map with timers 5053 * or pin such map in bpffs. When map is unpinned or file descriptor is 5054 * closed all timers in the map will be cancelled and freed. 5055 * 5056 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5057 * Description 5058 * Configure the timer to call *callback_fn* static function. 5059 * Return 5060 * 0 on success. 5061 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5062 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5063 * The user space should either hold a file descriptor to a map with timers 5064 * or pin such map in bpffs. When map is unpinned or file descriptor is 5065 * closed all timers in the map will be cancelled and freed. 5066 * 5067 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5068 * Description 5069 * Set timer expiration N nanoseconds from the current time. The 5070 * configured callback will be invoked in soft irq context on some cpu 5071 * and will not repeat unless another bpf_timer_start() is made. 5072 * In such case the next invocation can migrate to a different cpu. 5073 * Since struct bpf_timer is a field inside map element the map 5074 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5075 * of BPF program to make sure that callback_fn code stays valid. 5076 * When user space reference to a map reaches zero all timers 5077 * in a map are cancelled and corresponding program's refcnts are 5078 * decremented. This is done to make sure that Ctrl-C of a user 5079 * process doesn't leave any timers running. If map is pinned in 5080 * bpffs the callback_fn can re-arm itself indefinitely. 5081 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5082 * cancel and free the timer in the given map element. 5083 * The map can contain timers that invoke callback_fn-s from different 5084 * programs. The same callback_fn can serve different timers from 5085 * different maps if key/value layout matches across maps. 5086 * Every bpf_timer_set_callback() can have different callback_fn. 5087 * 5088 * *flags* can be one of: 5089 * 5090 * **BPF_F_TIMER_ABS** 5091 * Start the timer in absolute expire value instead of the 5092 * default relative one. 5093 * 5094 * Return 5095 * 0 on success. 5096 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5097 * or invalid *flags* are passed. 5098 * 5099 * long bpf_timer_cancel(struct bpf_timer *timer) 5100 * Description 5101 * Cancel the timer and wait for callback_fn to finish if it was running. 5102 * Return 5103 * 0 if the timer was not active. 5104 * 1 if the timer was active. 5105 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5106 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5107 * own timer which would have led to a deadlock otherwise. 5108 * 5109 * u64 bpf_get_func_ip(void *ctx) 5110 * Description 5111 * Get address of the traced function (for tracing and kprobe programs). 5112 * 5113 * When called for kprobe program attached as uprobe it returns 5114 * probe address for both entry and return uprobe. 5115 * 5116 * Return 5117 * Address of the traced function for kprobe. 5118 * 0 for kprobes placed within the function (not at the entry). 5119 * Address of the probe for uprobe and return uprobe. 5120 * 5121 * u64 bpf_get_attach_cookie(void *ctx) 5122 * Description 5123 * Get bpf_cookie value provided (optionally) during the program 5124 * attachment. It might be different for each individual 5125 * attachment, even if BPF program itself is the same. 5126 * Expects BPF program context *ctx* as a first argument. 5127 * 5128 * Supported for the following program types: 5129 * - kprobe/uprobe; 5130 * - tracepoint; 5131 * - perf_event. 5132 * Return 5133 * Value specified by user at BPF link creation/attachment time 5134 * or 0, if it was not specified. 5135 * 5136 * long bpf_task_pt_regs(struct task_struct *task) 5137 * Description 5138 * Get the struct pt_regs associated with **task**. 5139 * Return 5140 * A pointer to struct pt_regs. 5141 * 5142 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5143 * Description 5144 * Get branch trace from hardware engines like Intel LBR. The 5145 * hardware engine is stopped shortly after the helper is 5146 * called. Therefore, the user need to filter branch entries 5147 * based on the actual use case. To capture branch trace 5148 * before the trigger point of the BPF program, the helper 5149 * should be called at the beginning of the BPF program. 5150 * 5151 * The data is stored as struct perf_branch_entry into output 5152 * buffer *entries*. *size* is the size of *entries* in bytes. 5153 * *flags* is reserved for now and must be zero. 5154 * 5155 * Return 5156 * On success, number of bytes written to *buf*. On error, a 5157 * negative value. 5158 * 5159 * **-EINVAL** if *flags* is not zero. 5160 * 5161 * **-ENOENT** if architecture does not support branch records. 5162 * 5163 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5164 * Description 5165 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5166 * to format and can handle more format args as a result. 5167 * 5168 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5169 * Return 5170 * The number of bytes written to the buffer, or a negative error 5171 * in case of failure. 5172 * 5173 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5174 * Description 5175 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5176 * Return 5177 * *sk* if casting is valid, or **NULL** otherwise. 5178 * 5179 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5180 * Description 5181 * Get the address of a kernel symbol, returned in *res*. *res* is 5182 * set to 0 if the symbol is not found. 5183 * Return 5184 * On success, zero. On error, a negative value. 5185 * 5186 * **-EINVAL** if *flags* is not zero. 5187 * 5188 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5189 * 5190 * **-ENOENT** if symbol is not found. 5191 * 5192 * **-EPERM** if caller does not have permission to obtain kernel address. 5193 * 5194 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5195 * Description 5196 * Find vma of *task* that contains *addr*, call *callback_fn* 5197 * function with *task*, *vma*, and *callback_ctx*. 5198 * The *callback_fn* should be a static function and 5199 * the *callback_ctx* should be a pointer to the stack. 5200 * The *flags* is used to control certain aspects of the helper. 5201 * Currently, the *flags* must be 0. 5202 * 5203 * The expected callback signature is 5204 * 5205 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5206 * 5207 * Return 5208 * 0 on success. 5209 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5210 * **-EBUSY** if failed to try lock mmap_lock. 5211 * **-EINVAL** for invalid **flags**. 5212 * 5213 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5214 * Description 5215 * For **nr_loops**, call **callback_fn** function 5216 * with **callback_ctx** as the context parameter. 5217 * The **callback_fn** should be a static function and 5218 * the **callback_ctx** should be a pointer to the stack. 5219 * The **flags** is used to control certain aspects of the helper. 5220 * Currently, the **flags** must be 0. Currently, nr_loops is 5221 * limited to 1 << 23 (~8 million) loops. 5222 * 5223 * long (\*callback_fn)(u32 index, void \*ctx); 5224 * 5225 * where **index** is the current index in the loop. The index 5226 * is zero-indexed. 5227 * 5228 * If **callback_fn** returns 0, the helper will continue to the next 5229 * loop. If return value is 1, the helper will skip the rest of 5230 * the loops and return. Other return values are not used now, 5231 * and will be rejected by the verifier. 5232 * 5233 * Return 5234 * The number of loops performed, **-EINVAL** for invalid **flags**, 5235 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5236 * 5237 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5238 * Description 5239 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5240 * to be null-terminated and **s1_sz** is the maximum storage 5241 * size of **s1**. **s2** must be a read-only string. 5242 * Return 5243 * An integer less than, equal to, or greater than zero 5244 * if the first **s1_sz** bytes of **s1** is found to be 5245 * less than, to match, or be greater than **s2**. 5246 * 5247 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5248 * Description 5249 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5250 * returned in **value**. 5251 * 5252 * Return 5253 * 0 on success. 5254 * **-EINVAL** if n >= argument register count of traced function. 5255 * 5256 * long bpf_get_func_ret(void *ctx, u64 *value) 5257 * Description 5258 * Get return value of the traced function (for tracing programs) 5259 * in **value**. 5260 * 5261 * Return 5262 * 0 on success. 5263 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5264 * 5265 * long bpf_get_func_arg_cnt(void *ctx) 5266 * Description 5267 * Get number of registers of the traced function (for tracing programs) where 5268 * function arguments are stored in these registers. 5269 * 5270 * Return 5271 * The number of argument registers of the traced function. 5272 * 5273 * int bpf_get_retval(void) 5274 * Description 5275 * Get the BPF program's return value that will be returned to the upper layers. 5276 * 5277 * This helper is currently supported by cgroup programs and only by the hooks 5278 * where BPF program's return value is returned to the userspace via errno. 5279 * Return 5280 * The BPF program's return value. 5281 * 5282 * int bpf_set_retval(int retval) 5283 * Description 5284 * Set the BPF program's return value that will be returned to the upper layers. 5285 * 5286 * This helper is currently supported by cgroup programs and only by the hooks 5287 * where BPF program's return value is returned to the userspace via errno. 5288 * 5289 * Note that there is the following corner case where the program exports an error 5290 * via bpf_set_retval but signals success via 'return 1': 5291 * 5292 * bpf_set_retval(-EPERM); 5293 * return 1; 5294 * 5295 * In this case, the BPF program's return value will use helper's -EPERM. This 5296 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5297 * 5298 * Return 5299 * 0 on success, or a negative error in case of failure. 5300 * 5301 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5302 * Description 5303 * Get the total size of a given xdp buff (linear and paged area) 5304 * Return 5305 * The total size of a given xdp buffer. 5306 * 5307 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5308 * Description 5309 * This helper is provided as an easy way to load data from a 5310 * xdp buffer. It can be used to load *len* bytes from *offset* from 5311 * the frame associated to *xdp_md*, into the buffer pointed by 5312 * *buf*. 5313 * Return 5314 * 0 on success, or a negative error in case of failure. 5315 * 5316 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5317 * Description 5318 * Store *len* bytes from buffer *buf* into the frame 5319 * associated to *xdp_md*, at *offset*. 5320 * Return 5321 * 0 on success, or a negative error in case of failure. 5322 * 5323 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5324 * Description 5325 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5326 * address space, and stores the data in *dst*. *flags* is not 5327 * used yet and is provided for future extensibility. This helper 5328 * can only be used by sleepable programs. 5329 * Return 5330 * 0 on success, or a negative error in case of failure. On error 5331 * *dst* buffer is zeroed out. 5332 * 5333 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5334 * Description 5335 * Change the __sk_buff->tstamp_type to *tstamp_type* 5336 * and set *tstamp* to the __sk_buff->tstamp together. 5337 * 5338 * If there is no need to change the __sk_buff->tstamp_type, 5339 * the tstamp value can be directly written to __sk_buff->tstamp 5340 * instead. 5341 * 5342 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5343 * will be kept during bpf_redirect_*(). A non zero 5344 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5345 * *tstamp_type*. 5346 * 5347 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5348 * with a zero *tstamp*. 5349 * 5350 * Only IPv4 and IPv6 skb->protocol are supported. 5351 * 5352 * This function is most useful when it needs to set a 5353 * mono delivery time to __sk_buff->tstamp and then 5354 * bpf_redirect_*() to the egress of an iface. For example, 5355 * changing the (rcv) timestamp in __sk_buff->tstamp at 5356 * ingress to a mono delivery time and then bpf_redirect_*() 5357 * to sch_fq@phy-dev. 5358 * Return 5359 * 0 on success. 5360 * **-EINVAL** for invalid input 5361 * **-EOPNOTSUPP** for unsupported protocol 5362 * 5363 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5364 * Description 5365 * Returns a calculated IMA hash of the *file*. 5366 * If the hash is larger than *size*, then only *size* 5367 * bytes will be copied to *dst* 5368 * Return 5369 * The **hash_algo** is returned on success, 5370 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5371 * invalid arguments are passed. 5372 * 5373 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5374 * Description 5375 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5376 * old value. *ptr* can be NULL, otherwise it must be a referenced 5377 * pointer which will be released when this helper is called. 5378 * Return 5379 * The old value of kptr (which can be NULL). The returned pointer 5380 * if not NULL, is a reference which must be released using its 5381 * corresponding release function, or moved into a BPF map before 5382 * program exit. 5383 * 5384 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5385 * Description 5386 * Perform a lookup in *percpu map* for an entry associated to 5387 * *key* on *cpu*. 5388 * Return 5389 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5390 * was found or *cpu* is invalid. 5391 * 5392 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5393 * Description 5394 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5395 * Return 5396 * *sk* if casting is valid, or **NULL** otherwise. 5397 * 5398 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5399 * Description 5400 * Get a dynptr to local memory *data*. 5401 * 5402 * *data* must be a ptr to a map value. 5403 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5404 * *flags* is currently unused. 5405 * Return 5406 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5407 * -EINVAL if flags is not 0. 5408 * 5409 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5410 * Description 5411 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5412 * through the dynptr interface. *flags* must be 0. 5413 * 5414 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5415 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5416 * reservation fails. This is enforced by the verifier. 5417 * Return 5418 * 0 on success, or a negative error in case of failure. 5419 * 5420 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5421 * Description 5422 * Submit reserved ring buffer sample, pointed to by *data*, 5423 * through the dynptr interface. This is a no-op if the dynptr is 5424 * invalid/null. 5425 * 5426 * For more information on *flags*, please see 5427 * 'bpf_ringbuf_submit'. 5428 * Return 5429 * Nothing. Always succeeds. 5430 * 5431 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5432 * Description 5433 * Discard reserved ring buffer sample through the dynptr 5434 * interface. This is a no-op if the dynptr is invalid/null. 5435 * 5436 * For more information on *flags*, please see 5437 * 'bpf_ringbuf_discard'. 5438 * Return 5439 * Nothing. Always succeeds. 5440 * 5441 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5442 * Description 5443 * Read *len* bytes from *src* into *dst*, starting from *offset* 5444 * into *src*. 5445 * *flags* is currently unused. 5446 * Return 5447 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5448 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5449 * *flags* is not 0. 5450 * 5451 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5452 * Description 5453 * Write *len* bytes from *src* into *dst*, starting from *offset* 5454 * into *dst*. 5455 * 5456 * *flags* must be 0 except for skb-type dynptrs. 5457 * 5458 * For skb-type dynptrs: 5459 * * All data slices of the dynptr are automatically 5460 * invalidated after **bpf_dynptr_write**\ (). This is 5461 * because writing may pull the skb and change the 5462 * underlying packet buffer. 5463 * 5464 * * For *flags*, please see the flags accepted by 5465 * **bpf_skb_store_bytes**\ (). 5466 * Return 5467 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5468 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5469 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5470 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5471 * 5472 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5473 * Description 5474 * Get a pointer to the underlying dynptr data. 5475 * 5476 * *len* must be a statically known value. The returned data slice 5477 * is invalidated whenever the dynptr is invalidated. 5478 * 5479 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5480 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5481 * Return 5482 * Pointer to the underlying dynptr data, NULL if the dynptr is 5483 * read-only, if the dynptr is invalid, or if the offset and length 5484 * is out of bounds. 5485 * 5486 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5487 * Description 5488 * Try to issue a SYN cookie for the packet with corresponding 5489 * IPv4/TCP headers, *iph* and *th*, without depending on a 5490 * listening socket. 5491 * 5492 * *iph* points to the IPv4 header. 5493 * 5494 * *th* points to the start of the TCP header, while *th_len* 5495 * contains the length of the TCP header (at least 5496 * **sizeof**\ (**struct tcphdr**)). 5497 * Return 5498 * On success, lower 32 bits hold the generated SYN cookie in 5499 * followed by 16 bits which hold the MSS value for that cookie, 5500 * and the top 16 bits are unused. 5501 * 5502 * On failure, the returned value is one of the following: 5503 * 5504 * **-EINVAL** if *th_len* is invalid. 5505 * 5506 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5507 * Description 5508 * Try to issue a SYN cookie for the packet with corresponding 5509 * IPv6/TCP headers, *iph* and *th*, without depending on a 5510 * listening socket. 5511 * 5512 * *iph* points to the IPv6 header. 5513 * 5514 * *th* points to the start of the TCP header, while *th_len* 5515 * contains the length of the TCP header (at least 5516 * **sizeof**\ (**struct tcphdr**)). 5517 * Return 5518 * On success, lower 32 bits hold the generated SYN cookie in 5519 * followed by 16 bits which hold the MSS value for that cookie, 5520 * and the top 16 bits are unused. 5521 * 5522 * On failure, the returned value is one of the following: 5523 * 5524 * **-EINVAL** if *th_len* is invalid. 5525 * 5526 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5527 * 5528 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5529 * Description 5530 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5531 * without depending on a listening socket. 5532 * 5533 * *iph* points to the IPv4 header. 5534 * 5535 * *th* points to the TCP header. 5536 * Return 5537 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5538 * 5539 * On failure, the returned value is one of the following: 5540 * 5541 * **-EACCES** if the SYN cookie is not valid. 5542 * 5543 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5544 * Description 5545 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5546 * without depending on a listening socket. 5547 * 5548 * *iph* points to the IPv6 header. 5549 * 5550 * *th* points to the TCP header. 5551 * Return 5552 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5553 * 5554 * On failure, the returned value is one of the following: 5555 * 5556 * **-EACCES** if the SYN cookie is not valid. 5557 * 5558 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5559 * 5560 * u64 bpf_ktime_get_tai_ns(void) 5561 * Description 5562 * A nonsettable system-wide clock derived from wall-clock time but 5563 * ignoring leap seconds. This clock does not experience 5564 * discontinuities and backwards jumps caused by NTP inserting leap 5565 * seconds as CLOCK_REALTIME does. 5566 * 5567 * See: **clock_gettime**\ (**CLOCK_TAI**) 5568 * Return 5569 * Current *ktime*. 5570 * 5571 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5572 * Description 5573 * Drain samples from the specified user ring buffer, and invoke 5574 * the provided callback for each such sample: 5575 * 5576 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5577 * 5578 * If **callback_fn** returns 0, the helper will continue to try 5579 * and drain the next sample, up to a maximum of 5580 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5581 * the helper will skip the rest of the samples and return. Other 5582 * return values are not used now, and will be rejected by the 5583 * verifier. 5584 * Return 5585 * The number of drained samples if no error was encountered while 5586 * draining samples, or 0 if no samples were present in the ring 5587 * buffer. If a user-space producer was epoll-waiting on this map, 5588 * and at least one sample was drained, they will receive an event 5589 * notification notifying them of available space in the ring 5590 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5591 * function, no wakeup notification will be sent. If the 5592 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5593 * be sent even if no sample was drained. 5594 * 5595 * On failure, the returned value is one of the following: 5596 * 5597 * **-EBUSY** if the ring buffer is contended, and another calling 5598 * context was concurrently draining the ring buffer. 5599 * 5600 * **-EINVAL** if user-space is not properly tracking the ring 5601 * buffer due to the producer position not being aligned to 8 5602 * bytes, a sample not being aligned to 8 bytes, or the producer 5603 * position not matching the advertised length of a sample. 5604 * 5605 * **-E2BIG** if user-space has tried to publish a sample which is 5606 * larger than the size of the ring buffer, or which cannot fit 5607 * within a struct bpf_dynptr. 5608 * 5609 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5610 * Description 5611 * Get a bpf_local_storage from the *cgroup*. 5612 * 5613 * Logically, it could be thought of as getting the value from 5614 * a *map* with *cgroup* as the **key**. From this 5615 * perspective, the usage is not much different from 5616 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5617 * helper enforces the key must be a cgroup struct and the map must also 5618 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5619 * 5620 * In reality, the local-storage value is embedded directly inside of the 5621 * *cgroup* object itself, rather than being located in the 5622 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5623 * queried for some *map* on a *cgroup* object, the kernel will perform an 5624 * O(n) iteration over all of the live local-storage values for that 5625 * *cgroup* object until the local-storage value for the *map* is found. 5626 * 5627 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5628 * used such that a new bpf_local_storage will be 5629 * created if one does not exist. *value* can be used 5630 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5631 * the initial value of a bpf_local_storage. If *value* is 5632 * **NULL**, the new bpf_local_storage will be zero initialized. 5633 * Return 5634 * A bpf_local_storage pointer is returned on success. 5635 * 5636 * **NULL** if not found or there was an error in adding 5637 * a new bpf_local_storage. 5638 * 5639 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5640 * Description 5641 * Delete a bpf_local_storage from a *cgroup*. 5642 * Return 5643 * 0 on success. 5644 * 5645 * **-ENOENT** if the bpf_local_storage cannot be found. 5646 */ 5647 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5648 FN(unspec, 0, ##ctx) \ 5649 FN(map_lookup_elem, 1, ##ctx) \ 5650 FN(map_update_elem, 2, ##ctx) \ 5651 FN(map_delete_elem, 3, ##ctx) \ 5652 FN(probe_read, 4, ##ctx) \ 5653 FN(ktime_get_ns, 5, ##ctx) \ 5654 FN(trace_printk, 6, ##ctx) \ 5655 FN(get_prandom_u32, 7, ##ctx) \ 5656 FN(get_smp_processor_id, 8, ##ctx) \ 5657 FN(skb_store_bytes, 9, ##ctx) \ 5658 FN(l3_csum_replace, 10, ##ctx) \ 5659 FN(l4_csum_replace, 11, ##ctx) \ 5660 FN(tail_call, 12, ##ctx) \ 5661 FN(clone_redirect, 13, ##ctx) \ 5662 FN(get_current_pid_tgid, 14, ##ctx) \ 5663 FN(get_current_uid_gid, 15, ##ctx) \ 5664 FN(get_current_comm, 16, ##ctx) \ 5665 FN(get_cgroup_classid, 17, ##ctx) \ 5666 FN(skb_vlan_push, 18, ##ctx) \ 5667 FN(skb_vlan_pop, 19, ##ctx) \ 5668 FN(skb_get_tunnel_key, 20, ##ctx) \ 5669 FN(skb_set_tunnel_key, 21, ##ctx) \ 5670 FN(perf_event_read, 22, ##ctx) \ 5671 FN(redirect, 23, ##ctx) \ 5672 FN(get_route_realm, 24, ##ctx) \ 5673 FN(perf_event_output, 25, ##ctx) \ 5674 FN(skb_load_bytes, 26, ##ctx) \ 5675 FN(get_stackid, 27, ##ctx) \ 5676 FN(csum_diff, 28, ##ctx) \ 5677 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5678 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5679 FN(skb_change_proto, 31, ##ctx) \ 5680 FN(skb_change_type, 32, ##ctx) \ 5681 FN(skb_under_cgroup, 33, ##ctx) \ 5682 FN(get_hash_recalc, 34, ##ctx) \ 5683 FN(get_current_task, 35, ##ctx) \ 5684 FN(probe_write_user, 36, ##ctx) \ 5685 FN(current_task_under_cgroup, 37, ##ctx) \ 5686 FN(skb_change_tail, 38, ##ctx) \ 5687 FN(skb_pull_data, 39, ##ctx) \ 5688 FN(csum_update, 40, ##ctx) \ 5689 FN(set_hash_invalid, 41, ##ctx) \ 5690 FN(get_numa_node_id, 42, ##ctx) \ 5691 FN(skb_change_head, 43, ##ctx) \ 5692 FN(xdp_adjust_head, 44, ##ctx) \ 5693 FN(probe_read_str, 45, ##ctx) \ 5694 FN(get_socket_cookie, 46, ##ctx) \ 5695 FN(get_socket_uid, 47, ##ctx) \ 5696 FN(set_hash, 48, ##ctx) \ 5697 FN(setsockopt, 49, ##ctx) \ 5698 FN(skb_adjust_room, 50, ##ctx) \ 5699 FN(redirect_map, 51, ##ctx) \ 5700 FN(sk_redirect_map, 52, ##ctx) \ 5701 FN(sock_map_update, 53, ##ctx) \ 5702 FN(xdp_adjust_meta, 54, ##ctx) \ 5703 FN(perf_event_read_value, 55, ##ctx) \ 5704 FN(perf_prog_read_value, 56, ##ctx) \ 5705 FN(getsockopt, 57, ##ctx) \ 5706 FN(override_return, 58, ##ctx) \ 5707 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5708 FN(msg_redirect_map, 60, ##ctx) \ 5709 FN(msg_apply_bytes, 61, ##ctx) \ 5710 FN(msg_cork_bytes, 62, ##ctx) \ 5711 FN(msg_pull_data, 63, ##ctx) \ 5712 FN(bind, 64, ##ctx) \ 5713 FN(xdp_adjust_tail, 65, ##ctx) \ 5714 FN(skb_get_xfrm_state, 66, ##ctx) \ 5715 FN(get_stack, 67, ##ctx) \ 5716 FN(skb_load_bytes_relative, 68, ##ctx) \ 5717 FN(fib_lookup, 69, ##ctx) \ 5718 FN(sock_hash_update, 70, ##ctx) \ 5719 FN(msg_redirect_hash, 71, ##ctx) \ 5720 FN(sk_redirect_hash, 72, ##ctx) \ 5721 FN(lwt_push_encap, 73, ##ctx) \ 5722 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5723 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5724 FN(lwt_seg6_action, 76, ##ctx) \ 5725 FN(rc_repeat, 77, ##ctx) \ 5726 FN(rc_keydown, 78, ##ctx) \ 5727 FN(skb_cgroup_id, 79, ##ctx) \ 5728 FN(get_current_cgroup_id, 80, ##ctx) \ 5729 FN(get_local_storage, 81, ##ctx) \ 5730 FN(sk_select_reuseport, 82, ##ctx) \ 5731 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5732 FN(sk_lookup_tcp, 84, ##ctx) \ 5733 FN(sk_lookup_udp, 85, ##ctx) \ 5734 FN(sk_release, 86, ##ctx) \ 5735 FN(map_push_elem, 87, ##ctx) \ 5736 FN(map_pop_elem, 88, ##ctx) \ 5737 FN(map_peek_elem, 89, ##ctx) \ 5738 FN(msg_push_data, 90, ##ctx) \ 5739 FN(msg_pop_data, 91, ##ctx) \ 5740 FN(rc_pointer_rel, 92, ##ctx) \ 5741 FN(spin_lock, 93, ##ctx) \ 5742 FN(spin_unlock, 94, ##ctx) \ 5743 FN(sk_fullsock, 95, ##ctx) \ 5744 FN(tcp_sock, 96, ##ctx) \ 5745 FN(skb_ecn_set_ce, 97, ##ctx) \ 5746 FN(get_listener_sock, 98, ##ctx) \ 5747 FN(skc_lookup_tcp, 99, ##ctx) \ 5748 FN(tcp_check_syncookie, 100, ##ctx) \ 5749 FN(sysctl_get_name, 101, ##ctx) \ 5750 FN(sysctl_get_current_value, 102, ##ctx) \ 5751 FN(sysctl_get_new_value, 103, ##ctx) \ 5752 FN(sysctl_set_new_value, 104, ##ctx) \ 5753 FN(strtol, 105, ##ctx) \ 5754 FN(strtoul, 106, ##ctx) \ 5755 FN(sk_storage_get, 107, ##ctx) \ 5756 FN(sk_storage_delete, 108, ##ctx) \ 5757 FN(send_signal, 109, ##ctx) \ 5758 FN(tcp_gen_syncookie, 110, ##ctx) \ 5759 FN(skb_output, 111, ##ctx) \ 5760 FN(probe_read_user, 112, ##ctx) \ 5761 FN(probe_read_kernel, 113, ##ctx) \ 5762 FN(probe_read_user_str, 114, ##ctx) \ 5763 FN(probe_read_kernel_str, 115, ##ctx) \ 5764 FN(tcp_send_ack, 116, ##ctx) \ 5765 FN(send_signal_thread, 117, ##ctx) \ 5766 FN(jiffies64, 118, ##ctx) \ 5767 FN(read_branch_records, 119, ##ctx) \ 5768 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5769 FN(xdp_output, 121, ##ctx) \ 5770 FN(get_netns_cookie, 122, ##ctx) \ 5771 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5772 FN(sk_assign, 124, ##ctx) \ 5773 FN(ktime_get_boot_ns, 125, ##ctx) \ 5774 FN(seq_printf, 126, ##ctx) \ 5775 FN(seq_write, 127, ##ctx) \ 5776 FN(sk_cgroup_id, 128, ##ctx) \ 5777 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5778 FN(ringbuf_output, 130, ##ctx) \ 5779 FN(ringbuf_reserve, 131, ##ctx) \ 5780 FN(ringbuf_submit, 132, ##ctx) \ 5781 FN(ringbuf_discard, 133, ##ctx) \ 5782 FN(ringbuf_query, 134, ##ctx) \ 5783 FN(csum_level, 135, ##ctx) \ 5784 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5785 FN(skc_to_tcp_sock, 137, ##ctx) \ 5786 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5787 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5788 FN(skc_to_udp6_sock, 140, ##ctx) \ 5789 FN(get_task_stack, 141, ##ctx) \ 5790 FN(load_hdr_opt, 142, ##ctx) \ 5791 FN(store_hdr_opt, 143, ##ctx) \ 5792 FN(reserve_hdr_opt, 144, ##ctx) \ 5793 FN(inode_storage_get, 145, ##ctx) \ 5794 FN(inode_storage_delete, 146, ##ctx) \ 5795 FN(d_path, 147, ##ctx) \ 5796 FN(copy_from_user, 148, ##ctx) \ 5797 FN(snprintf_btf, 149, ##ctx) \ 5798 FN(seq_printf_btf, 150, ##ctx) \ 5799 FN(skb_cgroup_classid, 151, ##ctx) \ 5800 FN(redirect_neigh, 152, ##ctx) \ 5801 FN(per_cpu_ptr, 153, ##ctx) \ 5802 FN(this_cpu_ptr, 154, ##ctx) \ 5803 FN(redirect_peer, 155, ##ctx) \ 5804 FN(task_storage_get, 156, ##ctx) \ 5805 FN(task_storage_delete, 157, ##ctx) \ 5806 FN(get_current_task_btf, 158, ##ctx) \ 5807 FN(bprm_opts_set, 159, ##ctx) \ 5808 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5809 FN(ima_inode_hash, 161, ##ctx) \ 5810 FN(sock_from_file, 162, ##ctx) \ 5811 FN(check_mtu, 163, ##ctx) \ 5812 FN(for_each_map_elem, 164, ##ctx) \ 5813 FN(snprintf, 165, ##ctx) \ 5814 FN(sys_bpf, 166, ##ctx) \ 5815 FN(btf_find_by_name_kind, 167, ##ctx) \ 5816 FN(sys_close, 168, ##ctx) \ 5817 FN(timer_init, 169, ##ctx) \ 5818 FN(timer_set_callback, 170, ##ctx) \ 5819 FN(timer_start, 171, ##ctx) \ 5820 FN(timer_cancel, 172, ##ctx) \ 5821 FN(get_func_ip, 173, ##ctx) \ 5822 FN(get_attach_cookie, 174, ##ctx) \ 5823 FN(task_pt_regs, 175, ##ctx) \ 5824 FN(get_branch_snapshot, 176, ##ctx) \ 5825 FN(trace_vprintk, 177, ##ctx) \ 5826 FN(skc_to_unix_sock, 178, ##ctx) \ 5827 FN(kallsyms_lookup_name, 179, ##ctx) \ 5828 FN(find_vma, 180, ##ctx) \ 5829 FN(loop, 181, ##ctx) \ 5830 FN(strncmp, 182, ##ctx) \ 5831 FN(get_func_arg, 183, ##ctx) \ 5832 FN(get_func_ret, 184, ##ctx) \ 5833 FN(get_func_arg_cnt, 185, ##ctx) \ 5834 FN(get_retval, 186, ##ctx) \ 5835 FN(set_retval, 187, ##ctx) \ 5836 FN(xdp_get_buff_len, 188, ##ctx) \ 5837 FN(xdp_load_bytes, 189, ##ctx) \ 5838 FN(xdp_store_bytes, 190, ##ctx) \ 5839 FN(copy_from_user_task, 191, ##ctx) \ 5840 FN(skb_set_tstamp, 192, ##ctx) \ 5841 FN(ima_file_hash, 193, ##ctx) \ 5842 FN(kptr_xchg, 194, ##ctx) \ 5843 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5844 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5845 FN(dynptr_from_mem, 197, ##ctx) \ 5846 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5847 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5848 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5849 FN(dynptr_read, 201, ##ctx) \ 5850 FN(dynptr_write, 202, ##ctx) \ 5851 FN(dynptr_data, 203, ##ctx) \ 5852 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5853 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5854 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5855 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5856 FN(ktime_get_tai_ns, 208, ##ctx) \ 5857 FN(user_ringbuf_drain, 209, ##ctx) \ 5858 FN(cgrp_storage_get, 210, ##ctx) \ 5859 FN(cgrp_storage_delete, 211, ##ctx) \ 5860 /* */ 5861 5862 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 5863 * know or care about integer value that is now passed as second argument 5864 */ 5865 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 5866 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 5867 5868 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5869 * function eBPF program intends to call 5870 */ 5871 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 5872 enum bpf_func_id { 5873 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5874 __BPF_FUNC_MAX_ID, 5875 }; 5876 #undef __BPF_ENUM_FN 5877 5878 /* All flags used by eBPF helper functions, placed here. */ 5879 5880 /* BPF_FUNC_skb_store_bytes flags. */ 5881 enum { 5882 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5883 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5884 }; 5885 5886 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5887 * First 4 bits are for passing the header field size. 5888 */ 5889 enum { 5890 BPF_F_HDR_FIELD_MASK = 0xfULL, 5891 }; 5892 5893 /* BPF_FUNC_l4_csum_replace flags. */ 5894 enum { 5895 BPF_F_PSEUDO_HDR = (1ULL << 4), 5896 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5897 BPF_F_MARK_ENFORCE = (1ULL << 6), 5898 }; 5899 5900 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5901 enum { 5902 BPF_F_INGRESS = (1ULL << 0), 5903 }; 5904 5905 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5906 enum { 5907 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5908 }; 5909 5910 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5911 enum { 5912 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5913 BPF_F_USER_STACK = (1ULL << 8), 5914 /* flags used by BPF_FUNC_get_stackid only. */ 5915 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5916 BPF_F_REUSE_STACKID = (1ULL << 10), 5917 /* flags used by BPF_FUNC_get_stack only. */ 5918 BPF_F_USER_BUILD_ID = (1ULL << 11), 5919 }; 5920 5921 /* BPF_FUNC_skb_set_tunnel_key flags. */ 5922 enum { 5923 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5924 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5925 BPF_F_SEQ_NUMBER = (1ULL << 3), 5926 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 5927 }; 5928 5929 /* BPF_FUNC_skb_get_tunnel_key flags. */ 5930 enum { 5931 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 5932 }; 5933 5934 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5935 * BPF_FUNC_perf_event_read_value flags. 5936 */ 5937 enum { 5938 BPF_F_INDEX_MASK = 0xffffffffULL, 5939 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5940 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 5941 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5942 }; 5943 5944 /* Current network namespace */ 5945 enum { 5946 BPF_F_CURRENT_NETNS = (-1L), 5947 }; 5948 5949 /* BPF_FUNC_csum_level level values. */ 5950 enum { 5951 BPF_CSUM_LEVEL_QUERY, 5952 BPF_CSUM_LEVEL_INC, 5953 BPF_CSUM_LEVEL_DEC, 5954 BPF_CSUM_LEVEL_RESET, 5955 }; 5956 5957 /* BPF_FUNC_skb_adjust_room flags. */ 5958 enum { 5959 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5960 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5961 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5962 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5963 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5964 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5965 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5966 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 5967 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 5968 }; 5969 5970 enum { 5971 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5972 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5973 }; 5974 5975 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5976 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5977 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5978 5979 /* BPF_FUNC_sysctl_get_name flags. */ 5980 enum { 5981 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5982 }; 5983 5984 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 5985 enum { 5986 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5987 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5988 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5989 */ 5990 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5991 }; 5992 5993 /* BPF_FUNC_read_branch_records flags. */ 5994 enum { 5995 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5996 }; 5997 5998 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5999 * BPF_FUNC_bpf_ringbuf_output flags. 6000 */ 6001 enum { 6002 BPF_RB_NO_WAKEUP = (1ULL << 0), 6003 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6004 }; 6005 6006 /* BPF_FUNC_bpf_ringbuf_query flags */ 6007 enum { 6008 BPF_RB_AVAIL_DATA = 0, 6009 BPF_RB_RING_SIZE = 1, 6010 BPF_RB_CONS_POS = 2, 6011 BPF_RB_PROD_POS = 3, 6012 }; 6013 6014 /* BPF ring buffer constants */ 6015 enum { 6016 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6017 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6018 BPF_RINGBUF_HDR_SZ = 8, 6019 }; 6020 6021 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6022 enum { 6023 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6024 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6025 }; 6026 6027 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6028 enum bpf_adj_room_mode { 6029 BPF_ADJ_ROOM_NET, 6030 BPF_ADJ_ROOM_MAC, 6031 }; 6032 6033 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6034 enum bpf_hdr_start_off { 6035 BPF_HDR_START_MAC, 6036 BPF_HDR_START_NET, 6037 }; 6038 6039 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6040 enum bpf_lwt_encap_mode { 6041 BPF_LWT_ENCAP_SEG6, 6042 BPF_LWT_ENCAP_SEG6_INLINE, 6043 BPF_LWT_ENCAP_IP, 6044 }; 6045 6046 /* Flags for bpf_bprm_opts_set helper */ 6047 enum { 6048 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6049 }; 6050 6051 /* Flags for bpf_redirect_map helper */ 6052 enum { 6053 BPF_F_BROADCAST = (1ULL << 3), 6054 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 6055 }; 6056 6057 #define __bpf_md_ptr(type, name) \ 6058 union { \ 6059 type name; \ 6060 __u64 :64; \ 6061 } __attribute__((aligned(8))) 6062 6063 enum { 6064 BPF_SKB_TSTAMP_UNSPEC, 6065 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 6066 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 6067 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 6068 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 6069 */ 6070 }; 6071 6072 /* user accessible mirror of in-kernel sk_buff. 6073 * new fields can only be added to the end of this structure 6074 */ 6075 struct __sk_buff { 6076 __u32 len; 6077 __u32 pkt_type; 6078 __u32 mark; 6079 __u32 queue_mapping; 6080 __u32 protocol; 6081 __u32 vlan_present; 6082 __u32 vlan_tci; 6083 __u32 vlan_proto; 6084 __u32 priority; 6085 __u32 ingress_ifindex; 6086 __u32 ifindex; 6087 __u32 tc_index; 6088 __u32 cb[5]; 6089 __u32 hash; 6090 __u32 tc_classid; 6091 __u32 data; 6092 __u32 data_end; 6093 __u32 napi_id; 6094 6095 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6096 __u32 family; 6097 __u32 remote_ip4; /* Stored in network byte order */ 6098 __u32 local_ip4; /* Stored in network byte order */ 6099 __u32 remote_ip6[4]; /* Stored in network byte order */ 6100 __u32 local_ip6[4]; /* Stored in network byte order */ 6101 __u32 remote_port; /* Stored in network byte order */ 6102 __u32 local_port; /* stored in host byte order */ 6103 /* ... here. */ 6104 6105 __u32 data_meta; 6106 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6107 __u64 tstamp; 6108 __u32 wire_len; 6109 __u32 gso_segs; 6110 __bpf_md_ptr(struct bpf_sock *, sk); 6111 __u32 gso_size; 6112 __u8 tstamp_type; 6113 __u32 :24; /* Padding, future use. */ 6114 __u64 hwtstamp; 6115 }; 6116 6117 struct bpf_tunnel_key { 6118 __u32 tunnel_id; 6119 union { 6120 __u32 remote_ipv4; 6121 __u32 remote_ipv6[4]; 6122 }; 6123 __u8 tunnel_tos; 6124 __u8 tunnel_ttl; 6125 union { 6126 __u16 tunnel_ext; /* compat */ 6127 __be16 tunnel_flags; 6128 }; 6129 __u32 tunnel_label; 6130 union { 6131 __u32 local_ipv4; 6132 __u32 local_ipv6[4]; 6133 }; 6134 }; 6135 6136 /* user accessible mirror of in-kernel xfrm_state. 6137 * new fields can only be added to the end of this structure 6138 */ 6139 struct bpf_xfrm_state { 6140 __u32 reqid; 6141 __u32 spi; /* Stored in network byte order */ 6142 __u16 family; 6143 __u16 ext; /* Padding, future use. */ 6144 union { 6145 __u32 remote_ipv4; /* Stored in network byte order */ 6146 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6147 }; 6148 }; 6149 6150 /* Generic BPF return codes which all BPF program types may support. 6151 * The values are binary compatible with their TC_ACT_* counter-part to 6152 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6153 * programs. 6154 * 6155 * XDP is handled seprately, see XDP_*. 6156 */ 6157 enum bpf_ret_code { 6158 BPF_OK = 0, 6159 /* 1 reserved */ 6160 BPF_DROP = 2, 6161 /* 3-6 reserved */ 6162 BPF_REDIRECT = 7, 6163 /* >127 are reserved for prog type specific return codes. 6164 * 6165 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6166 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6167 * changed and should be routed based on its new L3 header. 6168 * (This is an L3 redirect, as opposed to L2 redirect 6169 * represented by BPF_REDIRECT above). 6170 */ 6171 BPF_LWT_REROUTE = 128, 6172 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6173 * to indicate that no custom dissection was performed, and 6174 * fallback to standard dissector is requested. 6175 */ 6176 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6177 }; 6178 6179 struct bpf_sock { 6180 __u32 bound_dev_if; 6181 __u32 family; 6182 __u32 type; 6183 __u32 protocol; 6184 __u32 mark; 6185 __u32 priority; 6186 /* IP address also allows 1 and 2 bytes access */ 6187 __u32 src_ip4; 6188 __u32 src_ip6[4]; 6189 __u32 src_port; /* host byte order */ 6190 __be16 dst_port; /* network byte order */ 6191 __u16 :16; /* zero padding */ 6192 __u32 dst_ip4; 6193 __u32 dst_ip6[4]; 6194 __u32 state; 6195 __s32 rx_queue_mapping; 6196 }; 6197 6198 struct bpf_tcp_sock { 6199 __u32 snd_cwnd; /* Sending congestion window */ 6200 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6201 __u32 rtt_min; 6202 __u32 snd_ssthresh; /* Slow start size threshold */ 6203 __u32 rcv_nxt; /* What we want to receive next */ 6204 __u32 snd_nxt; /* Next sequence we send */ 6205 __u32 snd_una; /* First byte we want an ack for */ 6206 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6207 __u32 ecn_flags; /* ECN status bits. */ 6208 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6209 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6210 __u32 packets_out; /* Packets which are "in flight" */ 6211 __u32 retrans_out; /* Retransmitted packets out */ 6212 __u32 total_retrans; /* Total retransmits for entire connection */ 6213 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6214 * total number of segments in. 6215 */ 6216 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6217 * total number of data segments in. 6218 */ 6219 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6220 * The total number of segments sent. 6221 */ 6222 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6223 * total number of data segments sent. 6224 */ 6225 __u32 lost_out; /* Lost packets */ 6226 __u32 sacked_out; /* SACK'd packets */ 6227 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6228 * sum(delta(rcv_nxt)), or how many bytes 6229 * were acked. 6230 */ 6231 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6232 * sum(delta(snd_una)), or how many bytes 6233 * were acked. 6234 */ 6235 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6236 * total number of DSACK blocks received 6237 */ 6238 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6239 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6240 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6241 }; 6242 6243 struct bpf_sock_tuple { 6244 union { 6245 struct { 6246 __be32 saddr; 6247 __be32 daddr; 6248 __be16 sport; 6249 __be16 dport; 6250 } ipv4; 6251 struct { 6252 __be32 saddr[4]; 6253 __be32 daddr[4]; 6254 __be16 sport; 6255 __be16 dport; 6256 } ipv6; 6257 }; 6258 }; 6259 6260 /* (Simplified) user return codes for tcx prog type. 6261 * A valid tcx program must return one of these defined values. All other 6262 * return codes are reserved for future use. Must remain compatible with 6263 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6264 * return codes are mapped to TCX_NEXT. 6265 */ 6266 enum tcx_action_base { 6267 TCX_NEXT = -1, 6268 TCX_PASS = 0, 6269 TCX_DROP = 2, 6270 TCX_REDIRECT = 7, 6271 }; 6272 6273 struct bpf_xdp_sock { 6274 __u32 queue_id; 6275 }; 6276 6277 #define XDP_PACKET_HEADROOM 256 6278 6279 /* User return codes for XDP prog type. 6280 * A valid XDP program must return one of these defined values. All other 6281 * return codes are reserved for future use. Unknown return codes will 6282 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6283 */ 6284 enum xdp_action { 6285 XDP_ABORTED = 0, 6286 XDP_DROP, 6287 XDP_PASS, 6288 XDP_TX, 6289 XDP_REDIRECT, 6290 }; 6291 6292 /* user accessible metadata for XDP packet hook 6293 * new fields must be added to the end of this structure 6294 */ 6295 struct xdp_md { 6296 __u32 data; 6297 __u32 data_end; 6298 __u32 data_meta; 6299 /* Below access go through struct xdp_rxq_info */ 6300 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6301 __u32 rx_queue_index; /* rxq->queue_index */ 6302 6303 __u32 egress_ifindex; /* txq->dev->ifindex */ 6304 }; 6305 6306 /* DEVMAP map-value layout 6307 * 6308 * The struct data-layout of map-value is a configuration interface. 6309 * New members can only be added to the end of this structure. 6310 */ 6311 struct bpf_devmap_val { 6312 __u32 ifindex; /* device index */ 6313 union { 6314 int fd; /* prog fd on map write */ 6315 __u32 id; /* prog id on map read */ 6316 } bpf_prog; 6317 }; 6318 6319 /* CPUMAP map-value layout 6320 * 6321 * The struct data-layout of map-value is a configuration interface. 6322 * New members can only be added to the end of this structure. 6323 */ 6324 struct bpf_cpumap_val { 6325 __u32 qsize; /* queue size to remote target CPU */ 6326 union { 6327 int fd; /* prog fd on map write */ 6328 __u32 id; /* prog id on map read */ 6329 } bpf_prog; 6330 }; 6331 6332 enum sk_action { 6333 SK_DROP = 0, 6334 SK_PASS, 6335 }; 6336 6337 /* user accessible metadata for SK_MSG packet hook, new fields must 6338 * be added to the end of this structure 6339 */ 6340 struct sk_msg_md { 6341 __bpf_md_ptr(void *, data); 6342 __bpf_md_ptr(void *, data_end); 6343 6344 __u32 family; 6345 __u32 remote_ip4; /* Stored in network byte order */ 6346 __u32 local_ip4; /* Stored in network byte order */ 6347 __u32 remote_ip6[4]; /* Stored in network byte order */ 6348 __u32 local_ip6[4]; /* Stored in network byte order */ 6349 __u32 remote_port; /* Stored in network byte order */ 6350 __u32 local_port; /* stored in host byte order */ 6351 __u32 size; /* Total size of sk_msg */ 6352 6353 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6354 }; 6355 6356 struct sk_reuseport_md { 6357 /* 6358 * Start of directly accessible data. It begins from 6359 * the tcp/udp header. 6360 */ 6361 __bpf_md_ptr(void *, data); 6362 /* End of directly accessible data */ 6363 __bpf_md_ptr(void *, data_end); 6364 /* 6365 * Total length of packet (starting from the tcp/udp header). 6366 * Note that the directly accessible bytes (data_end - data) 6367 * could be less than this "len". Those bytes could be 6368 * indirectly read by a helper "bpf_skb_load_bytes()". 6369 */ 6370 __u32 len; 6371 /* 6372 * Eth protocol in the mac header (network byte order). e.g. 6373 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6374 */ 6375 __u32 eth_protocol; 6376 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6377 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6378 __u32 hash; /* A hash of the packet 4 tuples */ 6379 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6380 * new incoming connection request (e.g. selecting a listen sk for 6381 * the received SYN in the TCP case). reuse->sk is one of the sk 6382 * in the reuseport group. The bpf prog can use reuse->sk to learn 6383 * the local listening ip/port without looking into the skb. 6384 * 6385 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6386 * reuse->migrating_sk is the socket that needs to be migrated 6387 * to another listening socket. migrating_sk could be a fullsock 6388 * sk that is fully established or a reqsk that is in-the-middle 6389 * of 3-way handshake. 6390 */ 6391 __bpf_md_ptr(struct bpf_sock *, sk); 6392 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6393 }; 6394 6395 #define BPF_TAG_SIZE 8 6396 6397 struct bpf_prog_info { 6398 __u32 type; 6399 __u32 id; 6400 __u8 tag[BPF_TAG_SIZE]; 6401 __u32 jited_prog_len; 6402 __u32 xlated_prog_len; 6403 __aligned_u64 jited_prog_insns; 6404 __aligned_u64 xlated_prog_insns; 6405 __u64 load_time; /* ns since boottime */ 6406 __u32 created_by_uid; 6407 __u32 nr_map_ids; 6408 __aligned_u64 map_ids; 6409 char name[BPF_OBJ_NAME_LEN]; 6410 __u32 ifindex; 6411 __u32 gpl_compatible:1; 6412 __u32 :31; /* alignment pad */ 6413 __u64 netns_dev; 6414 __u64 netns_ino; 6415 __u32 nr_jited_ksyms; 6416 __u32 nr_jited_func_lens; 6417 __aligned_u64 jited_ksyms; 6418 __aligned_u64 jited_func_lens; 6419 __u32 btf_id; 6420 __u32 func_info_rec_size; 6421 __aligned_u64 func_info; 6422 __u32 nr_func_info; 6423 __u32 nr_line_info; 6424 __aligned_u64 line_info; 6425 __aligned_u64 jited_line_info; 6426 __u32 nr_jited_line_info; 6427 __u32 line_info_rec_size; 6428 __u32 jited_line_info_rec_size; 6429 __u32 nr_prog_tags; 6430 __aligned_u64 prog_tags; 6431 __u64 run_time_ns; 6432 __u64 run_cnt; 6433 __u64 recursion_misses; 6434 __u32 verified_insns; 6435 __u32 attach_btf_obj_id; 6436 __u32 attach_btf_id; 6437 } __attribute__((aligned(8))); 6438 6439 struct bpf_map_info { 6440 __u32 type; 6441 __u32 id; 6442 __u32 key_size; 6443 __u32 value_size; 6444 __u32 max_entries; 6445 __u32 map_flags; 6446 char name[BPF_OBJ_NAME_LEN]; 6447 __u32 ifindex; 6448 __u32 btf_vmlinux_value_type_id; 6449 __u64 netns_dev; 6450 __u64 netns_ino; 6451 __u32 btf_id; 6452 __u32 btf_key_type_id; 6453 __u32 btf_value_type_id; 6454 __u32 :32; /* alignment pad */ 6455 __u64 map_extra; 6456 } __attribute__((aligned(8))); 6457 6458 struct bpf_btf_info { 6459 __aligned_u64 btf; 6460 __u32 btf_size; 6461 __u32 id; 6462 __aligned_u64 name; 6463 __u32 name_len; 6464 __u32 kernel_btf; 6465 } __attribute__((aligned(8))); 6466 6467 struct bpf_link_info { 6468 __u32 type; 6469 __u32 id; 6470 __u32 prog_id; 6471 union { 6472 struct { 6473 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6474 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6475 } raw_tracepoint; 6476 struct { 6477 __u32 attach_type; 6478 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6479 __u32 target_btf_id; /* BTF type id inside the object */ 6480 } tracing; 6481 struct { 6482 __u64 cgroup_id; 6483 __u32 attach_type; 6484 } cgroup; 6485 struct { 6486 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6487 __u32 target_name_len; /* in/out: target_name buffer len */ 6488 6489 /* If the iter specific field is 32 bits, it can be put 6490 * in the first or second union. Otherwise it should be 6491 * put in the second union. 6492 */ 6493 union { 6494 struct { 6495 __u32 map_id; 6496 } map; 6497 }; 6498 union { 6499 struct { 6500 __u64 cgroup_id; 6501 __u32 order; 6502 } cgroup; 6503 struct { 6504 __u32 tid; 6505 __u32 pid; 6506 } task; 6507 }; 6508 } iter; 6509 struct { 6510 __u32 netns_ino; 6511 __u32 attach_type; 6512 } netns; 6513 struct { 6514 __u32 ifindex; 6515 } xdp; 6516 struct { 6517 __u32 map_id; 6518 } struct_ops; 6519 struct { 6520 __u32 pf; 6521 __u32 hooknum; 6522 __s32 priority; 6523 __u32 flags; 6524 } netfilter; 6525 struct { 6526 __aligned_u64 addrs; 6527 __u32 count; /* in/out: kprobe_multi function count */ 6528 __u32 flags; 6529 } kprobe_multi; 6530 struct { 6531 __u32 type; /* enum bpf_perf_event_type */ 6532 __u32 :32; 6533 union { 6534 struct { 6535 __aligned_u64 file_name; /* in/out */ 6536 __u32 name_len; 6537 __u32 offset; /* offset from file_name */ 6538 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6539 struct { 6540 __aligned_u64 func_name; /* in/out */ 6541 __u32 name_len; 6542 __u32 offset; /* offset from func_name */ 6543 __u64 addr; 6544 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6545 struct { 6546 __aligned_u64 tp_name; /* in/out */ 6547 __u32 name_len; 6548 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6549 struct { 6550 __u64 config; 6551 __u32 type; 6552 } event; /* BPF_PERF_EVENT_EVENT */ 6553 }; 6554 } perf_event; 6555 struct { 6556 __u32 ifindex; 6557 __u32 attach_type; 6558 } tcx; 6559 }; 6560 } __attribute__((aligned(8))); 6561 6562 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6563 * by user and intended to be used by socket (e.g. to bind to, depends on 6564 * attach type). 6565 */ 6566 struct bpf_sock_addr { 6567 __u32 user_family; /* Allows 4-byte read, but no write. */ 6568 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6569 * Stored in network byte order. 6570 */ 6571 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6572 * Stored in network byte order. 6573 */ 6574 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6575 * Stored in network byte order 6576 */ 6577 __u32 family; /* Allows 4-byte read, but no write */ 6578 __u32 type; /* Allows 4-byte read, but no write */ 6579 __u32 protocol; /* Allows 4-byte read, but no write */ 6580 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6581 * Stored in network byte order. 6582 */ 6583 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6584 * Stored in network byte order. 6585 */ 6586 __bpf_md_ptr(struct bpf_sock *, sk); 6587 }; 6588 6589 /* User bpf_sock_ops struct to access socket values and specify request ops 6590 * and their replies. 6591 * Some of this fields are in network (bigendian) byte order and may need 6592 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6593 * New fields can only be added at the end of this structure 6594 */ 6595 struct bpf_sock_ops { 6596 __u32 op; 6597 union { 6598 __u32 args[4]; /* Optionally passed to bpf program */ 6599 __u32 reply; /* Returned by bpf program */ 6600 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6601 }; 6602 __u32 family; 6603 __u32 remote_ip4; /* Stored in network byte order */ 6604 __u32 local_ip4; /* Stored in network byte order */ 6605 __u32 remote_ip6[4]; /* Stored in network byte order */ 6606 __u32 local_ip6[4]; /* Stored in network byte order */ 6607 __u32 remote_port; /* Stored in network byte order */ 6608 __u32 local_port; /* stored in host byte order */ 6609 __u32 is_fullsock; /* Some TCP fields are only valid if 6610 * there is a full socket. If not, the 6611 * fields read as zero. 6612 */ 6613 __u32 snd_cwnd; 6614 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6615 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6616 __u32 state; 6617 __u32 rtt_min; 6618 __u32 snd_ssthresh; 6619 __u32 rcv_nxt; 6620 __u32 snd_nxt; 6621 __u32 snd_una; 6622 __u32 mss_cache; 6623 __u32 ecn_flags; 6624 __u32 rate_delivered; 6625 __u32 rate_interval_us; 6626 __u32 packets_out; 6627 __u32 retrans_out; 6628 __u32 total_retrans; 6629 __u32 segs_in; 6630 __u32 data_segs_in; 6631 __u32 segs_out; 6632 __u32 data_segs_out; 6633 __u32 lost_out; 6634 __u32 sacked_out; 6635 __u32 sk_txhash; 6636 __u64 bytes_received; 6637 __u64 bytes_acked; 6638 __bpf_md_ptr(struct bpf_sock *, sk); 6639 /* [skb_data, skb_data_end) covers the whole TCP header. 6640 * 6641 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6642 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6643 * header has not been written. 6644 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6645 * been written so far. 6646 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6647 * the 3WHS. 6648 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6649 * the 3WHS. 6650 * 6651 * bpf_load_hdr_opt() can also be used to read a particular option. 6652 */ 6653 __bpf_md_ptr(void *, skb_data); 6654 __bpf_md_ptr(void *, skb_data_end); 6655 __u32 skb_len; /* The total length of a packet. 6656 * It includes the header, options, 6657 * and payload. 6658 */ 6659 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6660 * an easy way to check for tcp_flags 6661 * without parsing skb_data. 6662 * 6663 * In particular, the skb_tcp_flags 6664 * will still be available in 6665 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6666 * the outgoing header has not 6667 * been written yet. 6668 */ 6669 __u64 skb_hwtstamp; 6670 }; 6671 6672 /* Definitions for bpf_sock_ops_cb_flags */ 6673 enum { 6674 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6675 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6676 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6677 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6678 /* Call bpf for all received TCP headers. The bpf prog will be 6679 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6680 * 6681 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6682 * for the header option related helpers that will be useful 6683 * to the bpf programs. 6684 * 6685 * It could be used at the client/active side (i.e. connect() side) 6686 * when the server told it that the server was in syncookie 6687 * mode and required the active side to resend the bpf-written 6688 * options. The active side can keep writing the bpf-options until 6689 * it received a valid packet from the server side to confirm 6690 * the earlier packet (and options) has been received. The later 6691 * example patch is using it like this at the active side when the 6692 * server is in syncookie mode. 6693 * 6694 * The bpf prog will usually turn this off in the common cases. 6695 */ 6696 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6697 /* Call bpf when kernel has received a header option that 6698 * the kernel cannot handle. The bpf prog will be called under 6699 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6700 * 6701 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6702 * for the header option related helpers that will be useful 6703 * to the bpf programs. 6704 */ 6705 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6706 /* Call bpf when the kernel is writing header options for the 6707 * outgoing packet. The bpf prog will first be called 6708 * to reserve space in a skb under 6709 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6710 * the bpf prog will be called to write the header option(s) 6711 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6712 * 6713 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6714 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6715 * related helpers that will be useful to the bpf programs. 6716 * 6717 * The kernel gets its chance to reserve space and write 6718 * options first before the BPF program does. 6719 */ 6720 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6721 /* Mask of all currently supported cb flags */ 6722 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6723 }; 6724 6725 /* List of known BPF sock_ops operators. 6726 * New entries can only be added at the end 6727 */ 6728 enum { 6729 BPF_SOCK_OPS_VOID, 6730 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6731 * -1 if default value should be used 6732 */ 6733 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6734 * window (in packets) or -1 if default 6735 * value should be used 6736 */ 6737 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6738 * active connection is initialized 6739 */ 6740 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6741 * active connection is 6742 * established 6743 */ 6744 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6745 * passive connection is 6746 * established 6747 */ 6748 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6749 * needs ECN 6750 */ 6751 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6752 * based on the path and may be 6753 * dependent on the congestion control 6754 * algorithm. In general it indicates 6755 * a congestion threshold. RTTs above 6756 * this indicate congestion 6757 */ 6758 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6759 * Arg1: value of icsk_retransmits 6760 * Arg2: value of icsk_rto 6761 * Arg3: whether RTO has expired 6762 */ 6763 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6764 * Arg1: sequence number of 1st byte 6765 * Arg2: # segments 6766 * Arg3: return value of 6767 * tcp_transmit_skb (0 => success) 6768 */ 6769 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6770 * Arg1: old_state 6771 * Arg2: new_state 6772 */ 6773 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6774 * socket transition to LISTEN state. 6775 */ 6776 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6777 */ 6778 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6779 * It will be called to handle 6780 * the packets received at 6781 * an already established 6782 * connection. 6783 * 6784 * sock_ops->skb_data: 6785 * Referring to the received skb. 6786 * It covers the TCP header only. 6787 * 6788 * bpf_load_hdr_opt() can also 6789 * be used to search for a 6790 * particular option. 6791 */ 6792 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6793 * header option later in 6794 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6795 * Arg1: bool want_cookie. (in 6796 * writing SYNACK only) 6797 * 6798 * sock_ops->skb_data: 6799 * Not available because no header has 6800 * been written yet. 6801 * 6802 * sock_ops->skb_tcp_flags: 6803 * The tcp_flags of the 6804 * outgoing skb. (e.g. SYN, ACK, FIN). 6805 * 6806 * bpf_reserve_hdr_opt() should 6807 * be used to reserve space. 6808 */ 6809 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6810 * Arg1: bool want_cookie. (in 6811 * writing SYNACK only) 6812 * 6813 * sock_ops->skb_data: 6814 * Referring to the outgoing skb. 6815 * It covers the TCP header 6816 * that has already been written 6817 * by the kernel and the 6818 * earlier bpf-progs. 6819 * 6820 * sock_ops->skb_tcp_flags: 6821 * The tcp_flags of the outgoing 6822 * skb. (e.g. SYN, ACK, FIN). 6823 * 6824 * bpf_store_hdr_opt() should 6825 * be used to write the 6826 * option. 6827 * 6828 * bpf_load_hdr_opt() can also 6829 * be used to search for a 6830 * particular option that 6831 * has already been written 6832 * by the kernel or the 6833 * earlier bpf-progs. 6834 */ 6835 }; 6836 6837 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6838 * changes between the TCP and BPF versions. Ideally this should never happen. 6839 * If it does, we need to add code to convert them before calling 6840 * the BPF sock_ops function. 6841 */ 6842 enum { 6843 BPF_TCP_ESTABLISHED = 1, 6844 BPF_TCP_SYN_SENT, 6845 BPF_TCP_SYN_RECV, 6846 BPF_TCP_FIN_WAIT1, 6847 BPF_TCP_FIN_WAIT2, 6848 BPF_TCP_TIME_WAIT, 6849 BPF_TCP_CLOSE, 6850 BPF_TCP_CLOSE_WAIT, 6851 BPF_TCP_LAST_ACK, 6852 BPF_TCP_LISTEN, 6853 BPF_TCP_CLOSING, /* Now a valid state */ 6854 BPF_TCP_NEW_SYN_RECV, 6855 6856 BPF_TCP_MAX_STATES /* Leave at the end! */ 6857 }; 6858 6859 enum { 6860 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6861 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6862 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6863 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6864 /* Copy the SYN pkt to optval 6865 * 6866 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6867 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6868 * to only getting from the saved_syn. It can either get the 6869 * syn packet from: 6870 * 6871 * 1. the just-received SYN packet (only available when writing the 6872 * SYNACK). It will be useful when it is not necessary to 6873 * save the SYN packet for latter use. It is also the only way 6874 * to get the SYN during syncookie mode because the syn 6875 * packet cannot be saved during syncookie. 6876 * 6877 * OR 6878 * 6879 * 2. the earlier saved syn which was done by 6880 * bpf_setsockopt(TCP_SAVE_SYN). 6881 * 6882 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 6883 * SYN packet is obtained. 6884 * 6885 * If the bpf-prog does not need the IP[46] header, the 6886 * bpf-prog can avoid parsing the IP header by using 6887 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 6888 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 6889 * 6890 * >0: Total number of bytes copied 6891 * -ENOSPC: Not enough space in optval. Only optlen number of 6892 * bytes is copied. 6893 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 6894 * is not saved by setsockopt(TCP_SAVE_SYN). 6895 */ 6896 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 6897 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6898 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 6899 }; 6900 6901 enum { 6902 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 6903 }; 6904 6905 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 6906 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6907 */ 6908 enum { 6909 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 6910 * total option spaces 6911 * required for an established 6912 * sk in order to calculate the 6913 * MSS. No skb is actually 6914 * sent. 6915 */ 6916 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 6917 * when sending a SYN. 6918 */ 6919 }; 6920 6921 struct bpf_perf_event_value { 6922 __u64 counter; 6923 __u64 enabled; 6924 __u64 running; 6925 }; 6926 6927 enum { 6928 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 6929 BPF_DEVCG_ACC_READ = (1ULL << 1), 6930 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 6931 }; 6932 6933 enum { 6934 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 6935 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 6936 }; 6937 6938 struct bpf_cgroup_dev_ctx { 6939 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 6940 __u32 access_type; 6941 __u32 major; 6942 __u32 minor; 6943 }; 6944 6945 struct bpf_raw_tracepoint_args { 6946 __u64 args[0]; 6947 }; 6948 6949 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 6950 * OUTPUT: Do lookup from egress perspective; default is ingress 6951 */ 6952 enum { 6953 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 6954 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 6955 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 6956 BPF_FIB_LOOKUP_TBID = (1U << 3), 6957 }; 6958 6959 enum { 6960 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 6961 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 6962 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 6963 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 6964 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 6965 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 6966 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 6967 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 6968 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6969 }; 6970 6971 struct bpf_fib_lookup { 6972 /* input: network family for lookup (AF_INET, AF_INET6) 6973 * output: network family of egress nexthop 6974 */ 6975 __u8 family; 6976 6977 /* set if lookup is to consider L4 data - e.g., FIB rules */ 6978 __u8 l4_protocol; 6979 __be16 sport; 6980 __be16 dport; 6981 6982 union { /* used for MTU check */ 6983 /* input to lookup */ 6984 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 6985 6986 /* output: MTU value */ 6987 __u16 mtu_result; 6988 }; 6989 /* input: L3 device index for lookup 6990 * output: device index from FIB lookup 6991 */ 6992 __u32 ifindex; 6993 6994 union { 6995 /* inputs to lookup */ 6996 __u8 tos; /* AF_INET */ 6997 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 6998 6999 /* output: metric of fib result (IPv4/IPv6 only) */ 7000 __u32 rt_metric; 7001 }; 7002 7003 union { 7004 __be32 ipv4_src; 7005 __u32 ipv6_src[4]; /* in6_addr; network order */ 7006 }; 7007 7008 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7009 * network header. output: bpf_fib_lookup sets to gateway address 7010 * if FIB lookup returns gateway route 7011 */ 7012 union { 7013 __be32 ipv4_dst; 7014 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7015 }; 7016 7017 union { 7018 struct { 7019 /* output */ 7020 __be16 h_vlan_proto; 7021 __be16 h_vlan_TCI; 7022 }; 7023 /* input: when accompanied with the 7024 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7025 * specific routing table to use for the fib lookup. 7026 */ 7027 __u32 tbid; 7028 }; 7029 7030 __u8 smac[6]; /* ETH_ALEN */ 7031 __u8 dmac[6]; /* ETH_ALEN */ 7032 }; 7033 7034 struct bpf_redir_neigh { 7035 /* network family for lookup (AF_INET, AF_INET6) */ 7036 __u32 nh_family; 7037 /* network address of nexthop; skips fib lookup to find gateway */ 7038 union { 7039 __be32 ipv4_nh; 7040 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7041 }; 7042 }; 7043 7044 /* bpf_check_mtu flags*/ 7045 enum bpf_check_mtu_flags { 7046 BPF_MTU_CHK_SEGS = (1U << 0), 7047 }; 7048 7049 enum bpf_check_mtu_ret { 7050 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7051 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7052 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7053 }; 7054 7055 enum bpf_task_fd_type { 7056 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7057 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7058 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7059 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7060 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7061 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7062 }; 7063 7064 enum { 7065 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7066 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7067 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7068 }; 7069 7070 struct bpf_flow_keys { 7071 __u16 nhoff; 7072 __u16 thoff; 7073 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7074 __u8 is_frag; 7075 __u8 is_first_frag; 7076 __u8 is_encap; 7077 __u8 ip_proto; 7078 __be16 n_proto; 7079 __be16 sport; 7080 __be16 dport; 7081 union { 7082 struct { 7083 __be32 ipv4_src; 7084 __be32 ipv4_dst; 7085 }; 7086 struct { 7087 __u32 ipv6_src[4]; /* in6_addr; network order */ 7088 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7089 }; 7090 }; 7091 __u32 flags; 7092 __be32 flow_label; 7093 }; 7094 7095 struct bpf_func_info { 7096 __u32 insn_off; 7097 __u32 type_id; 7098 }; 7099 7100 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7101 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7102 7103 struct bpf_line_info { 7104 __u32 insn_off; 7105 __u32 file_name_off; 7106 __u32 line_off; 7107 __u32 line_col; 7108 }; 7109 7110 struct bpf_spin_lock { 7111 __u32 val; 7112 }; 7113 7114 struct bpf_timer { 7115 __u64 :64; 7116 __u64 :64; 7117 } __attribute__((aligned(8))); 7118 7119 struct bpf_dynptr { 7120 __u64 :64; 7121 __u64 :64; 7122 } __attribute__((aligned(8))); 7123 7124 struct bpf_list_head { 7125 __u64 :64; 7126 __u64 :64; 7127 } __attribute__((aligned(8))); 7128 7129 struct bpf_list_node { 7130 __u64 :64; 7131 __u64 :64; 7132 __u64 :64; 7133 } __attribute__((aligned(8))); 7134 7135 struct bpf_rb_root { 7136 __u64 :64; 7137 __u64 :64; 7138 } __attribute__((aligned(8))); 7139 7140 struct bpf_rb_node { 7141 __u64 :64; 7142 __u64 :64; 7143 __u64 :64; 7144 __u64 :64; 7145 } __attribute__((aligned(8))); 7146 7147 struct bpf_refcount { 7148 __u32 :32; 7149 } __attribute__((aligned(4))); 7150 7151 struct bpf_sysctl { 7152 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7153 * Allows 1,2,4-byte read, but no write. 7154 */ 7155 __u32 file_pos; /* Sysctl file position to read from, write to. 7156 * Allows 1,2,4-byte read an 4-byte write. 7157 */ 7158 }; 7159 7160 struct bpf_sockopt { 7161 __bpf_md_ptr(struct bpf_sock *, sk); 7162 __bpf_md_ptr(void *, optval); 7163 __bpf_md_ptr(void *, optval_end); 7164 7165 __s32 level; 7166 __s32 optname; 7167 __s32 optlen; 7168 __s32 retval; 7169 }; 7170 7171 struct bpf_pidns_info { 7172 __u32 pid; 7173 __u32 tgid; 7174 }; 7175 7176 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7177 struct bpf_sk_lookup { 7178 union { 7179 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7180 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7181 }; 7182 7183 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7184 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7185 __u32 remote_ip4; /* Network byte order */ 7186 __u32 remote_ip6[4]; /* Network byte order */ 7187 __be16 remote_port; /* Network byte order */ 7188 __u16 :16; /* Zero padding */ 7189 __u32 local_ip4; /* Network byte order */ 7190 __u32 local_ip6[4]; /* Network byte order */ 7191 __u32 local_port; /* Host byte order */ 7192 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7193 }; 7194 7195 /* 7196 * struct btf_ptr is used for typed pointer representation; the 7197 * type id is used to render the pointer data as the appropriate type 7198 * via the bpf_snprintf_btf() helper described above. A flags field - 7199 * potentially to specify additional details about the BTF pointer 7200 * (rather than its mode of display) - is included for future use. 7201 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7202 */ 7203 struct btf_ptr { 7204 void *ptr; 7205 __u32 type_id; 7206 __u32 flags; /* BTF ptr flags; unused at present. */ 7207 }; 7208 7209 /* 7210 * Flags to control bpf_snprintf_btf() behaviour. 7211 * - BTF_F_COMPACT: no formatting around type information 7212 * - BTF_F_NONAME: no struct/union member names/types 7213 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7214 * equivalent to %px. 7215 * - BTF_F_ZERO: show zero-valued struct/union members; they 7216 * are not displayed by default 7217 */ 7218 enum { 7219 BTF_F_COMPACT = (1ULL << 0), 7220 BTF_F_NONAME = (1ULL << 1), 7221 BTF_F_PTR_RAW = (1ULL << 2), 7222 BTF_F_ZERO = (1ULL << 3), 7223 }; 7224 7225 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7226 * has to be adjusted by relocations. It is emitted by llvm and passed to 7227 * libbpf and later to the kernel. 7228 */ 7229 enum bpf_core_relo_kind { 7230 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7231 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7232 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7233 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7234 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7235 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7236 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7237 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7238 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7239 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7240 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7241 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7242 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7243 }; 7244 7245 /* 7246 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7247 * and from libbpf to the kernel. 7248 * 7249 * CO-RE relocation captures the following data: 7250 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7251 * its insn->imm field to be relocated with actual field info; 7252 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7253 * type or field; 7254 * - access_str_off - offset into corresponding .BTF string section. String 7255 * interpretation depends on specific relocation kind: 7256 * - for field-based relocations, string encodes an accessed field using 7257 * a sequence of field and array indices, separated by colon (:). It's 7258 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7259 * arguments for identifying offset to a field. 7260 * - for type-based relocations, strings is expected to be just "0"; 7261 * - for enum value-based relocations, string contains an index of enum 7262 * value within its enum type; 7263 * - kind - one of enum bpf_core_relo_kind; 7264 * 7265 * Example: 7266 * struct sample { 7267 * int a; 7268 * struct { 7269 * int b[10]; 7270 * }; 7271 * }; 7272 * 7273 * struct sample *s = ...; 7274 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7275 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7276 * // b is field #0 inside anon struct, accessing elem #5) 7277 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7278 * 7279 * type_id for all relocs in this example will capture BTF type id of 7280 * `struct sample`. 7281 * 7282 * Such relocation is emitted when using __builtin_preserve_access_index() 7283 * Clang built-in, passing expression that captures field address, e.g.: 7284 * 7285 * bpf_probe_read(&dst, sizeof(dst), 7286 * __builtin_preserve_access_index(&src->a.b.c)); 7287 * 7288 * In this case Clang will emit field relocation recording necessary data to 7289 * be able to find offset of embedded `a.b.c` field within `src` struct. 7290 * 7291 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7292 */ 7293 struct bpf_core_relo { 7294 __u32 insn_off; 7295 __u32 type_id; 7296 __u32 access_str_off; 7297 enum bpf_core_relo_kind kind; 7298 }; 7299 7300 /* 7301 * Flags to control bpf_timer_start() behaviour. 7302 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7303 * relative to current time. 7304 */ 7305 enum { 7306 BPF_F_TIMER_ABS = (1ULL << 0), 7307 }; 7308 7309 /* BPF numbers iterator state */ 7310 struct bpf_iter_num { 7311 /* opaque iterator state; having __u64 here allows to preserve correct 7312 * alignment requirements in vmlinux.h, generated from BTF 7313 */ 7314 __u64 __opaque[1]; 7315 } __attribute__((aligned(8)));