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