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