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