Where Online Learning is simpler!
The C and C++ Include Header Files
/usr/include/linux/bpf.h
$ cat -n /usr/include/linux/bpf.h 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 __LINUX_BPF_H__ 9 #define __LINUX_BPF_H__ 10 11 #include
12 #include
13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_MEMSX 0x80 /* load with sign extension */ 23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 25 26 /* alu/jmp fields */ 27 #define BPF_MOV 0xb0 /* mov reg to reg */ 28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 29 30 /* change endianness of a register */ 31 #define BPF_END 0xd0 /* flags for endianness conversion: */ 32 #define BPF_TO_LE 0x00 /* convert to little-endian */ 33 #define BPF_TO_BE 0x08 /* convert to big-endian */ 34 #define BPF_FROM_LE BPF_TO_LE 35 #define BPF_FROM_BE BPF_TO_BE 36 37 /* jmp encodings */ 38 #define BPF_JNE 0x50 /* jump != */ 39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 45 #define BPF_CALL 0x80 /* function call */ 46 #define BPF_EXIT 0x90 /* function return */ 47 48 /* atomic op type fields (stored in immediate) */ 49 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 50 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 51 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 52 53 /* Register numbers */ 54 enum { 55 BPF_REG_0 = 0, 56 BPF_REG_1, 57 BPF_REG_2, 58 BPF_REG_3, 59 BPF_REG_4, 60 BPF_REG_5, 61 BPF_REG_6, 62 BPF_REG_7, 63 BPF_REG_8, 64 BPF_REG_9, 65 BPF_REG_10, 66 __MAX_BPF_REG, 67 }; 68 69 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 70 #define MAX_BPF_REG __MAX_BPF_REG 71 72 struct bpf_insn { 73 __u8 code; /* opcode */ 74 __u8 dst_reg:4; /* dest register */ 75 __u8 src_reg:4; /* source register */ 76 __s16 off; /* signed offset */ 77 __s32 imm; /* signed immediate constant */ 78 }; 79 80 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 81 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 82 * the trailing flexible array member) instead. 83 */ 84 struct bpf_lpm_trie_key { 85 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 86 __u8 data[0]; /* Arbitrary size */ 87 }; 88 89 /* Header for bpf_lpm_trie_key structs */ 90 struct bpf_lpm_trie_key_hdr { 91 __u32 prefixlen; 92 }; 93 94 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 95 struct bpf_lpm_trie_key_u8 { 96 union { 97 struct bpf_lpm_trie_key_hdr hdr; 98 __u32 prefixlen; 99 }; 100 __u8 data[]; /* Arbitrary size */ 101 }; 102 103 struct bpf_cgroup_storage_key { 104 __u64 cgroup_inode_id; /* cgroup inode id */ 105 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 106 }; 107 108 enum bpf_cgroup_iter_order { 109 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 110 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 111 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 112 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 113 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 114 }; 115 116 union bpf_iter_link_info { 117 struct { 118 __u32 map_fd; 119 } map; 120 struct { 121 enum bpf_cgroup_iter_order order; 122 123 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 124 * both are zero, the walk starts from the default cgroup v2 125 * root. For walking v1 hierarchy, one should always explicitly 126 * specify cgroup_fd. 127 */ 128 __u32 cgroup_fd; 129 __u64 cgroup_id; 130 } cgroup; 131 /* Parameters of task iterators. */ 132 struct { 133 __u32 tid; 134 __u32 pid; 135 __u32 pid_fd; 136 } task; 137 }; 138 139 /* BPF syscall commands, see bpf(2) man-page for more details. */ 140 /** 141 * DOC: eBPF Syscall Preamble 142 * 143 * The operation to be performed by the **bpf**\ () system call is determined 144 * by the *cmd* argument. Each operation takes an accompanying argument, 145 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 146 * below). The size argument is the size of the union pointed to by *attr*. 147 */ 148 /** 149 * DOC: eBPF Syscall Commands 150 * 151 * BPF_MAP_CREATE 152 * Description 153 * Create a map and return a file descriptor that refers to the 154 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 155 * is automatically enabled for the new file descriptor. 156 * 157 * Applying **close**\ (2) to the file descriptor returned by 158 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 159 * 160 * Return 161 * A new file descriptor (a nonnegative integer), or -1 if an 162 * error occurred (in which case, *errno* is set appropriately). 163 * 164 * BPF_MAP_LOOKUP_ELEM 165 * Description 166 * Look up an element with a given *key* in the map referred to 167 * by the file descriptor *map_fd*. 168 * 169 * The *flags* argument may be specified as one of the 170 * following: 171 * 172 * **BPF_F_LOCK** 173 * Look up the value of a spin-locked map without 174 * returning the lock. This must be specified if the 175 * elements contain a spinlock. 176 * 177 * Return 178 * Returns zero on success. On error, -1 is returned and *errno* 179 * is set appropriately. 180 * 181 * BPF_MAP_UPDATE_ELEM 182 * Description 183 * Create or update an element (key/value pair) in a specified map. 184 * 185 * The *flags* argument should be specified as one of the 186 * following: 187 * 188 * **BPF_ANY** 189 * Create a new element or update an existing element. 190 * **BPF_NOEXIST** 191 * Create a new element only if it did not exist. 192 * **BPF_EXIST** 193 * Update an existing element. 194 * **BPF_F_LOCK** 195 * Update a spin_lock-ed map element. 196 * 197 * Return 198 * Returns zero on success. On error, -1 is returned and *errno* 199 * is set appropriately. 200 * 201 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 202 * **E2BIG**, **EEXIST**, or **ENOENT**. 203 * 204 * **E2BIG** 205 * The number of elements in the map reached the 206 * *max_entries* limit specified at map creation time. 207 * **EEXIST** 208 * If *flags* specifies **BPF_NOEXIST** and the element 209 * with *key* already exists in the map. 210 * **ENOENT** 211 * If *flags* specifies **BPF_EXIST** and the element with 212 * *key* does not exist in the map. 213 * 214 * BPF_MAP_DELETE_ELEM 215 * Description 216 * Look up and delete an element by key in a specified map. 217 * 218 * Return 219 * Returns zero on success. On error, -1 is returned and *errno* 220 * is set appropriately. 221 * 222 * BPF_MAP_GET_NEXT_KEY 223 * Description 224 * Look up an element by key in a specified map and return the key 225 * of the next element. Can be used to iterate over all elements 226 * in the map. 227 * 228 * Return 229 * Returns zero on success. On error, -1 is returned and *errno* 230 * is set appropriately. 231 * 232 * The following cases can be used to iterate over all elements of 233 * the map: 234 * 235 * * If *key* is not found, the operation returns zero and sets 236 * the *next_key* pointer to the key of the first element. 237 * * If *key* is found, the operation returns zero and sets the 238 * *next_key* pointer to the key of the next element. 239 * * If *key* is the last element, returns -1 and *errno* is set 240 * to **ENOENT**. 241 * 242 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 243 * **EINVAL** on error. 244 * 245 * BPF_PROG_LOAD 246 * Description 247 * Verify and load an eBPF program, returning a new file 248 * descriptor associated with the program. 249 * 250 * Applying **close**\ (2) to the file descriptor returned by 251 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 252 * 253 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 254 * automatically enabled for the new file descriptor. 255 * 256 * Return 257 * A new file descriptor (a nonnegative integer), or -1 if an 258 * error occurred (in which case, *errno* is set appropriately). 259 * 260 * BPF_OBJ_PIN 261 * Description 262 * Pin an eBPF program or map referred by the specified *bpf_fd* 263 * to the provided *pathname* on the filesystem. 264 * 265 * The *pathname* argument must not contain a dot ("."). 266 * 267 * On success, *pathname* retains a reference to the eBPF object, 268 * preventing deallocation of the object when the original 269 * *bpf_fd* is closed. This allow the eBPF object to live beyond 270 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 271 * process. 272 * 273 * Applying **unlink**\ (2) or similar calls to the *pathname* 274 * unpins the object from the filesystem, removing the reference. 275 * If no other file descriptors or filesystem nodes refer to the 276 * same object, it will be deallocated (see NOTES). 277 * 278 * The filesystem type for the parent directory of *pathname* must 279 * be **BPF_FS_MAGIC**. 280 * 281 * Return 282 * Returns zero on success. On error, -1 is returned and *errno* 283 * is set appropriately. 284 * 285 * BPF_OBJ_GET 286 * Description 287 * Open a file descriptor for the eBPF object pinned to the 288 * specified *pathname*. 289 * 290 * Return 291 * A new file descriptor (a nonnegative integer), or -1 if an 292 * error occurred (in which case, *errno* is set appropriately). 293 * 294 * BPF_PROG_ATTACH 295 * Description 296 * Attach an eBPF program to a *target_fd* at the specified 297 * *attach_type* hook. 298 * 299 * The *attach_type* specifies the eBPF attachment point to 300 * attach the program to, and must be one of *bpf_attach_type* 301 * (see below). 302 * 303 * The *attach_bpf_fd* must be a valid file descriptor for a 304 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 305 * or sock_ops type corresponding to the specified *attach_type*. 306 * 307 * The *target_fd* must be a valid file descriptor for a kernel 308 * object which depends on the attach type of *attach_bpf_fd*: 309 * 310 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 311 * **BPF_PROG_TYPE_CGROUP_SKB**, 312 * **BPF_PROG_TYPE_CGROUP_SOCK**, 313 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 314 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 315 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 316 * **BPF_PROG_TYPE_SOCK_OPS** 317 * 318 * Control Group v2 hierarchy with the eBPF controller 319 * enabled. Requires the kernel to be compiled with 320 * **CONFIG_CGROUP_BPF**. 321 * 322 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 323 * 324 * Network namespace (eg /proc/self/ns/net). 325 * 326 * **BPF_PROG_TYPE_LIRC_MODE2** 327 * 328 * LIRC device path (eg /dev/lircN). Requires the kernel 329 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 330 * 331 * **BPF_PROG_TYPE_SK_SKB**, 332 * **BPF_PROG_TYPE_SK_MSG** 333 * 334 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 335 * 336 * Return 337 * Returns zero on success. On error, -1 is returned and *errno* 338 * is set appropriately. 339 * 340 * BPF_PROG_DETACH 341 * Description 342 * Detach the eBPF program associated with the *target_fd* at the 343 * hook specified by *attach_type*. The program must have been 344 * previously attached using **BPF_PROG_ATTACH**. 345 * 346 * Return 347 * Returns zero on success. On error, -1 is returned and *errno* 348 * is set appropriately. 349 * 350 * BPF_PROG_TEST_RUN 351 * Description 352 * Run the eBPF program associated with the *prog_fd* a *repeat* 353 * number of times against a provided program context *ctx_in* and 354 * data *data_in*, and return the modified program context 355 * *ctx_out*, *data_out* (for example, packet data), result of the 356 * execution *retval*, and *duration* of the test run. 357 * 358 * The sizes of the buffers provided as input and output 359 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 360 * be provided in the corresponding variables *ctx_size_in*, 361 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 362 * of these parameters are not provided (ie set to NULL), the 363 * corresponding size field must be zero. 364 * 365 * Some program types have particular requirements: 366 * 367 * **BPF_PROG_TYPE_SK_LOOKUP** 368 * *data_in* and *data_out* must be NULL. 369 * 370 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 371 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 372 * 373 * *ctx_out*, *data_in* and *data_out* must be NULL. 374 * *repeat* must be zero. 375 * 376 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 377 * 378 * Return 379 * Returns zero on success. On error, -1 is returned and *errno* 380 * is set appropriately. 381 * 382 * **ENOSPC** 383 * Either *data_size_out* or *ctx_size_out* is too small. 384 * **ENOTSUPP** 385 * This command is not supported by the program type of 386 * the program referred to by *prog_fd*. 387 * 388 * BPF_PROG_GET_NEXT_ID 389 * Description 390 * Fetch the next eBPF program currently loaded into the kernel. 391 * 392 * Looks for the eBPF program with an id greater than *start_id* 393 * and updates *next_id* on success. If no other eBPF programs 394 * remain with ids higher than *start_id*, returns -1 and sets 395 * *errno* to **ENOENT**. 396 * 397 * Return 398 * Returns zero on success. On error, or when no id remains, -1 399 * is returned and *errno* is set appropriately. 400 * 401 * BPF_MAP_GET_NEXT_ID 402 * Description 403 * Fetch the next eBPF map currently loaded into the kernel. 404 * 405 * Looks for the eBPF map with an id greater than *start_id* 406 * and updates *next_id* on success. If no other eBPF maps 407 * remain with ids higher than *start_id*, returns -1 and sets 408 * *errno* to **ENOENT**. 409 * 410 * Return 411 * Returns zero on success. On error, or when no id remains, -1 412 * is returned and *errno* is set appropriately. 413 * 414 * BPF_PROG_GET_FD_BY_ID 415 * Description 416 * Open a file descriptor for the eBPF program corresponding to 417 * *prog_id*. 418 * 419 * Return 420 * A new file descriptor (a nonnegative integer), or -1 if an 421 * error occurred (in which case, *errno* is set appropriately). 422 * 423 * BPF_MAP_GET_FD_BY_ID 424 * Description 425 * Open a file descriptor for the eBPF map corresponding to 426 * *map_id*. 427 * 428 * Return 429 * A new file descriptor (a nonnegative integer), or -1 if an 430 * error occurred (in which case, *errno* is set appropriately). 431 * 432 * BPF_OBJ_GET_INFO_BY_FD 433 * Description 434 * Obtain information about the eBPF object corresponding to 435 * *bpf_fd*. 436 * 437 * Populates up to *info_len* bytes of *info*, which will be in 438 * one of the following formats depending on the eBPF object type 439 * of *bpf_fd*: 440 * 441 * * **struct bpf_prog_info** 442 * * **struct bpf_map_info** 443 * * **struct bpf_btf_info** 444 * * **struct bpf_link_info** 445 * 446 * Return 447 * Returns zero on success. On error, -1 is returned and *errno* 448 * is set appropriately. 449 * 450 * BPF_PROG_QUERY 451 * Description 452 * Obtain information about eBPF programs associated with the 453 * specified *attach_type* hook. 454 * 455 * The *target_fd* must be a valid file descriptor for a kernel 456 * object which depends on the attach type of *attach_bpf_fd*: 457 * 458 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 459 * **BPF_PROG_TYPE_CGROUP_SKB**, 460 * **BPF_PROG_TYPE_CGROUP_SOCK**, 461 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 462 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 463 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 464 * **BPF_PROG_TYPE_SOCK_OPS** 465 * 466 * Control Group v2 hierarchy with the eBPF controller 467 * enabled. Requires the kernel to be compiled with 468 * **CONFIG_CGROUP_BPF**. 469 * 470 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 471 * 472 * Network namespace (eg /proc/self/ns/net). 473 * 474 * **BPF_PROG_TYPE_LIRC_MODE2** 475 * 476 * LIRC device path (eg /dev/lircN). Requires the kernel 477 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 478 * 479 * **BPF_PROG_QUERY** always fetches the number of programs 480 * attached and the *attach_flags* which were used to attach those 481 * programs. Additionally, if *prog_ids* is nonzero and the number 482 * of attached programs is less than *prog_cnt*, populates 483 * *prog_ids* with the eBPF program ids of the programs attached 484 * at *target_fd*. 485 * 486 * The following flags may alter the result: 487 * 488 * **BPF_F_QUERY_EFFECTIVE** 489 * Only return information regarding programs which are 490 * currently effective at the specified *target_fd*. 491 * 492 * Return 493 * Returns zero on success. On error, -1 is returned and *errno* 494 * is set appropriately. 495 * 496 * BPF_RAW_TRACEPOINT_OPEN 497 * Description 498 * Attach an eBPF program to a tracepoint *name* to access kernel 499 * internal arguments of the tracepoint in their raw form. 500 * 501 * The *prog_fd* must be a valid file descriptor associated with 502 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 503 * 504 * No ABI guarantees are made about the content of tracepoint 505 * arguments exposed to the corresponding eBPF program. 506 * 507 * Applying **close**\ (2) to the file descriptor returned by 508 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 509 * 510 * Return 511 * A new file descriptor (a nonnegative integer), or -1 if an 512 * error occurred (in which case, *errno* is set appropriately). 513 * 514 * BPF_BTF_LOAD 515 * Description 516 * Verify and load BPF Type Format (BTF) metadata into the kernel, 517 * returning a new file descriptor associated with the metadata. 518 * BTF is described in more detail at 519 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 520 * 521 * The *btf* parameter must point to valid memory providing 522 * *btf_size* bytes of BTF binary metadata. 523 * 524 * The returned file descriptor can be passed to other **bpf**\ () 525 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 526 * associate the BTF with those objects. 527 * 528 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 529 * parameters to specify a *btf_log_buf*, *btf_log_size* and 530 * *btf_log_level* which allow the kernel to return freeform log 531 * output regarding the BTF verification process. 532 * 533 * Return 534 * A new file descriptor (a nonnegative integer), or -1 if an 535 * error occurred (in which case, *errno* is set appropriately). 536 * 537 * BPF_BTF_GET_FD_BY_ID 538 * Description 539 * Open a file descriptor for the BPF Type Format (BTF) 540 * corresponding to *btf_id*. 541 * 542 * Return 543 * A new file descriptor (a nonnegative integer), or -1 if an 544 * error occurred (in which case, *errno* is set appropriately). 545 * 546 * BPF_TASK_FD_QUERY 547 * Description 548 * Obtain information about eBPF programs associated with the 549 * target process identified by *pid* and *fd*. 550 * 551 * If the *pid* and *fd* are associated with a tracepoint, kprobe 552 * or uprobe perf event, then the *prog_id* and *fd_type* will 553 * be populated with the eBPF program id and file descriptor type 554 * of type **bpf_task_fd_type**. If associated with a kprobe or 555 * uprobe, the *probe_offset* and *probe_addr* will also be 556 * populated. Optionally, if *buf* is provided, then up to 557 * *buf_len* bytes of *buf* will be populated with the name of 558 * the tracepoint, kprobe or uprobe. 559 * 560 * The resulting *prog_id* may be introspected in deeper detail 561 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 562 * 563 * Return 564 * Returns zero on success. On error, -1 is returned and *errno* 565 * is set appropriately. 566 * 567 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 568 * Description 569 * Look up an element with the given *key* in the map referred to 570 * by the file descriptor *fd*, and if found, delete the element. 571 * 572 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 573 * types, the *flags* argument needs to be set to 0, but for other 574 * map types, it may be specified as: 575 * 576 * **BPF_F_LOCK** 577 * Look up and delete the value of a spin-locked map 578 * without returning the lock. This must be specified if 579 * the elements contain a spinlock. 580 * 581 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 582 * implement this command as a "pop" operation, deleting the top 583 * element rather than one corresponding to *key*. 584 * The *key* and *key_len* parameters should be zeroed when 585 * issuing this operation for these map types. 586 * 587 * This command is only valid for the following map types: 588 * * **BPF_MAP_TYPE_QUEUE** 589 * * **BPF_MAP_TYPE_STACK** 590 * * **BPF_MAP_TYPE_HASH** 591 * * **BPF_MAP_TYPE_PERCPU_HASH** 592 * * **BPF_MAP_TYPE_LRU_HASH** 593 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 594 * 595 * Return 596 * Returns zero on success. On error, -1 is returned and *errno* 597 * is set appropriately. 598 * 599 * BPF_MAP_FREEZE 600 * Description 601 * Freeze the permissions of the specified map. 602 * 603 * Write permissions may be frozen by passing zero *flags*. 604 * Upon success, no future syscall invocations may alter the 605 * map state of *map_fd*. Write operations from eBPF programs 606 * are still possible for a frozen map. 607 * 608 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 609 * 610 * Return 611 * Returns zero on success. On error, -1 is returned and *errno* 612 * is set appropriately. 613 * 614 * BPF_BTF_GET_NEXT_ID 615 * Description 616 * Fetch the next BPF Type Format (BTF) object currently loaded 617 * into the kernel. 618 * 619 * Looks for the BTF object with an id greater than *start_id* 620 * and updates *next_id* on success. If no other BTF objects 621 * remain with ids higher than *start_id*, returns -1 and sets 622 * *errno* to **ENOENT**. 623 * 624 * Return 625 * Returns zero on success. On error, or when no id remains, -1 626 * is returned and *errno* is set appropriately. 627 * 628 * BPF_MAP_LOOKUP_BATCH 629 * Description 630 * Iterate and fetch multiple elements in a map. 631 * 632 * Two opaque values are used to manage batch operations, 633 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 634 * to NULL to begin the batched operation. After each subsequent 635 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 636 * *out_batch* as the *in_batch* for the next operation to 637 * continue iteration from the current point. 638 * 639 * The *keys* and *values* are output parameters which must point 640 * to memory large enough to hold *count* items based on the key 641 * and value size of the map *map_fd*. The *keys* buffer must be 642 * of *key_size* * *count*. The *values* buffer must be of 643 * *value_size* * *count*. 644 * 645 * The *elem_flags* argument may be specified as one of the 646 * following: 647 * 648 * **BPF_F_LOCK** 649 * Look up the value of a spin-locked map without 650 * returning the lock. This must be specified if the 651 * elements contain a spinlock. 652 * 653 * On success, *count* elements from the map are copied into the 654 * user buffer, with the keys copied into *keys* and the values 655 * copied into the corresponding indices in *values*. 656 * 657 * If an error is returned and *errno* is not **EFAULT**, *count* 658 * is set to the number of successfully processed elements. 659 * 660 * Return 661 * Returns zero on success. On error, -1 is returned and *errno* 662 * is set appropriately. 663 * 664 * May set *errno* to **ENOSPC** to indicate that *keys* or 665 * *values* is too small to dump an entire bucket during 666 * iteration of a hash-based map type. 667 * 668 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 669 * Description 670 * Iterate and delete all elements in a map. 671 * 672 * This operation has the same behavior as 673 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 674 * 675 * * Every element that is successfully returned is also deleted 676 * from the map. This is at least *count* elements. Note that 677 * *count* is both an input and an output parameter. 678 * * Upon returning with *errno* set to **EFAULT**, up to 679 * *count* elements may be deleted without returning the keys 680 * and values of the deleted elements. 681 * 682 * Return 683 * Returns zero on success. On error, -1 is returned and *errno* 684 * is set appropriately. 685 * 686 * BPF_MAP_UPDATE_BATCH 687 * Description 688 * Update multiple elements in a map by *key*. 689 * 690 * The *keys* and *values* are input parameters which must point 691 * to memory large enough to hold *count* items based on the key 692 * and value size of the map *map_fd*. The *keys* buffer must be 693 * of *key_size* * *count*. The *values* buffer must be of 694 * *value_size* * *count*. 695 * 696 * Each element specified in *keys* is sequentially updated to the 697 * value in the corresponding index in *values*. The *in_batch* 698 * and *out_batch* parameters are ignored and should be zeroed. 699 * 700 * The *elem_flags* argument should be specified as one of the 701 * following: 702 * 703 * **BPF_ANY** 704 * Create new elements or update a existing elements. 705 * **BPF_NOEXIST** 706 * Create new elements only if they do not exist. 707 * **BPF_EXIST** 708 * Update existing elements. 709 * **BPF_F_LOCK** 710 * Update spin_lock-ed map elements. This must be 711 * specified if the map value contains a spinlock. 712 * 713 * On success, *count* elements from the map are updated. 714 * 715 * If an error is returned and *errno* is not **EFAULT**, *count* 716 * is set to the number of successfully processed elements. 717 * 718 * Return 719 * Returns zero on success. On error, -1 is returned and *errno* 720 * is set appropriately. 721 * 722 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 723 * **E2BIG**. **E2BIG** indicates that the number of elements in 724 * the map reached the *max_entries* limit specified at map 725 * creation time. 726 * 727 * May set *errno* to one of the following error codes under 728 * specific circumstances: 729 * 730 * **EEXIST** 731 * If *flags* specifies **BPF_NOEXIST** and the element 732 * with *key* already exists in the map. 733 * **ENOENT** 734 * If *flags* specifies **BPF_EXIST** and the element with 735 * *key* does not exist in the map. 736 * 737 * BPF_MAP_DELETE_BATCH 738 * Description 739 * Delete multiple elements in a map by *key*. 740 * 741 * The *keys* parameter is an input parameter which must point 742 * to memory large enough to hold *count* items based on the key 743 * size of the map *map_fd*, that is, *key_size* * *count*. 744 * 745 * Each element specified in *keys* is sequentially deleted. The 746 * *in_batch*, *out_batch*, and *values* parameters are ignored 747 * and should be zeroed. 748 * 749 * The *elem_flags* argument may be specified as one of the 750 * following: 751 * 752 * **BPF_F_LOCK** 753 * Look up the value of a spin-locked map without 754 * returning the lock. This must be specified if the 755 * elements contain a spinlock. 756 * 757 * On success, *count* elements from the map are updated. 758 * 759 * If an error is returned and *errno* is not **EFAULT**, *count* 760 * is set to the number of successfully processed elements. If 761 * *errno* is **EFAULT**, up to *count* elements may be been 762 * deleted. 763 * 764 * Return 765 * Returns zero on success. On error, -1 is returned and *errno* 766 * is set appropriately. 767 * 768 * BPF_LINK_CREATE 769 * Description 770 * Attach an eBPF program to a *target_fd* at the specified 771 * *attach_type* hook and return a file descriptor handle for 772 * managing the link. 773 * 774 * Return 775 * A new file descriptor (a nonnegative integer), or -1 if an 776 * error occurred (in which case, *errno* is set appropriately). 777 * 778 * BPF_LINK_UPDATE 779 * Description 780 * Update the eBPF program in the specified *link_fd* to 781 * *new_prog_fd*. 782 * 783 * Return 784 * Returns zero on success. On error, -1 is returned and *errno* 785 * is set appropriately. 786 * 787 * BPF_LINK_GET_FD_BY_ID 788 * Description 789 * Open a file descriptor for the eBPF Link corresponding to 790 * *link_id*. 791 * 792 * Return 793 * A new file descriptor (a nonnegative integer), or -1 if an 794 * error occurred (in which case, *errno* is set appropriately). 795 * 796 * BPF_LINK_GET_NEXT_ID 797 * Description 798 * Fetch the next eBPF link currently loaded into the kernel. 799 * 800 * Looks for the eBPF link with an id greater than *start_id* 801 * and updates *next_id* on success. If no other eBPF links 802 * remain with ids higher than *start_id*, returns -1 and sets 803 * *errno* to **ENOENT**. 804 * 805 * Return 806 * Returns zero on success. On error, or when no id remains, -1 807 * is returned and *errno* is set appropriately. 808 * 809 * BPF_ENABLE_STATS 810 * Description 811 * Enable eBPF runtime statistics gathering. 812 * 813 * Runtime statistics gathering for the eBPF runtime is disabled 814 * by default to minimize the corresponding performance overhead. 815 * This command enables statistics globally. 816 * 817 * Multiple programs may independently enable statistics. 818 * After gathering the desired statistics, eBPF runtime statistics 819 * may be disabled again by calling **close**\ (2) for the file 820 * descriptor returned by this function. Statistics will only be 821 * disabled system-wide when all outstanding file descriptors 822 * returned by prior calls for this subcommand are closed. 823 * 824 * Return 825 * A new file descriptor (a nonnegative integer), or -1 if an 826 * error occurred (in which case, *errno* is set appropriately). 827 * 828 * BPF_ITER_CREATE 829 * Description 830 * Create an iterator on top of the specified *link_fd* (as 831 * previously created using **BPF_LINK_CREATE**) and return a 832 * file descriptor that can be used to trigger the iteration. 833 * 834 * If the resulting file descriptor is pinned to the filesystem 835 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 836 * for that path will trigger the iterator to read kernel state 837 * using the eBPF program attached to *link_fd*. 838 * 839 * Return 840 * A new file descriptor (a nonnegative integer), or -1 if an 841 * error occurred (in which case, *errno* is set appropriately). 842 * 843 * BPF_LINK_DETACH 844 * Description 845 * Forcefully detach the specified *link_fd* from its 846 * corresponding attachment point. 847 * 848 * Return 849 * Returns zero on success. On error, -1 is returned and *errno* 850 * is set appropriately. 851 * 852 * BPF_PROG_BIND_MAP 853 * Description 854 * Bind a map to the lifetime of an eBPF program. 855 * 856 * The map identified by *map_fd* is bound to the program 857 * identified by *prog_fd* and only released when *prog_fd* is 858 * released. This may be used in cases where metadata should be 859 * associated with a program which otherwise does not contain any 860 * references to the map (for example, embedded in the eBPF 861 * program instructions). 862 * 863 * Return 864 * Returns zero on success. On error, -1 is returned and *errno* 865 * is set appropriately. 866 * 867 * NOTES 868 * eBPF objects (maps and programs) can be shared between processes. 869 * 870 * * After **fork**\ (2), the child inherits file descriptors 871 * referring to the same eBPF objects. 872 * * File descriptors referring to eBPF objects can be transferred over 873 * **unix**\ (7) domain sockets. 874 * * File descriptors referring to eBPF objects can be duplicated in the 875 * usual way, using **dup**\ (2) and similar calls. 876 * * File descriptors referring to eBPF objects can be pinned to the 877 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 878 * 879 * An eBPF object is deallocated only after all file descriptors referring 880 * to the object have been closed and no references remain pinned to the 881 * filesystem or attached (for example, bound to a program or device). 882 */ 883 enum bpf_cmd { 884 BPF_MAP_CREATE, 885 BPF_MAP_LOOKUP_ELEM, 886 BPF_MAP_UPDATE_ELEM, 887 BPF_MAP_DELETE_ELEM, 888 BPF_MAP_GET_NEXT_KEY, 889 BPF_PROG_LOAD, 890 BPF_OBJ_PIN, 891 BPF_OBJ_GET, 892 BPF_PROG_ATTACH, 893 BPF_PROG_DETACH, 894 BPF_PROG_TEST_RUN, 895 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 896 BPF_PROG_GET_NEXT_ID, 897 BPF_MAP_GET_NEXT_ID, 898 BPF_PROG_GET_FD_BY_ID, 899 BPF_MAP_GET_FD_BY_ID, 900 BPF_OBJ_GET_INFO_BY_FD, 901 BPF_PROG_QUERY, 902 BPF_RAW_TRACEPOINT_OPEN, 903 BPF_BTF_LOAD, 904 BPF_BTF_GET_FD_BY_ID, 905 BPF_TASK_FD_QUERY, 906 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 907 BPF_MAP_FREEZE, 908 BPF_BTF_GET_NEXT_ID, 909 BPF_MAP_LOOKUP_BATCH, 910 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 911 BPF_MAP_UPDATE_BATCH, 912 BPF_MAP_DELETE_BATCH, 913 BPF_LINK_CREATE, 914 BPF_LINK_UPDATE, 915 BPF_LINK_GET_FD_BY_ID, 916 BPF_LINK_GET_NEXT_ID, 917 BPF_ENABLE_STATS, 918 BPF_ITER_CREATE, 919 BPF_LINK_DETACH, 920 BPF_PROG_BIND_MAP, 921 }; 922 923 enum bpf_map_type { 924 BPF_MAP_TYPE_UNSPEC, 925 BPF_MAP_TYPE_HASH, 926 BPF_MAP_TYPE_ARRAY, 927 BPF_MAP_TYPE_PROG_ARRAY, 928 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 929 BPF_MAP_TYPE_PERCPU_HASH, 930 BPF_MAP_TYPE_PERCPU_ARRAY, 931 BPF_MAP_TYPE_STACK_TRACE, 932 BPF_MAP_TYPE_CGROUP_ARRAY, 933 BPF_MAP_TYPE_LRU_HASH, 934 BPF_MAP_TYPE_LRU_PERCPU_HASH, 935 BPF_MAP_TYPE_LPM_TRIE, 936 BPF_MAP_TYPE_ARRAY_OF_MAPS, 937 BPF_MAP_TYPE_HASH_OF_MAPS, 938 BPF_MAP_TYPE_DEVMAP, 939 BPF_MAP_TYPE_SOCKMAP, 940 BPF_MAP_TYPE_CPUMAP, 941 BPF_MAP_TYPE_XSKMAP, 942 BPF_MAP_TYPE_SOCKHASH, 943 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 944 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 945 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 946 * both cgroup-attached and other progs and supports all functionality 947 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 948 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 949 */ 950 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 951 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 952 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 953 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 954 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 955 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 956 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 957 * deprecated. 958 */ 959 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 960 BPF_MAP_TYPE_QUEUE, 961 BPF_MAP_TYPE_STACK, 962 BPF_MAP_TYPE_SK_STORAGE, 963 BPF_MAP_TYPE_DEVMAP_HASH, 964 BPF_MAP_TYPE_STRUCT_OPS, 965 BPF_MAP_TYPE_RINGBUF, 966 BPF_MAP_TYPE_INODE_STORAGE, 967 BPF_MAP_TYPE_TASK_STORAGE, 968 BPF_MAP_TYPE_BLOOM_FILTER, 969 BPF_MAP_TYPE_USER_RINGBUF, 970 BPF_MAP_TYPE_CGRP_STORAGE, 971 }; 972 973 /* Note that tracing related programs such as 974 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 975 * are not subject to a stable API since kernel internal data 976 * structures can change from release to release and may 977 * therefore break existing tracing BPF programs. Tracing BPF 978 * programs correspond to /a/ specific kernel which is to be 979 * analyzed, and not /a/ specific kernel /and/ all future ones. 980 */ 981 enum bpf_prog_type { 982 BPF_PROG_TYPE_UNSPEC, 983 BPF_PROG_TYPE_SOCKET_FILTER, 984 BPF_PROG_TYPE_KPROBE, 985 BPF_PROG_TYPE_SCHED_CLS, 986 BPF_PROG_TYPE_SCHED_ACT, 987 BPF_PROG_TYPE_TRACEPOINT, 988 BPF_PROG_TYPE_XDP, 989 BPF_PROG_TYPE_PERF_EVENT, 990 BPF_PROG_TYPE_CGROUP_SKB, 991 BPF_PROG_TYPE_CGROUP_SOCK, 992 BPF_PROG_TYPE_LWT_IN, 993 BPF_PROG_TYPE_LWT_OUT, 994 BPF_PROG_TYPE_LWT_XMIT, 995 BPF_PROG_TYPE_SOCK_OPS, 996 BPF_PROG_TYPE_SK_SKB, 997 BPF_PROG_TYPE_CGROUP_DEVICE, 998 BPF_PROG_TYPE_SK_MSG, 999 BPF_PROG_TYPE_RAW_TRACEPOINT, 1000 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1001 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1002 BPF_PROG_TYPE_LIRC_MODE2, 1003 BPF_PROG_TYPE_SK_REUSEPORT, 1004 BPF_PROG_TYPE_FLOW_DISSECTOR, 1005 BPF_PROG_TYPE_CGROUP_SYSCTL, 1006 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1007 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1008 BPF_PROG_TYPE_TRACING, 1009 BPF_PROG_TYPE_STRUCT_OPS, 1010 BPF_PROG_TYPE_EXT, 1011 BPF_PROG_TYPE_LSM, 1012 BPF_PROG_TYPE_SK_LOOKUP, 1013 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1014 BPF_PROG_TYPE_NETFILTER, 1015 }; 1016 1017 enum bpf_attach_type { 1018 BPF_CGROUP_INET_INGRESS, 1019 BPF_CGROUP_INET_EGRESS, 1020 BPF_CGROUP_INET_SOCK_CREATE, 1021 BPF_CGROUP_SOCK_OPS, 1022 BPF_SK_SKB_STREAM_PARSER, 1023 BPF_SK_SKB_STREAM_VERDICT, 1024 BPF_CGROUP_DEVICE, 1025 BPF_SK_MSG_VERDICT, 1026 BPF_CGROUP_INET4_BIND, 1027 BPF_CGROUP_INET6_BIND, 1028 BPF_CGROUP_INET4_CONNECT, 1029 BPF_CGROUP_INET6_CONNECT, 1030 BPF_CGROUP_INET4_POST_BIND, 1031 BPF_CGROUP_INET6_POST_BIND, 1032 BPF_CGROUP_UDP4_SENDMSG, 1033 BPF_CGROUP_UDP6_SENDMSG, 1034 BPF_LIRC_MODE2, 1035 BPF_FLOW_DISSECTOR, 1036 BPF_CGROUP_SYSCTL, 1037 BPF_CGROUP_UDP4_RECVMSG, 1038 BPF_CGROUP_UDP6_RECVMSG, 1039 BPF_CGROUP_GETSOCKOPT, 1040 BPF_CGROUP_SETSOCKOPT, 1041 BPF_TRACE_RAW_TP, 1042 BPF_TRACE_FENTRY, 1043 BPF_TRACE_FEXIT, 1044 BPF_MODIFY_RETURN, 1045 BPF_LSM_MAC, 1046 BPF_TRACE_ITER, 1047 BPF_CGROUP_INET4_GETPEERNAME, 1048 BPF_CGROUP_INET6_GETPEERNAME, 1049 BPF_CGROUP_INET4_GETSOCKNAME, 1050 BPF_CGROUP_INET6_GETSOCKNAME, 1051 BPF_XDP_DEVMAP, 1052 BPF_CGROUP_INET_SOCK_RELEASE, 1053 BPF_XDP_CPUMAP, 1054 BPF_SK_LOOKUP, 1055 BPF_XDP, 1056 BPF_SK_SKB_VERDICT, 1057 BPF_SK_REUSEPORT_SELECT, 1058 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1059 BPF_PERF_EVENT, 1060 BPF_TRACE_KPROBE_MULTI, 1061 BPF_LSM_CGROUP, 1062 BPF_STRUCT_OPS, 1063 BPF_NETFILTER, 1064 BPF_TCX_INGRESS, 1065 BPF_TCX_EGRESS, 1066 BPF_TRACE_UPROBE_MULTI, 1067 BPF_CGROUP_UNIX_CONNECT, 1068 BPF_CGROUP_UNIX_SENDMSG, 1069 BPF_CGROUP_UNIX_RECVMSG, 1070 BPF_CGROUP_UNIX_GETPEERNAME, 1071 BPF_CGROUP_UNIX_GETSOCKNAME, 1072 BPF_NETKIT_PRIMARY, 1073 BPF_NETKIT_PEER, 1074 __MAX_BPF_ATTACH_TYPE 1075 }; 1076 1077 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1078 1079 enum bpf_link_type { 1080 BPF_LINK_TYPE_UNSPEC = 0, 1081 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1082 BPF_LINK_TYPE_TRACING = 2, 1083 BPF_LINK_TYPE_CGROUP = 3, 1084 BPF_LINK_TYPE_ITER = 4, 1085 BPF_LINK_TYPE_NETNS = 5, 1086 BPF_LINK_TYPE_XDP = 6, 1087 BPF_LINK_TYPE_PERF_EVENT = 7, 1088 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1089 BPF_LINK_TYPE_STRUCT_OPS = 9, 1090 BPF_LINK_TYPE_NETFILTER = 10, 1091 BPF_LINK_TYPE_TCX = 11, 1092 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1093 BPF_LINK_TYPE_NETKIT = 13, 1094 __MAX_BPF_LINK_TYPE, 1095 }; 1096 1097 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1098 1099 enum bpf_perf_event_type { 1100 BPF_PERF_EVENT_UNSPEC = 0, 1101 BPF_PERF_EVENT_UPROBE = 1, 1102 BPF_PERF_EVENT_URETPROBE = 2, 1103 BPF_PERF_EVENT_KPROBE = 3, 1104 BPF_PERF_EVENT_KRETPROBE = 4, 1105 BPF_PERF_EVENT_TRACEPOINT = 5, 1106 BPF_PERF_EVENT_EVENT = 6, 1107 }; 1108 1109 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1110 * 1111 * NONE(default): No further bpf programs allowed in the subtree. 1112 * 1113 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1114 * the program in this cgroup yields to sub-cgroup program. 1115 * 1116 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1117 * that cgroup program gets run in addition to the program in this cgroup. 1118 * 1119 * Only one program is allowed to be attached to a cgroup with 1120 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1121 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1122 * release old program and attach the new one. Attach flags has to match. 1123 * 1124 * Multiple programs are allowed to be attached to a cgroup with 1125 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1126 * (those that were attached first, run first) 1127 * The programs of sub-cgroup are executed first, then programs of 1128 * this cgroup and then programs of parent cgroup. 1129 * When children program makes decision (like picking TCP CA or sock bind) 1130 * parent program has a chance to override it. 1131 * 1132 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1133 * programs for a cgroup. Though it's possible to replace an old program at 1134 * any position by also specifying BPF_F_REPLACE flag and position itself in 1135 * replace_bpf_fd attribute. Old program at this position will be released. 1136 * 1137 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1138 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1139 * Ex1: 1140 * cgrp1 (MULTI progs A, B) -> 1141 * cgrp2 (OVERRIDE prog C) -> 1142 * cgrp3 (MULTI prog D) -> 1143 * cgrp4 (OVERRIDE prog E) -> 1144 * cgrp5 (NONE prog F) 1145 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1146 * if prog F is detached, the execution is E,D,A,B 1147 * if prog F and D are detached, the execution is E,A,B 1148 * if prog F, E and D are detached, the execution is C,A,B 1149 * 1150 * All eligible programs are executed regardless of return code from 1151 * earlier programs. 1152 */ 1153 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1154 #define BPF_F_ALLOW_MULTI (1U << 1) 1155 /* Generic attachment flags. */ 1156 #define BPF_F_REPLACE (1U << 2) 1157 #define BPF_F_BEFORE (1U << 3) 1158 #define BPF_F_AFTER (1U << 4) 1159 #define BPF_F_ID (1U << 5) 1160 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1161 1162 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1163 * verifier will perform strict alignment checking as if the kernel 1164 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1165 * and NET_IP_ALIGN defined to 2. 1166 */ 1167 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1168 1169 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1170 * verifier will allow any alignment whatsoever. On platforms 1171 * with strict alignment requirements for loads ands stores (such 1172 * as sparc and mips) the verifier validates that all loads and 1173 * stores provably follow this requirement. This flag turns that 1174 * checking and enforcement off. 1175 * 1176 * It is mostly used for testing when we want to validate the 1177 * context and memory access aspects of the verifier, but because 1178 * of an unaligned access the alignment check would trigger before 1179 * the one we are interested in. 1180 */ 1181 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1182 1183 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1184 * Verifier does sub-register def/use analysis and identifies instructions whose 1185 * def only matters for low 32-bit, high 32-bit is never referenced later 1186 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1187 * that it is safe to ignore clearing high 32-bit for these instructions. This 1188 * saves some back-ends a lot of code-gen. However such optimization is not 1189 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1190 * hence hasn't used verifier's analysis result. But, we really want to have a 1191 * way to be able to verify the correctness of the described optimization on 1192 * x86_64 on which testsuites are frequently exercised. 1193 * 1194 * So, this flag is introduced. Once it is set, verifier will randomize high 1195 * 32-bit for those instructions who has been identified as safe to ignore them. 1196 * Then, if verifier is not doing correct analysis, such randomization will 1197 * regress tests to expose bugs. 1198 */ 1199 #define BPF_F_TEST_RND_HI32 (1U << 2) 1200 1201 /* The verifier internal test flag. Behavior is undefined */ 1202 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1203 1204 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1205 * restrict map and helper usage for such programs. Sleepable BPF programs can 1206 * only be attached to hooks where kernel execution context allows sleeping. 1207 * Such programs are allowed to use helpers that may sleep like 1208 * bpf_copy_from_user(). 1209 */ 1210 #define BPF_F_SLEEPABLE (1U << 4) 1211 1212 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1213 * fully support xdp frags. 1214 */ 1215 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1216 1217 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1218 * program becomes device-bound but can access XDP metadata. 1219 */ 1220 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1221 1222 /* The verifier internal test flag. Behavior is undefined */ 1223 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1224 1225 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1226 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1227 */ 1228 enum { 1229 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1230 }; 1231 1232 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1233 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1234 */ 1235 enum { 1236 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1237 }; 1238 1239 /* link_create.netfilter.flags used in LINK_CREATE command for 1240 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1241 */ 1242 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1243 1244 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1245 * the following extensions: 1246 * 1247 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1248 * insn[0].imm: map fd or fd_idx 1249 * insn[1].imm: 0 1250 * insn[0].off: 0 1251 * insn[1].off: 0 1252 * ldimm64 rewrite: address of map 1253 * verifier type: CONST_PTR_TO_MAP 1254 */ 1255 #define BPF_PSEUDO_MAP_FD 1 1256 #define BPF_PSEUDO_MAP_IDX 5 1257 1258 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1259 * insn[0].imm: map fd or fd_idx 1260 * insn[1].imm: offset into value 1261 * insn[0].off: 0 1262 * insn[1].off: 0 1263 * ldimm64 rewrite: address of map[0]+offset 1264 * verifier type: PTR_TO_MAP_VALUE 1265 */ 1266 #define BPF_PSEUDO_MAP_VALUE 2 1267 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1268 1269 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1270 * insn[0].imm: kernel btd id of VAR 1271 * insn[1].imm: 0 1272 * insn[0].off: 0 1273 * insn[1].off: 0 1274 * ldimm64 rewrite: address of the kernel variable 1275 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1276 * is struct/union. 1277 */ 1278 #define BPF_PSEUDO_BTF_ID 3 1279 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1280 * insn[0].imm: insn offset to the func 1281 * insn[1].imm: 0 1282 * insn[0].off: 0 1283 * insn[1].off: 0 1284 * ldimm64 rewrite: address of the function 1285 * verifier type: PTR_TO_FUNC. 1286 */ 1287 #define BPF_PSEUDO_FUNC 4 1288 1289 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1290 * offset to another bpf function 1291 */ 1292 #define BPF_PSEUDO_CALL 1 1293 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1294 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1295 */ 1296 #define BPF_PSEUDO_KFUNC_CALL 2 1297 1298 /* flags for BPF_MAP_UPDATE_ELEM command */ 1299 enum { 1300 BPF_ANY = 0, /* create new element or update existing */ 1301 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1302 BPF_EXIST = 2, /* update existing element */ 1303 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1304 }; 1305 1306 /* flags for BPF_MAP_CREATE command */ 1307 enum { 1308 BPF_F_NO_PREALLOC = (1U << 0), 1309 /* Instead of having one common LRU list in the 1310 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1311 * which can scale and perform better. 1312 * Note, the LRU nodes (including free nodes) cannot be moved 1313 * across different LRU lists. 1314 */ 1315 BPF_F_NO_COMMON_LRU = (1U << 1), 1316 /* Specify numa node during map creation */ 1317 BPF_F_NUMA_NODE = (1U << 2), 1318 1319 /* Flags for accessing BPF object from syscall side. */ 1320 BPF_F_RDONLY = (1U << 3), 1321 BPF_F_WRONLY = (1U << 4), 1322 1323 /* Flag for stack_map, store build_id+offset instead of pointer */ 1324 BPF_F_STACK_BUILD_ID = (1U << 5), 1325 1326 /* Zero-initialize hash function seed. This should only be used for testing. */ 1327 BPF_F_ZERO_SEED = (1U << 6), 1328 1329 /* Flags for accessing BPF object from program side. */ 1330 BPF_F_RDONLY_PROG = (1U << 7), 1331 BPF_F_WRONLY_PROG = (1U << 8), 1332 1333 /* Clone map from listener for newly accepted socket */ 1334 BPF_F_CLONE = (1U << 9), 1335 1336 /* Enable memory-mapping BPF map */ 1337 BPF_F_MMAPABLE = (1U << 10), 1338 1339 /* Share perf_event among processes */ 1340 BPF_F_PRESERVE_ELEMS = (1U << 11), 1341 1342 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1343 BPF_F_INNER_MAP = (1U << 12), 1344 1345 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1346 BPF_F_LINK = (1U << 13), 1347 1348 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1349 BPF_F_PATH_FD = (1U << 14), 1350 }; 1351 1352 /* Flags for BPF_PROG_QUERY. */ 1353 1354 /* Query effective (directly attached + inherited from ancestor cgroups) 1355 * programs that will be executed for events within a cgroup. 1356 * attach_flags with this flag are always returned 0. 1357 */ 1358 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1359 1360 /* Flags for BPF_PROG_TEST_RUN */ 1361 1362 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1363 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1364 /* If set, XDP frames will be transmitted after processing */ 1365 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1366 1367 /* type for BPF_ENABLE_STATS */ 1368 enum bpf_stats_type { 1369 /* enabled run_time_ns and run_cnt */ 1370 BPF_STATS_RUN_TIME = 0, 1371 }; 1372 1373 enum bpf_stack_build_id_status { 1374 /* user space need an empty entry to identify end of a trace */ 1375 BPF_STACK_BUILD_ID_EMPTY = 0, 1376 /* with valid build_id and offset */ 1377 BPF_STACK_BUILD_ID_VALID = 1, 1378 /* couldn't get build_id, fallback to ip */ 1379 BPF_STACK_BUILD_ID_IP = 2, 1380 }; 1381 1382 #define BPF_BUILD_ID_SIZE 20 1383 struct bpf_stack_build_id { 1384 __s32 status; 1385 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1386 union { 1387 __u64 offset; 1388 __u64 ip; 1389 }; 1390 }; 1391 1392 #define BPF_OBJ_NAME_LEN 16U 1393 1394 union bpf_attr { 1395 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1396 __u32 map_type; /* one of enum bpf_map_type */ 1397 __u32 key_size; /* size of key in bytes */ 1398 __u32 value_size; /* size of value in bytes */ 1399 __u32 max_entries; /* max number of entries in a map */ 1400 __u32 map_flags; /* BPF_MAP_CREATE related 1401 * flags defined above. 1402 */ 1403 __u32 inner_map_fd; /* fd pointing to the inner map */ 1404 __u32 numa_node; /* numa node (effective only if 1405 * BPF_F_NUMA_NODE is set). 1406 */ 1407 char map_name[BPF_OBJ_NAME_LEN]; 1408 __u32 map_ifindex; /* ifindex of netdev to create on */ 1409 __u32 btf_fd; /* fd pointing to a BTF type data */ 1410 __u32 btf_key_type_id; /* BTF type_id of the key */ 1411 __u32 btf_value_type_id; /* BTF type_id of the value */ 1412 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1413 * struct stored as the 1414 * map value 1415 */ 1416 /* Any per-map-type extra fields 1417 * 1418 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1419 * number of hash functions (if 0, the bloom filter will default 1420 * to using 5 hash functions). 1421 */ 1422 __u64 map_extra; 1423 }; 1424 1425 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1426 __u32 map_fd; 1427 __aligned_u64 key; 1428 union { 1429 __aligned_u64 value; 1430 __aligned_u64 next_key; 1431 }; 1432 __u64 flags; 1433 }; 1434 1435 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1436 __aligned_u64 in_batch; /* start batch, 1437 * NULL to start from beginning 1438 */ 1439 __aligned_u64 out_batch; /* output: next start batch */ 1440 __aligned_u64 keys; 1441 __aligned_u64 values; 1442 __u32 count; /* input/output: 1443 * input: # of key/value 1444 * elements 1445 * output: # of filled elements 1446 */ 1447 __u32 map_fd; 1448 __u64 elem_flags; 1449 __u64 flags; 1450 } batch; 1451 1452 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1453 __u32 prog_type; /* one of enum bpf_prog_type */ 1454 __u32 insn_cnt; 1455 __aligned_u64 insns; 1456 __aligned_u64 license; 1457 __u32 log_level; /* verbosity level of verifier */ 1458 __u32 log_size; /* size of user buffer */ 1459 __aligned_u64 log_buf; /* user supplied buffer */ 1460 __u32 kern_version; /* not used */ 1461 __u32 prog_flags; 1462 char prog_name[BPF_OBJ_NAME_LEN]; 1463 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1464 /* For some prog types expected attach type must be known at 1465 * load time to verify attach type specific parts of prog 1466 * (context accesses, allowed helpers, etc). 1467 */ 1468 __u32 expected_attach_type; 1469 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1470 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1471 __aligned_u64 func_info; /* func info */ 1472 __u32 func_info_cnt; /* number of bpf_func_info records */ 1473 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1474 __aligned_u64 line_info; /* line info */ 1475 __u32 line_info_cnt; /* number of bpf_line_info records */ 1476 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1477 union { 1478 /* valid prog_fd to attach to bpf prog */ 1479 __u32 attach_prog_fd; 1480 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1481 __u32 attach_btf_obj_fd; 1482 }; 1483 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1484 __aligned_u64 fd_array; /* array of FDs */ 1485 __aligned_u64 core_relos; 1486 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1487 /* output: actual total log contents size (including termintaing zero). 1488 * It could be both larger than original log_size (if log was 1489 * truncated), or smaller (if log buffer wasn't filled completely). 1490 */ 1491 __u32 log_true_size; 1492 }; 1493 1494 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1495 __aligned_u64 pathname; 1496 __u32 bpf_fd; 1497 __u32 file_flags; 1498 /* Same as dirfd in openat() syscall; see openat(2) 1499 * manpage for details of path FD and pathname semantics; 1500 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1501 * file_flags field, otherwise it should be set to zero; 1502 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1503 */ 1504 __s32 path_fd; 1505 }; 1506 1507 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1508 union { 1509 __u32 target_fd; /* target object to attach to or ... */ 1510 __u32 target_ifindex; /* target ifindex */ 1511 }; 1512 __u32 attach_bpf_fd; 1513 __u32 attach_type; 1514 __u32 attach_flags; 1515 __u32 replace_bpf_fd; 1516 union { 1517 __u32 relative_fd; 1518 __u32 relative_id; 1519 }; 1520 __u64 expected_revision; 1521 }; 1522 1523 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1524 __u32 prog_fd; 1525 __u32 retval; 1526 __u32 data_size_in; /* input: len of data_in */ 1527 __u32 data_size_out; /* input/output: len of data_out 1528 * returns ENOSPC if data_out 1529 * is too small. 1530 */ 1531 __aligned_u64 data_in; 1532 __aligned_u64 data_out; 1533 __u32 repeat; 1534 __u32 duration; 1535 __u32 ctx_size_in; /* input: len of ctx_in */ 1536 __u32 ctx_size_out; /* input/output: len of ctx_out 1537 * returns ENOSPC if ctx_out 1538 * is too small. 1539 */ 1540 __aligned_u64 ctx_in; 1541 __aligned_u64 ctx_out; 1542 __u32 flags; 1543 __u32 cpu; 1544 __u32 batch_size; 1545 } test; 1546 1547 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1548 union { 1549 __u32 start_id; 1550 __u32 prog_id; 1551 __u32 map_id; 1552 __u32 btf_id; 1553 __u32 link_id; 1554 }; 1555 __u32 next_id; 1556 __u32 open_flags; 1557 }; 1558 1559 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1560 __u32 bpf_fd; 1561 __u32 info_len; 1562 __aligned_u64 info; 1563 } info; 1564 1565 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1566 union { 1567 __u32 target_fd; /* target object to query or ... */ 1568 __u32 target_ifindex; /* target ifindex */ 1569 }; 1570 __u32 attach_type; 1571 __u32 query_flags; 1572 __u32 attach_flags; 1573 __aligned_u64 prog_ids; 1574 union { 1575 __u32 prog_cnt; 1576 __u32 count; 1577 }; 1578 __u32 :32; 1579 /* output: per-program attach_flags. 1580 * not allowed to be set during effective query. 1581 */ 1582 __aligned_u64 prog_attach_flags; 1583 __aligned_u64 link_ids; 1584 __aligned_u64 link_attach_flags; 1585 __u64 revision; 1586 } query; 1587 1588 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1589 __u64 name; 1590 __u32 prog_fd; 1591 } raw_tracepoint; 1592 1593 struct { /* anonymous struct for BPF_BTF_LOAD */ 1594 __aligned_u64 btf; 1595 __aligned_u64 btf_log_buf; 1596 __u32 btf_size; 1597 __u32 btf_log_size; 1598 __u32 btf_log_level; 1599 /* output: actual total log contents size (including termintaing zero). 1600 * It could be both larger than original log_size (if log was 1601 * truncated), or smaller (if log buffer wasn't filled completely). 1602 */ 1603 __u32 btf_log_true_size; 1604 }; 1605 1606 struct { 1607 __u32 pid; /* input: pid */ 1608 __u32 fd; /* input: fd */ 1609 __u32 flags; /* input: flags */ 1610 __u32 buf_len; /* input/output: buf len */ 1611 __aligned_u64 buf; /* input/output: 1612 * tp_name for tracepoint 1613 * symbol for kprobe 1614 * filename for uprobe 1615 */ 1616 __u32 prog_id; /* output: prod_id */ 1617 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1618 __u64 probe_offset; /* output: probe_offset */ 1619 __u64 probe_addr; /* output: probe_addr */ 1620 } task_fd_query; 1621 1622 struct { /* struct used by BPF_LINK_CREATE command */ 1623 union { 1624 __u32 prog_fd; /* eBPF program to attach */ 1625 __u32 map_fd; /* struct_ops to attach */ 1626 }; 1627 union { 1628 __u32 target_fd; /* target object to attach to or ... */ 1629 __u32 target_ifindex; /* target ifindex */ 1630 }; 1631 __u32 attach_type; /* attach type */ 1632 __u32 flags; /* extra flags */ 1633 union { 1634 __u32 target_btf_id; /* btf_id of target to attach to */ 1635 struct { 1636 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1637 __u32 iter_info_len; /* iter_info length */ 1638 }; 1639 struct { 1640 /* black box user-provided value passed through 1641 * to BPF program at the execution time and 1642 * accessible through bpf_get_attach_cookie() BPF helper 1643 */ 1644 __u64 bpf_cookie; 1645 } perf_event; 1646 struct { 1647 __u32 flags; 1648 __u32 cnt; 1649 __aligned_u64 syms; 1650 __aligned_u64 addrs; 1651 __aligned_u64 cookies; 1652 } kprobe_multi; 1653 struct { 1654 /* this is overlaid with the target_btf_id above. */ 1655 __u32 target_btf_id; 1656 /* black box user-provided value passed through 1657 * to BPF program at the execution time and 1658 * accessible through bpf_get_attach_cookie() BPF helper 1659 */ 1660 __u64 cookie; 1661 } tracing; 1662 struct { 1663 __u32 pf; 1664 __u32 hooknum; 1665 __s32 priority; 1666 __u32 flags; 1667 } netfilter; 1668 struct { 1669 union { 1670 __u32 relative_fd; 1671 __u32 relative_id; 1672 }; 1673 __u64 expected_revision; 1674 } tcx; 1675 struct { 1676 __aligned_u64 path; 1677 __aligned_u64 offsets; 1678 __aligned_u64 ref_ctr_offsets; 1679 __aligned_u64 cookies; 1680 __u32 cnt; 1681 __u32 flags; 1682 __u32 pid; 1683 } uprobe_multi; 1684 struct { 1685 union { 1686 __u32 relative_fd; 1687 __u32 relative_id; 1688 }; 1689 __u64 expected_revision; 1690 } netkit; 1691 }; 1692 } link_create; 1693 1694 struct { /* struct used by BPF_LINK_UPDATE command */ 1695 __u32 link_fd; /* link fd */ 1696 union { 1697 /* new program fd to update link with */ 1698 __u32 new_prog_fd; 1699 /* new struct_ops map fd to update link with */ 1700 __u32 new_map_fd; 1701 }; 1702 __u32 flags; /* extra flags */ 1703 union { 1704 /* expected link's program fd; is specified only if 1705 * BPF_F_REPLACE flag is set in flags. 1706 */ 1707 __u32 old_prog_fd; 1708 /* expected link's map fd; is specified only 1709 * if BPF_F_REPLACE flag is set. 1710 */ 1711 __u32 old_map_fd; 1712 }; 1713 } link_update; 1714 1715 struct { 1716 __u32 link_fd; 1717 } link_detach; 1718 1719 struct { /* struct used by BPF_ENABLE_STATS command */ 1720 __u32 type; 1721 } enable_stats; 1722 1723 struct { /* struct used by BPF_ITER_CREATE command */ 1724 __u32 link_fd; 1725 __u32 flags; 1726 } iter_create; 1727 1728 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1729 __u32 prog_fd; 1730 __u32 map_fd; 1731 __u32 flags; /* extra flags */ 1732 } prog_bind_map; 1733 1734 } __attribute__((aligned(8))); 1735 1736 /* The description below is an attempt at providing documentation to eBPF 1737 * developers about the multiple available eBPF helper functions. It can be 1738 * parsed and used to produce a manual page. The workflow is the following, 1739 * and requires the rst2man utility: 1740 * 1741 * $ ./scripts/bpf_doc.py \ 1742 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1743 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1744 * $ man /tmp/bpf-helpers.7 1745 * 1746 * Note that in order to produce this external documentation, some RST 1747 * formatting is used in the descriptions to get "bold" and "italics" in 1748 * manual pages. Also note that the few trailing white spaces are 1749 * intentional, removing them would break paragraphs for rst2man. 1750 * 1751 * Start of BPF helper function descriptions: 1752 * 1753 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1754 * Description 1755 * Perform a lookup in *map* for an entry associated to *key*. 1756 * Return 1757 * Map value associated to *key*, or **NULL** if no entry was 1758 * found. 1759 * 1760 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1761 * Description 1762 * Add or update the value of the entry associated to *key* in 1763 * *map* with *value*. *flags* is one of: 1764 * 1765 * **BPF_NOEXIST** 1766 * The entry for *key* must not exist in the map. 1767 * **BPF_EXIST** 1768 * The entry for *key* must already exist in the map. 1769 * **BPF_ANY** 1770 * No condition on the existence of the entry for *key*. 1771 * 1772 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1773 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1774 * elements always exist), the helper would return an error. 1775 * Return 1776 * 0 on success, or a negative error in case of failure. 1777 * 1778 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1779 * Description 1780 * Delete entry with *key* from *map*. 1781 * Return 1782 * 0 on success, or a negative error in case of failure. 1783 * 1784 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1785 * Description 1786 * For tracing programs, safely attempt to read *size* bytes from 1787 * kernel space address *unsafe_ptr* and store the data in *dst*. 1788 * 1789 * Generally, use **bpf_probe_read_user**\ () or 1790 * **bpf_probe_read_kernel**\ () instead. 1791 * Return 1792 * 0 on success, or a negative error in case of failure. 1793 * 1794 * u64 bpf_ktime_get_ns(void) 1795 * Description 1796 * Return the time elapsed since system boot, in nanoseconds. 1797 * Does not include time the system was suspended. 1798 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1799 * Return 1800 * Current *ktime*. 1801 * 1802 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1803 * Description 1804 * This helper is a "printk()-like" facility for debugging. It 1805 * prints a message defined by format *fmt* (of size *fmt_size*) 1806 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1807 * available. It can take up to three additional **u64** 1808 * arguments (as an eBPF helpers, the total number of arguments is 1809 * limited to five). 1810 * 1811 * Each time the helper is called, it appends a line to the trace. 1812 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1813 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1814 * The format of the trace is customizable, and the exact output 1815 * one will get depends on the options set in 1816 * *\/sys/kernel/tracing/trace_options* (see also the 1817 * *README* file under the same directory). However, it usually 1818 * defaults to something like: 1819 * 1820 * :: 1821 * 1822 * telnet-470 [001] .N.. 419421.045894: 0x00000001:
1823 * 1824 * In the above: 1825 * 1826 * * ``telnet`` is the name of the current task. 1827 * * ``470`` is the PID of the current task. 1828 * * ``001`` is the CPU number on which the task is 1829 * running. 1830 * * In ``.N..``, each character refers to a set of 1831 * options (whether irqs are enabled, scheduling 1832 * options, whether hard/softirqs are running, level of 1833 * preempt_disabled respectively). **N** means that 1834 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1835 * are set. 1836 * * ``419421.045894`` is a timestamp. 1837 * * ``0x00000001`` is a fake value used by BPF for the 1838 * instruction pointer register. 1839 * * ``
`` is the message formatted with 1840 * *fmt*. 1841 * 1842 * The conversion specifiers supported by *fmt* are similar, but 1843 * more limited than for printk(). They are **%d**, **%i**, 1844 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1845 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1846 * of field, padding with zeroes, etc.) is available, and the 1847 * helper will return **-EINVAL** (but print nothing) if it 1848 * encounters an unknown specifier. 1849 * 1850 * Also, note that **bpf_trace_printk**\ () is slow, and should 1851 * only be used for debugging purposes. For this reason, a notice 1852 * block (spanning several lines) is printed to kernel logs and 1853 * states that the helper should not be used "for production use" 1854 * the first time this helper is used (or more precisely, when 1855 * **trace_printk**\ () buffers are allocated). For passing values 1856 * to user space, perf events should be preferred. 1857 * Return 1858 * The number of bytes written to the buffer, or a negative error 1859 * in case of failure. 1860 * 1861 * u32 bpf_get_prandom_u32(void) 1862 * Description 1863 * Get a pseudo-random number. 1864 * 1865 * From a security point of view, this helper uses its own 1866 * pseudo-random internal state, and cannot be used to infer the 1867 * seed of other random functions in the kernel. However, it is 1868 * essential to note that the generator used by the helper is not 1869 * cryptographically secure. 1870 * Return 1871 * A random 32-bit unsigned value. 1872 * 1873 * u32 bpf_get_smp_processor_id(void) 1874 * Description 1875 * Get the SMP (symmetric multiprocessing) processor id. Note that 1876 * all programs run with migration disabled, which means that the 1877 * SMP processor id is stable during all the execution of the 1878 * program. 1879 * Return 1880 * The SMP id of the processor running the program. 1881 * 1882 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1883 * Description 1884 * Store *len* bytes from address *from* into the packet 1885 * associated to *skb*, at *offset*. *flags* are a combination of 1886 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1887 * checksum for the packet after storing the bytes) and 1888 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1889 * **->swhash** and *skb*\ **->l4hash** to 0). 1890 * 1891 * A call to this helper is susceptible to change the underlying 1892 * packet buffer. Therefore, at load time, all checks on pointers 1893 * previously done by the verifier are invalidated and must be 1894 * performed again, if the helper is used in combination with 1895 * direct packet access. 1896 * Return 1897 * 0 on success, or a negative error in case of failure. 1898 * 1899 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1900 * Description 1901 * Recompute the layer 3 (e.g. IP) checksum for the packet 1902 * associated to *skb*. Computation is incremental, so the helper 1903 * must know the former value of the header field that was 1904 * modified (*from*), the new value of this field (*to*), and the 1905 * number of bytes (2 or 4) for this field, stored in *size*. 1906 * Alternatively, it is possible to store the difference between 1907 * the previous and the new values of the header field in *to*, by 1908 * setting *from* and *size* to 0. For both methods, *offset* 1909 * indicates the location of the IP checksum within the packet. 1910 * 1911 * This helper works in combination with **bpf_csum_diff**\ (), 1912 * which does not update the checksum in-place, but offers more 1913 * flexibility and can handle sizes larger than 2 or 4 for the 1914 * checksum to update. 1915 * 1916 * A call to this helper is susceptible to change the underlying 1917 * packet buffer. Therefore, at load time, all checks on pointers 1918 * previously done by the verifier are invalidated and must be 1919 * performed again, if the helper is used in combination with 1920 * direct packet access. 1921 * Return 1922 * 0 on success, or a negative error in case of failure. 1923 * 1924 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1925 * Description 1926 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1927 * packet associated to *skb*. Computation is incremental, so the 1928 * helper must know the former value of the header field that was 1929 * modified (*from*), the new value of this field (*to*), and the 1930 * number of bytes (2 or 4) for this field, stored on the lowest 1931 * four bits of *flags*. Alternatively, it is possible to store 1932 * the difference between the previous and the new values of the 1933 * header field in *to*, by setting *from* and the four lowest 1934 * bits of *flags* to 0. For both methods, *offset* indicates the 1935 * location of the IP checksum within the packet. In addition to 1936 * the size of the field, *flags* can be added (bitwise OR) actual 1937 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1938 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1939 * for updates resulting in a null checksum the value is set to 1940 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1941 * the checksum is to be computed against a pseudo-header. 1942 * 1943 * This helper works in combination with **bpf_csum_diff**\ (), 1944 * which does not update the checksum in-place, but offers more 1945 * flexibility and can handle sizes larger than 2 or 4 for the 1946 * checksum to update. 1947 * 1948 * A call to this helper is susceptible to change the underlying 1949 * packet buffer. Therefore, at load time, all checks on pointers 1950 * previously done by the verifier are invalidated and must be 1951 * performed again, if the helper is used in combination with 1952 * direct packet access. 1953 * Return 1954 * 0 on success, or a negative error in case of failure. 1955 * 1956 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1957 * Description 1958 * This special helper is used to trigger a "tail call", or in 1959 * other words, to jump into another eBPF program. The same stack 1960 * frame is used (but values on stack and in registers for the 1961 * caller are not accessible to the callee). This mechanism allows 1962 * for program chaining, either for raising the maximum number of 1963 * available eBPF instructions, or to execute given programs in 1964 * conditional blocks. For security reasons, there is an upper 1965 * limit to the number of successive tail calls that can be 1966 * performed. 1967 * 1968 * Upon call of this helper, the program attempts to jump into a 1969 * program referenced at index *index* in *prog_array_map*, a 1970 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1971 * *ctx*, a pointer to the context. 1972 * 1973 * If the call succeeds, the kernel immediately runs the first 1974 * instruction of the new program. This is not a function call, 1975 * and it never returns to the previous program. If the call 1976 * fails, then the helper has no effect, and the caller continues 1977 * to run its subsequent instructions. A call can fail if the 1978 * destination program for the jump does not exist (i.e. *index* 1979 * is superior to the number of entries in *prog_array_map*), or 1980 * if the maximum number of tail calls has been reached for this 1981 * chain of programs. This limit is defined in the kernel by the 1982 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1983 * which is currently set to 33. 1984 * Return 1985 * 0 on success, or a negative error in case of failure. 1986 * 1987 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1988 * Description 1989 * Clone and redirect the packet associated to *skb* to another 1990 * net device of index *ifindex*. Both ingress and egress 1991 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1992 * value in *flags* is used to make the distinction (ingress path 1993 * is selected if the flag is present, egress path otherwise). 1994 * This is the only flag supported for now. 1995 * 1996 * In comparison with **bpf_redirect**\ () helper, 1997 * **bpf_clone_redirect**\ () has the associated cost of 1998 * duplicating the packet buffer, but this can be executed out of 1999 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2000 * efficient, but it is handled through an action code where the 2001 * redirection happens only after the eBPF program has returned. 2002 * 2003 * A call to this helper is susceptible to change the underlying 2004 * packet buffer. Therefore, at load time, all checks on pointers 2005 * previously done by the verifier are invalidated and must be 2006 * performed again, if the helper is used in combination with 2007 * direct packet access. 2008 * Return 2009 * 0 on success, or a negative error in case of failure. Positive 2010 * error indicates a potential drop or congestion in the target 2011 * device. The particular positive error codes are not defined. 2012 * 2013 * u64 bpf_get_current_pid_tgid(void) 2014 * Description 2015 * Get the current pid and tgid. 2016 * Return 2017 * A 64-bit integer containing the current tgid and pid, and 2018 * created as such: 2019 * *current_task*\ **->tgid << 32 \|** 2020 * *current_task*\ **->pid**. 2021 * 2022 * u64 bpf_get_current_uid_gid(void) 2023 * Description 2024 * Get the current uid and gid. 2025 * Return 2026 * A 64-bit integer containing the current GID and UID, and 2027 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2028 * 2029 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2030 * Description 2031 * Copy the **comm** attribute of the current task into *buf* of 2032 * *size_of_buf*. The **comm** attribute contains the name of 2033 * the executable (excluding the path) for the current task. The 2034 * *size_of_buf* must be strictly positive. On success, the 2035 * helper makes sure that the *buf* is NUL-terminated. On failure, 2036 * it is filled with zeroes. 2037 * Return 2038 * 0 on success, or a negative error in case of failure. 2039 * 2040 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2041 * Description 2042 * Retrieve the classid for the current task, i.e. for the net_cls 2043 * cgroup to which *skb* belongs. 2044 * 2045 * This helper can be used on TC egress path, but not on ingress. 2046 * 2047 * The net_cls cgroup provides an interface to tag network packets 2048 * based on a user-provided identifier for all traffic coming from 2049 * the tasks belonging to the related cgroup. See also the related 2050 * kernel documentation, available from the Linux sources in file 2051 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2052 * 2053 * The Linux kernel has two versions for cgroups: there are 2054 * cgroups v1 and cgroups v2. Both are available to users, who can 2055 * use a mixture of them, but note that the net_cls cgroup is for 2056 * cgroup v1 only. This makes it incompatible with BPF programs 2057 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2058 * only hold data for one version of cgroups at a time). 2059 * 2060 * This helper is only available is the kernel was compiled with 2061 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2062 * "**y**" or to "**m**". 2063 * Return 2064 * The classid, or 0 for the default unconfigured classid. 2065 * 2066 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2067 * Description 2068 * Push a *vlan_tci* (VLAN tag control information) of protocol 2069 * *vlan_proto* to the packet associated to *skb*, then update 2070 * the checksum. Note that if *vlan_proto* is different from 2071 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2072 * be **ETH_P_8021Q**. 2073 * 2074 * A call to this helper is susceptible to change the underlying 2075 * packet buffer. Therefore, at load time, all checks on pointers 2076 * previously done by the verifier are invalidated and must be 2077 * performed again, if the helper is used in combination with 2078 * direct packet access. 2079 * Return 2080 * 0 on success, or a negative error in case of failure. 2081 * 2082 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2083 * Description 2084 * Pop a VLAN header from the packet associated to *skb*. 2085 * 2086 * A call to this helper is susceptible to change the underlying 2087 * packet buffer. Therefore, at load time, all checks on pointers 2088 * previously done by the verifier are invalidated and must be 2089 * performed again, if the helper is used in combination with 2090 * direct packet access. 2091 * Return 2092 * 0 on success, or a negative error in case of failure. 2093 * 2094 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2095 * Description 2096 * Get tunnel metadata. This helper takes a pointer *key* to an 2097 * empty **struct bpf_tunnel_key** of **size**, that will be 2098 * filled with tunnel metadata for the packet associated to *skb*. 2099 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2100 * indicates that the tunnel is based on IPv6 protocol instead of 2101 * IPv4. 2102 * 2103 * The **struct bpf_tunnel_key** is an object that generalizes the 2104 * principal parameters used by various tunneling protocols into a 2105 * single struct. This way, it can be used to easily make a 2106 * decision based on the contents of the encapsulation header, 2107 * "summarized" in this struct. In particular, it holds the IP 2108 * address of the remote end (IPv4 or IPv6, depending on the case) 2109 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2110 * this struct exposes the *key*\ **->tunnel_id**, which is 2111 * generally mapped to a VNI (Virtual Network Identifier), making 2112 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2113 * () helper. 2114 * 2115 * Let's imagine that the following code is part of a program 2116 * attached to the TC ingress interface, on one end of a GRE 2117 * tunnel, and is supposed to filter out all messages coming from 2118 * remote ends with IPv4 address other than 10.0.0.1: 2119 * 2120 * :: 2121 * 2122 * int ret; 2123 * struct bpf_tunnel_key key = {}; 2124 * 2125 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2126 * if (ret < 0) 2127 * return TC_ACT_SHOT; // drop packet 2128 * 2129 * if (key.remote_ipv4 != 0x0a000001) 2130 * return TC_ACT_SHOT; // drop packet 2131 * 2132 * return TC_ACT_OK; // accept packet 2133 * 2134 * This interface can also be used with all encapsulation devices 2135 * that can operate in "collect metadata" mode: instead of having 2136 * one network device per specific configuration, the "collect 2137 * metadata" mode only requires a single device where the 2138 * configuration can be extracted from this helper. 2139 * 2140 * This can be used together with various tunnels such as VXLan, 2141 * Geneve, GRE or IP in IP (IPIP). 2142 * Return 2143 * 0 on success, or a negative error in case of failure. 2144 * 2145 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2146 * Description 2147 * Populate tunnel metadata for packet associated to *skb.* The 2148 * tunnel metadata is set to the contents of *key*, of *size*. The 2149 * *flags* can be set to a combination of the following values: 2150 * 2151 * **BPF_F_TUNINFO_IPV6** 2152 * Indicate that the tunnel is based on IPv6 protocol 2153 * instead of IPv4. 2154 * **BPF_F_ZERO_CSUM_TX** 2155 * For IPv4 packets, add a flag to tunnel metadata 2156 * indicating that checksum computation should be skipped 2157 * and checksum set to zeroes. 2158 * **BPF_F_DONT_FRAGMENT** 2159 * Add a flag to tunnel metadata indicating that the 2160 * packet should not be fragmented. 2161 * **BPF_F_SEQ_NUMBER** 2162 * Add a flag to tunnel metadata indicating that a 2163 * sequence number should be added to tunnel header before 2164 * sending the packet. This flag was added for GRE 2165 * encapsulation, but might be used with other protocols 2166 * as well in the future. 2167 * **BPF_F_NO_TUNNEL_KEY** 2168 * Add a flag to tunnel metadata indicating that no tunnel 2169 * key should be set in the resulting tunnel header. 2170 * 2171 * Here is a typical usage on the transmit path: 2172 * 2173 * :: 2174 * 2175 * struct bpf_tunnel_key key; 2176 * populate key ... 2177 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2178 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2179 * 2180 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2181 * helper for additional information. 2182 * Return 2183 * 0 on success, or a negative error in case of failure. 2184 * 2185 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2186 * Description 2187 * Read the value of a perf event counter. This helper relies on a 2188 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2189 * the perf event counter is selected when *map* is updated with 2190 * perf event file descriptors. The *map* is an array whose size 2191 * is the number of available CPUs, and each cell contains a value 2192 * relative to one CPU. The value to retrieve is indicated by 2193 * *flags*, that contains the index of the CPU to look up, masked 2194 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2195 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2196 * current CPU should be retrieved. 2197 * 2198 * Note that before Linux 4.13, only hardware perf event can be 2199 * retrieved. 2200 * 2201 * Also, be aware that the newer helper 2202 * **bpf_perf_event_read_value**\ () is recommended over 2203 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2204 * quirks where error and counter value are used as a return code 2205 * (which is wrong to do since ranges may overlap). This issue is 2206 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2207 * time provides more features over the **bpf_perf_event_read**\ 2208 * () interface. Please refer to the description of 2209 * **bpf_perf_event_read_value**\ () for details. 2210 * Return 2211 * The value of the perf event counter read from the map, or a 2212 * negative error code in case of failure. 2213 * 2214 * long bpf_redirect(u32 ifindex, u64 flags) 2215 * Description 2216 * Redirect the packet to another net device of index *ifindex*. 2217 * This helper is somewhat similar to **bpf_clone_redirect**\ 2218 * (), except that the packet is not cloned, which provides 2219 * increased performance. 2220 * 2221 * Except for XDP, both ingress and egress interfaces can be used 2222 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2223 * to make the distinction (ingress path is selected if the flag 2224 * is present, egress path otherwise). Currently, XDP only 2225 * supports redirection to the egress interface, and accepts no 2226 * flag at all. 2227 * 2228 * The same effect can also be attained with the more generic 2229 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2230 * redirect target instead of providing it directly to the helper. 2231 * Return 2232 * For XDP, the helper returns **XDP_REDIRECT** on success or 2233 * **XDP_ABORTED** on error. For other program types, the values 2234 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2235 * error. 2236 * 2237 * u32 bpf_get_route_realm(struct sk_buff *skb) 2238 * Description 2239 * Retrieve the realm or the route, that is to say the 2240 * **tclassid** field of the destination for the *skb*. The 2241 * identifier retrieved is a user-provided tag, similar to the 2242 * one used with the net_cls cgroup (see description for 2243 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2244 * held by a route (a destination entry), not by a task. 2245 * 2246 * Retrieving this identifier works with the clsact TC egress hook 2247 * (see also **tc-bpf(8)**), or alternatively on conventional 2248 * classful egress qdiscs, but not on TC ingress path. In case of 2249 * clsact TC egress hook, this has the advantage that, internally, 2250 * the destination entry has not been dropped yet in the transmit 2251 * path. Therefore, the destination entry does not need to be 2252 * artificially held via **netif_keep_dst**\ () for a classful 2253 * qdisc until the *skb* is freed. 2254 * 2255 * This helper is available only if the kernel was compiled with 2256 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2257 * Return 2258 * The realm of the route for the packet associated to *skb*, or 0 2259 * if none was found. 2260 * 2261 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2262 * Description 2263 * Write raw *data* blob into a special BPF perf event held by 2264 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2265 * event must have the following attributes: **PERF_SAMPLE_RAW** 2266 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2267 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2268 * 2269 * The *flags* are used to indicate the index in *map* for which 2270 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2271 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2272 * to indicate that the index of the current CPU core should be 2273 * used. 2274 * 2275 * The value to write, of *size*, is passed through eBPF stack and 2276 * pointed by *data*. 2277 * 2278 * The context of the program *ctx* needs also be passed to the 2279 * helper. 2280 * 2281 * On user space, a program willing to read the values needs to 2282 * call **perf_event_open**\ () on the perf event (either for 2283 * one or for all CPUs) and to store the file descriptor into the 2284 * *map*. This must be done before the eBPF program can send data 2285 * into it. An example is available in file 2286 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2287 * tree (the eBPF program counterpart is in 2288 * *samples/bpf/trace_output_kern.c*). 2289 * 2290 * **bpf_perf_event_output**\ () achieves better performance 2291 * than **bpf_trace_printk**\ () for sharing data with user 2292 * space, and is much better suitable for streaming data from eBPF 2293 * programs. 2294 * 2295 * Note that this helper is not restricted to tracing use cases 2296 * and can be used with programs attached to TC or XDP as well, 2297 * where it allows for passing data to user space listeners. Data 2298 * can be: 2299 * 2300 * * Only custom structs, 2301 * * Only the packet payload, or 2302 * * A combination of both. 2303 * Return 2304 * 0 on success, or a negative error in case of failure. 2305 * 2306 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2307 * Description 2308 * This helper was provided as an easy way to load data from a 2309 * packet. It can be used to load *len* bytes from *offset* from 2310 * the packet associated to *skb*, into the buffer pointed by 2311 * *to*. 2312 * 2313 * Since Linux 4.7, usage of this helper has mostly been replaced 2314 * by "direct packet access", enabling packet data to be 2315 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2316 * pointing respectively to the first byte of packet data and to 2317 * the byte after the last byte of packet data. However, it 2318 * remains useful if one wishes to read large quantities of data 2319 * at once from a packet into the eBPF stack. 2320 * Return 2321 * 0 on success, or a negative error in case of failure. 2322 * 2323 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2324 * Description 2325 * Walk a user or a kernel stack and return its id. To achieve 2326 * this, the helper needs *ctx*, which is a pointer to the context 2327 * on which the tracing program is executed, and a pointer to a 2328 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2329 * 2330 * The last argument, *flags*, holds the number of stack frames to 2331 * skip (from 0 to 255), masked with 2332 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2333 * a combination of the following flags: 2334 * 2335 * **BPF_F_USER_STACK** 2336 * Collect a user space stack instead of a kernel stack. 2337 * **BPF_F_FAST_STACK_CMP** 2338 * Compare stacks by hash only. 2339 * **BPF_F_REUSE_STACKID** 2340 * If two different stacks hash into the same *stackid*, 2341 * discard the old one. 2342 * 2343 * The stack id retrieved is a 32 bit long integer handle which 2344 * can be further combined with other data (including other stack 2345 * ids) and used as a key into maps. This can be useful for 2346 * generating a variety of graphs (such as flame graphs or off-cpu 2347 * graphs). 2348 * 2349 * For walking a stack, this helper is an improvement over 2350 * **bpf_probe_read**\ (), which can be used with unrolled loops 2351 * but is not efficient and consumes a lot of eBPF instructions. 2352 * Instead, **bpf_get_stackid**\ () can collect up to 2353 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2354 * this limit can be controlled with the **sysctl** program, and 2355 * that it should be manually increased in order to profile long 2356 * user stacks (such as stacks for Java programs). To do so, use: 2357 * 2358 * :: 2359 * 2360 * # sysctl kernel.perf_event_max_stack=
2361 * Return 2362 * The positive or null stack id on success, or a negative error 2363 * in case of failure. 2364 * 2365 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2366 * Description 2367 * Compute a checksum difference, from the raw buffer pointed by 2368 * *from*, of length *from_size* (that must be a multiple of 4), 2369 * towards the raw buffer pointed by *to*, of size *to_size* 2370 * (same remark). An optional *seed* can be added to the value 2371 * (this can be cascaded, the seed may come from a previous call 2372 * to the helper). 2373 * 2374 * This is flexible enough to be used in several ways: 2375 * 2376 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2377 * checksum, it can be used when pushing new data. 2378 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2379 * checksum, it can be used when removing data from a packet. 2380 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2381 * can be used to compute a diff. Note that *from_size* and 2382 * *to_size* do not need to be equal. 2383 * 2384 * This helper can be used in combination with 2385 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2386 * which one can feed in the difference computed with 2387 * **bpf_csum_diff**\ (). 2388 * Return 2389 * The checksum result, or a negative error code in case of 2390 * failure. 2391 * 2392 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2393 * Description 2394 * Retrieve tunnel options metadata for the packet associated to 2395 * *skb*, and store the raw tunnel option data to the buffer *opt* 2396 * of *size*. 2397 * 2398 * This helper can be used with encapsulation devices that can 2399 * operate in "collect metadata" mode (please refer to the related 2400 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2401 * more details). A particular example where this can be used is 2402 * in combination with the Geneve encapsulation protocol, where it 2403 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2404 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2405 * the eBPF program. This allows for full customization of these 2406 * headers. 2407 * Return 2408 * The size of the option data retrieved. 2409 * 2410 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2411 * Description 2412 * Set tunnel options metadata for the packet associated to *skb* 2413 * to the option data contained in the raw buffer *opt* of *size*. 2414 * 2415 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2416 * helper for additional information. 2417 * Return 2418 * 0 on success, or a negative error in case of failure. 2419 * 2420 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2421 * Description 2422 * Change the protocol of the *skb* to *proto*. Currently 2423 * supported are transition from IPv4 to IPv6, and from IPv6 to 2424 * IPv4. The helper takes care of the groundwork for the 2425 * transition, including resizing the socket buffer. The eBPF 2426 * program is expected to fill the new headers, if any, via 2427 * **skb_store_bytes**\ () and to recompute the checksums with 2428 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2429 * (). The main case for this helper is to perform NAT64 2430 * operations out of an eBPF program. 2431 * 2432 * Internally, the GSO type is marked as dodgy so that headers are 2433 * checked and segments are recalculated by the GSO/GRO engine. 2434 * The size for GSO target is adapted as well. 2435 * 2436 * All values for *flags* are reserved for future usage, and must 2437 * be left at zero. 2438 * 2439 * A call to this helper is susceptible to change the underlying 2440 * packet buffer. Therefore, at load time, all checks on pointers 2441 * previously done by the verifier are invalidated and must be 2442 * performed again, if the helper is used in combination with 2443 * direct packet access. 2444 * Return 2445 * 0 on success, or a negative error in case of failure. 2446 * 2447 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2448 * Description 2449 * Change the packet type for the packet associated to *skb*. This 2450 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2451 * the eBPF program does not have a write access to *skb*\ 2452 * **->pkt_type** beside this helper. Using a helper here allows 2453 * for graceful handling of errors. 2454 * 2455 * The major use case is to change incoming *skb*s to 2456 * **PACKET_HOST** in a programmatic way instead of having to 2457 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2458 * example. 2459 * 2460 * Note that *type* only allows certain values. At this time, they 2461 * are: 2462 * 2463 * **PACKET_HOST** 2464 * Packet is for us. 2465 * **PACKET_BROADCAST** 2466 * Send packet to all. 2467 * **PACKET_MULTICAST** 2468 * Send packet to group. 2469 * **PACKET_OTHERHOST** 2470 * Send packet to someone else. 2471 * Return 2472 * 0 on success, or a negative error in case of failure. 2473 * 2474 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2475 * Description 2476 * Check whether *skb* is a descendant of the cgroup2 held by 2477 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2478 * Return 2479 * The return value depends on the result of the test, and can be: 2480 * 2481 * * 0, if the *skb* failed the cgroup2 descendant test. 2482 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2483 * * A negative error code, if an error occurred. 2484 * 2485 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2486 * Description 2487 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2488 * not set, in particular if the hash was cleared due to mangling, 2489 * recompute this hash. Later accesses to the hash can be done 2490 * directly with *skb*\ **->hash**. 2491 * 2492 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2493 * prototype with **bpf_skb_change_proto**\ (), or calling 2494 * **bpf_skb_store_bytes**\ () with the 2495 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2496 * the hash and to trigger a new computation for the next call to 2497 * **bpf_get_hash_recalc**\ (). 2498 * Return 2499 * The 32-bit hash. 2500 * 2501 * u64 bpf_get_current_task(void) 2502 * Description 2503 * Get the current task. 2504 * Return 2505 * A pointer to the current task struct. 2506 * 2507 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2508 * Description 2509 * Attempt in a safe way to write *len* bytes from the buffer 2510 * *src* to *dst* in memory. It only works for threads that are in 2511 * user context, and *dst* must be a valid user space address. 2512 * 2513 * This helper should not be used to implement any kind of 2514 * security mechanism because of TOC-TOU attacks, but rather to 2515 * debug, divert, and manipulate execution of semi-cooperative 2516 * processes. 2517 * 2518 * Keep in mind that this feature is meant for experiments, and it 2519 * has a risk of crashing the system and running programs. 2520 * Therefore, when an eBPF program using this helper is attached, 2521 * a warning including PID and process name is printed to kernel 2522 * logs. 2523 * Return 2524 * 0 on success, or a negative error in case of failure. 2525 * 2526 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2527 * Description 2528 * Check whether the probe is being run is the context of a given 2529 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2530 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2531 * Return 2532 * The return value depends on the result of the test, and can be: 2533 * 2534 * * 1, if current task belongs to the cgroup2. 2535 * * 0, if current task does not belong to the cgroup2. 2536 * * A negative error code, if an error occurred. 2537 * 2538 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2539 * Description 2540 * Resize (trim or grow) the packet associated to *skb* to the 2541 * new *len*. The *flags* are reserved for future usage, and must 2542 * be left at zero. 2543 * 2544 * The basic idea is that the helper performs the needed work to 2545 * change the size of the packet, then the eBPF program rewrites 2546 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2547 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2548 * and others. This helper is a slow path utility intended for 2549 * replies with control messages. And because it is targeted for 2550 * slow path, the helper itself can afford to be slow: it 2551 * implicitly linearizes, unclones and drops offloads from the 2552 * *skb*. 2553 * 2554 * A call to this helper is susceptible to change the underlying 2555 * packet buffer. Therefore, at load time, all checks on pointers 2556 * previously done by the verifier are invalidated and must be 2557 * performed again, if the helper is used in combination with 2558 * direct packet access. 2559 * Return 2560 * 0 on success, or a negative error in case of failure. 2561 * 2562 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2563 * Description 2564 * Pull in non-linear data in case the *skb* is non-linear and not 2565 * all of *len* are part of the linear section. Make *len* bytes 2566 * from *skb* readable and writable. If a zero value is passed for 2567 * *len*, then all bytes in the linear part of *skb* will be made 2568 * readable and writable. 2569 * 2570 * This helper is only needed for reading and writing with direct 2571 * packet access. 2572 * 2573 * For direct packet access, testing that offsets to access 2574 * are within packet boundaries (test on *skb*\ **->data_end**) is 2575 * susceptible to fail if offsets are invalid, or if the requested 2576 * data is in non-linear parts of the *skb*. On failure the 2577 * program can just bail out, or in the case of a non-linear 2578 * buffer, use a helper to make the data available. The 2579 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2580 * the data. Another one consists in using **bpf_skb_pull_data** 2581 * to pull in once the non-linear parts, then retesting and 2582 * eventually access the data. 2583 * 2584 * At the same time, this also makes sure the *skb* is uncloned, 2585 * which is a necessary condition for direct write. As this needs 2586 * to be an invariant for the write part only, the verifier 2587 * detects writes and adds a prologue that is calling 2588 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2589 * the very beginning in case it is indeed cloned. 2590 * 2591 * A call to this helper is susceptible to change the underlying 2592 * packet buffer. Therefore, at load time, all checks on pointers 2593 * previously done by the verifier are invalidated and must be 2594 * performed again, if the helper is used in combination with 2595 * direct packet access. 2596 * Return 2597 * 0 on success, or a negative error in case of failure. 2598 * 2599 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2600 * Description 2601 * Add the checksum *csum* into *skb*\ **->csum** in case the 2602 * driver has supplied a checksum for the entire packet into that 2603 * field. Return an error otherwise. This helper is intended to be 2604 * used in combination with **bpf_csum_diff**\ (), in particular 2605 * when the checksum needs to be updated after data has been 2606 * written into the packet through direct packet access. 2607 * Return 2608 * The checksum on success, or a negative error code in case of 2609 * failure. 2610 * 2611 * void bpf_set_hash_invalid(struct sk_buff *skb) 2612 * Description 2613 * Invalidate the current *skb*\ **->hash**. It can be used after 2614 * mangling on headers through direct packet access, in order to 2615 * indicate that the hash is outdated and to trigger a 2616 * recalculation the next time the kernel tries to access this 2617 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2618 * Return 2619 * void. 2620 * 2621 * long bpf_get_numa_node_id(void) 2622 * Description 2623 * Return the id of the current NUMA node. The primary use case 2624 * for this helper is the selection of sockets for the local NUMA 2625 * node, when the program is attached to sockets using the 2626 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2627 * but the helper is also available to other eBPF program types, 2628 * similarly to **bpf_get_smp_processor_id**\ (). 2629 * Return 2630 * The id of current NUMA node. 2631 * 2632 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2633 * Description 2634 * Grows headroom of packet associated to *skb* and adjusts the 2635 * offset of the MAC header accordingly, adding *len* bytes of 2636 * space. It automatically extends and reallocates memory as 2637 * required. 2638 * 2639 * This helper can be used on a layer 3 *skb* to push a MAC header 2640 * for redirection into a layer 2 device. 2641 * 2642 * All values for *flags* are reserved for future usage, and must 2643 * be left at zero. 2644 * 2645 * A call to this helper is susceptible to change the underlying 2646 * packet buffer. Therefore, at load time, all checks on pointers 2647 * previously done by the verifier are invalidated and must be 2648 * performed again, if the helper is used in combination with 2649 * direct packet access. 2650 * Return 2651 * 0 on success, or a negative error in case of failure. 2652 * 2653 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2654 * Description 2655 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2656 * it is possible to use a negative value for *delta*. This helper 2657 * can be used to prepare the packet for pushing or popping 2658 * headers. 2659 * 2660 * A call to this helper is susceptible to change the underlying 2661 * packet buffer. Therefore, at load time, all checks on pointers 2662 * previously done by the verifier are invalidated and must be 2663 * performed again, if the helper is used in combination with 2664 * direct packet access. 2665 * Return 2666 * 0 on success, or a negative error in case of failure. 2667 * 2668 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2669 * Description 2670 * Copy a NUL terminated string from an unsafe kernel address 2671 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2672 * more details. 2673 * 2674 * Generally, use **bpf_probe_read_user_str**\ () or 2675 * **bpf_probe_read_kernel_str**\ () instead. 2676 * Return 2677 * On success, the strictly positive length of the string, 2678 * including the trailing NUL character. On error, a negative 2679 * value. 2680 * 2681 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2682 * Description 2683 * If the **struct sk_buff** pointed by *skb* has a known socket, 2684 * retrieve the cookie (generated by the kernel) of this socket. 2685 * If no cookie has been set yet, generate a new cookie. Once 2686 * generated, the socket cookie remains stable for the life of the 2687 * socket. This helper can be useful for monitoring per socket 2688 * networking traffic statistics as it provides a global socket 2689 * identifier that can be assumed unique. 2690 * Return 2691 * A 8-byte long unique number on success, or 0 if the socket 2692 * field is missing inside *skb*. 2693 * 2694 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2695 * Description 2696 * Equivalent to bpf_get_socket_cookie() helper that accepts 2697 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2698 * Return 2699 * A 8-byte long unique number. 2700 * 2701 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2702 * Description 2703 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2704 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2705 * Return 2706 * A 8-byte long unique number. 2707 * 2708 * u64 bpf_get_socket_cookie(struct sock *sk) 2709 * Description 2710 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2711 * *sk*, but gets socket from a BTF **struct sock**. This helper 2712 * also works for sleepable programs. 2713 * Return 2714 * A 8-byte long unique number or 0 if *sk* is NULL. 2715 * 2716 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2717 * Description 2718 * Get the owner UID of the socked associated to *skb*. 2719 * Return 2720 * The owner UID of the socket associated to *skb*. If the socket 2721 * is **NULL**, or if it is not a full socket (i.e. if it is a 2722 * time-wait or a request socket instead), **overflowuid** value 2723 * is returned (note that **overflowuid** might also be the actual 2724 * UID value for the socket). 2725 * 2726 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2727 * Description 2728 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2729 * to value *hash*. 2730 * Return 2731 * 0 2732 * 2733 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2734 * Description 2735 * Emulate a call to **setsockopt()** on the socket associated to 2736 * *bpf_socket*, which must be a full socket. The *level* at 2737 * which the option resides and the name *optname* of the option 2738 * must be specified, see **setsockopt(2)** for more information. 2739 * The option value of length *optlen* is pointed by *optval*. 2740 * 2741 * *bpf_socket* should be one of the following: 2742 * 2743 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2744 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2745 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2746 * 2747 * This helper actually implements a subset of **setsockopt()**. 2748 * It supports the following *level*\ s: 2749 * 2750 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2751 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2752 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2753 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2754 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2755 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2756 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2757 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2758 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2759 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2760 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2761 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2762 * **TCP_BPF_RTO_MIN**. 2763 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2764 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2765 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2766 * Return 2767 * 0 on success, or a negative error in case of failure. 2768 * 2769 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2770 * Description 2771 * Grow or shrink the room for data in the packet associated to 2772 * *skb* by *len_diff*, and according to the selected *mode*. 2773 * 2774 * By default, the helper will reset any offloaded checksum 2775 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2776 * by the following flag: 2777 * 2778 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2779 * checksum data of the skb to CHECKSUM_NONE. 2780 * 2781 * There are two supported modes at this time: 2782 * 2783 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2784 * (room space is added or removed between the layer 2 and 2785 * layer 3 headers). 2786 * 2787 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2788 * (room space is added or removed between the layer 3 and 2789 * layer 4 headers). 2790 * 2791 * The following flags are supported at this time: 2792 * 2793 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2794 * Adjusting mss in this way is not allowed for datagrams. 2795 * 2796 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2797 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2798 * Any new space is reserved to hold a tunnel header. 2799 * Configure skb offsets and other fields accordingly. 2800 * 2801 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2802 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2803 * Use with ENCAP_L3 flags to further specify the tunnel type. 2804 * 2805 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2806 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2807 * type; *len* is the length of the inner MAC header. 2808 * 2809 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2810 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2811 * L2 type as Ethernet. 2812 * 2813 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2814 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2815 * Indicate the new IP header version after decapsulating the outer 2816 * IP header. Used when the inner and outer IP versions are different. 2817 * 2818 * A call to this helper is susceptible to change the underlying 2819 * packet buffer. Therefore, at load time, all checks on pointers 2820 * previously done by the verifier are invalidated and must be 2821 * performed again, if the helper is used in combination with 2822 * direct packet access. 2823 * Return 2824 * 0 on success, or a negative error in case of failure. 2825 * 2826 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2827 * Description 2828 * Redirect the packet to the endpoint referenced by *map* at 2829 * index *key*. Depending on its type, this *map* can contain 2830 * references to net devices (for forwarding packets through other 2831 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2832 * but this is only implemented for native XDP (with driver 2833 * support) as of this writing). 2834 * 2835 * The lower two bits of *flags* are used as the return code if 2836 * the map lookup fails. This is so that the return value can be 2837 * one of the XDP program return codes up to **XDP_TX**, as chosen 2838 * by the caller. The higher bits of *flags* can be set to 2839 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2840 * 2841 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2842 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2843 * interface will be excluded when do broadcasting. 2844 * 2845 * See also **bpf_redirect**\ (), which only supports redirecting 2846 * to an ifindex, but doesn't require a map to do so. 2847 * Return 2848 * **XDP_REDIRECT** on success, or the value of the two lower bits 2849 * of the *flags* argument on error. 2850 * 2851 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2852 * Description 2853 * Redirect the packet to the socket referenced by *map* (of type 2854 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2855 * egress interfaces can be used for redirection. The 2856 * **BPF_F_INGRESS** value in *flags* is used to make the 2857 * distinction (ingress path is selected if the flag is present, 2858 * egress path otherwise). This is the only flag supported for now. 2859 * Return 2860 * **SK_PASS** on success, or **SK_DROP** on error. 2861 * 2862 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2863 * Description 2864 * Add an entry to, or update a *map* referencing sockets. The 2865 * *skops* is used as a new value for the entry associated to 2866 * *key*. *flags* is one of: 2867 * 2868 * **BPF_NOEXIST** 2869 * The entry for *key* must not exist in the map. 2870 * **BPF_EXIST** 2871 * The entry for *key* must already exist in the map. 2872 * **BPF_ANY** 2873 * No condition on the existence of the entry for *key*. 2874 * 2875 * If the *map* has eBPF programs (parser and verdict), those will 2876 * be inherited by the socket being added. If the socket is 2877 * already attached to eBPF programs, this results in an error. 2878 * Return 2879 * 0 on success, or a negative error in case of failure. 2880 * 2881 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2882 * Description 2883 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2884 * *delta* (which can be positive or negative). Note that this 2885 * operation modifies the address stored in *xdp_md*\ **->data**, 2886 * so the latter must be loaded only after the helper has been 2887 * called. 2888 * 2889 * The use of *xdp_md*\ **->data_meta** is optional and programs 2890 * are not required to use it. The rationale is that when the 2891 * packet is processed with XDP (e.g. as DoS filter), it is 2892 * possible to push further meta data along with it before passing 2893 * to the stack, and to give the guarantee that an ingress eBPF 2894 * program attached as a TC classifier on the same device can pick 2895 * this up for further post-processing. Since TC works with socket 2896 * buffers, it remains possible to set from XDP the **mark** or 2897 * **priority** pointers, or other pointers for the socket buffer. 2898 * Having this scratch space generic and programmable allows for 2899 * more flexibility as the user is free to store whatever meta 2900 * data they need. 2901 * 2902 * A call to this helper is susceptible to change the underlying 2903 * packet buffer. Therefore, at load time, all checks on pointers 2904 * previously done by the verifier are invalidated and must be 2905 * performed again, if the helper is used in combination with 2906 * direct packet access. 2907 * Return 2908 * 0 on success, or a negative error in case of failure. 2909 * 2910 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2911 * Description 2912 * Read the value of a perf event counter, and store it into *buf* 2913 * of size *buf_size*. This helper relies on a *map* of type 2914 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2915 * counter is selected when *map* is updated with perf event file 2916 * descriptors. The *map* is an array whose size is the number of 2917 * available CPUs, and each cell contains a value relative to one 2918 * CPU. The value to retrieve is indicated by *flags*, that 2919 * contains the index of the CPU to look up, masked with 2920 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2921 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2922 * current CPU should be retrieved. 2923 * 2924 * This helper behaves in a way close to 2925 * **bpf_perf_event_read**\ () helper, save that instead of 2926 * just returning the value observed, it fills the *buf* 2927 * structure. This allows for additional data to be retrieved: in 2928 * particular, the enabled and running times (in *buf*\ 2929 * **->enabled** and *buf*\ **->running**, respectively) are 2930 * copied. In general, **bpf_perf_event_read_value**\ () is 2931 * recommended over **bpf_perf_event_read**\ (), which has some 2932 * ABI issues and provides fewer functionalities. 2933 * 2934 * These values are interesting, because hardware PMU (Performance 2935 * Monitoring Unit) counters are limited resources. When there are 2936 * more PMU based perf events opened than available counters, 2937 * kernel will multiplex these events so each event gets certain 2938 * percentage (but not all) of the PMU time. In case that 2939 * multiplexing happens, the number of samples or counter value 2940 * will not reflect the case compared to when no multiplexing 2941 * occurs. This makes comparison between different runs difficult. 2942 * Typically, the counter value should be normalized before 2943 * comparing to other experiments. The usual normalization is done 2944 * as follows. 2945 * 2946 * :: 2947 * 2948 * normalized_counter = counter * t_enabled / t_running 2949 * 2950 * Where t_enabled is the time enabled for event and t_running is 2951 * the time running for event since last normalization. The 2952 * enabled and running times are accumulated since the perf event 2953 * open. To achieve scaling factor between two invocations of an 2954 * eBPF program, users can use CPU id as the key (which is 2955 * typical for perf array usage model) to remember the previous 2956 * value and do the calculation inside the eBPF program. 2957 * Return 2958 * 0 on success, or a negative error in case of failure. 2959 * 2960 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2961 * Description 2962 * For an eBPF program attached to a perf event, retrieve the 2963 * value of the event counter associated to *ctx* and store it in 2964 * the structure pointed by *buf* and of size *buf_size*. Enabled 2965 * and running times are also stored in the structure (see 2966 * description of helper **bpf_perf_event_read_value**\ () for 2967 * more details). 2968 * Return 2969 * 0 on success, or a negative error in case of failure. 2970 * 2971 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2972 * Description 2973 * Emulate a call to **getsockopt()** on the socket associated to 2974 * *bpf_socket*, which must be a full socket. The *level* at 2975 * which the option resides and the name *optname* of the option 2976 * must be specified, see **getsockopt(2)** for more information. 2977 * The retrieved value is stored in the structure pointed by 2978 * *opval* and of length *optlen*. 2979 * 2980 * *bpf_socket* should be one of the following: 2981 * 2982 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2983 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2984 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2985 * 2986 * This helper actually implements a subset of **getsockopt()**. 2987 * It supports the same set of *optname*\ s that is supported by 2988 * the **bpf_setsockopt**\ () helper. The exceptions are 2989 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 2990 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 2991 * Return 2992 * 0 on success, or a negative error in case of failure. 2993 * 2994 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2995 * Description 2996 * Used for error injection, this helper uses kprobes to override 2997 * the return value of the probed function, and to set it to *rc*. 2998 * The first argument is the context *regs* on which the kprobe 2999 * works. 3000 * 3001 * This helper works by setting the PC (program counter) 3002 * to an override function which is run in place of the original 3003 * probed function. This means the probed function is not run at 3004 * all. The replacement function just returns with the required 3005 * value. 3006 * 3007 * This helper has security implications, and thus is subject to 3008 * restrictions. It is only available if the kernel was compiled 3009 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3010 * option, and in this case it only works on functions tagged with 3011 * **ALLOW_ERROR_INJECTION** in the kernel code. 3012 * 3013 * Also, the helper is only available for the architectures having 3014 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 3015 * x86 architecture is the only one to support this feature. 3016 * Return 3017 * 0 3018 * 3019 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3020 * Description 3021 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3022 * for the full TCP socket associated to *bpf_sock_ops* to 3023 * *argval*. 3024 * 3025 * The primary use of this field is to determine if there should 3026 * be calls to eBPF programs of type 3027 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3028 * code. A program of the same type can change its value, per 3029 * connection and as necessary, when the connection is 3030 * established. This field is directly accessible for reading, but 3031 * this helper must be used for updates in order to return an 3032 * error if an eBPF program tries to set a callback that is not 3033 * supported in the current kernel. 3034 * 3035 * *argval* is a flag array which can combine these flags: 3036 * 3037 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3038 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3039 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3040 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3041 * 3042 * Therefore, this function can be used to clear a callback flag by 3043 * setting the appropriate bit to zero. e.g. to disable the RTO 3044 * callback: 3045 * 3046 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3047 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3048 * 3049 * Here are some examples of where one could call such eBPF 3050 * program: 3051 * 3052 * * When RTO fires. 3053 * * When a packet is retransmitted. 3054 * * When the connection terminates. 3055 * * When a packet is sent. 3056 * * When a packet is received. 3057 * Return 3058 * Code **-EINVAL** if the socket is not a full TCP socket; 3059 * otherwise, a positive number containing the bits that could not 3060 * be set is returned (which comes down to 0 if all bits were set 3061 * as required). 3062 * 3063 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3064 * Description 3065 * This helper is used in programs implementing policies at the 3066 * socket level. If the message *msg* is allowed to pass (i.e. if 3067 * the verdict eBPF program returns **SK_PASS**), redirect it to 3068 * the socket referenced by *map* (of type 3069 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3070 * egress interfaces can be used for redirection. The 3071 * **BPF_F_INGRESS** value in *flags* is used to make the 3072 * distinction (ingress path is selected if the flag is present, 3073 * egress path otherwise). This is the only flag supported for now. 3074 * Return 3075 * **SK_PASS** on success, or **SK_DROP** on error. 3076 * 3077 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3078 * Description 3079 * For socket policies, apply the verdict of the eBPF program to 3080 * the next *bytes* (number of bytes) of message *msg*. 3081 * 3082 * For example, this helper can be used in the following cases: 3083 * 3084 * * A single **sendmsg**\ () or **sendfile**\ () system call 3085 * contains multiple logical messages that the eBPF program is 3086 * supposed to read and for which it should apply a verdict. 3087 * * An eBPF program only cares to read the first *bytes* of a 3088 * *msg*. If the message has a large payload, then setting up 3089 * and calling the eBPF program repeatedly for all bytes, even 3090 * though the verdict is already known, would create unnecessary 3091 * overhead. 3092 * 3093 * When called from within an eBPF program, the helper sets a 3094 * counter internal to the BPF infrastructure, that is used to 3095 * apply the last verdict to the next *bytes*. If *bytes* is 3096 * smaller than the current data being processed from a 3097 * **sendmsg**\ () or **sendfile**\ () system call, the first 3098 * *bytes* will be sent and the eBPF program will be re-run with 3099 * the pointer for start of data pointing to byte number *bytes* 3100 * **+ 1**. If *bytes* is larger than the current data being 3101 * processed, then the eBPF verdict will be applied to multiple 3102 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3103 * consumed. 3104 * 3105 * Note that if a socket closes with the internal counter holding 3106 * a non-zero value, this is not a problem because data is not 3107 * being buffered for *bytes* and is sent as it is received. 3108 * Return 3109 * 0 3110 * 3111 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3112 * Description 3113 * For socket policies, prevent the execution of the verdict eBPF 3114 * program for message *msg* until *bytes* (byte number) have been 3115 * accumulated. 3116 * 3117 * This can be used when one needs a specific number of bytes 3118 * before a verdict can be assigned, even if the data spans 3119 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3120 * case would be a user calling **sendmsg**\ () repeatedly with 3121 * 1-byte long message segments. Obviously, this is bad for 3122 * performance, but it is still valid. If the eBPF program needs 3123 * *bytes* bytes to validate a header, this helper can be used to 3124 * prevent the eBPF program to be called again until *bytes* have 3125 * been accumulated. 3126 * Return 3127 * 0 3128 * 3129 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3130 * Description 3131 * For socket policies, pull in non-linear data from user space 3132 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3133 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3134 * respectively. 3135 * 3136 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3137 * *msg* it can only parse data that the (**data**, **data_end**) 3138 * pointers have already consumed. For **sendmsg**\ () hooks this 3139 * is likely the first scatterlist element. But for calls relying 3140 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3141 * be the range (**0**, **0**) because the data is shared with 3142 * user space and by default the objective is to avoid allowing 3143 * user space to modify data while (or after) eBPF verdict is 3144 * being decided. This helper can be used to pull in data and to 3145 * set the start and end pointer to given values. Data will be 3146 * copied if necessary (i.e. if data was not linear and if start 3147 * and end pointers do not point to the same chunk). 3148 * 3149 * A call to this helper is susceptible to change the underlying 3150 * packet buffer. Therefore, at load time, all checks on pointers 3151 * previously done by the verifier are invalidated and must be 3152 * performed again, if the helper is used in combination with 3153 * direct packet access. 3154 * 3155 * All values for *flags* are reserved for future usage, and must 3156 * be left at zero. 3157 * Return 3158 * 0 on success, or a negative error in case of failure. 3159 * 3160 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3161 * Description 3162 * Bind the socket associated to *ctx* to the address pointed by 3163 * *addr*, of length *addr_len*. This allows for making outgoing 3164 * connection from the desired IP address, which can be useful for 3165 * example when all processes inside a cgroup should use one 3166 * single IP address on a host that has multiple IP configured. 3167 * 3168 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3169 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3170 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3171 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3172 * behavior and lets the kernel efficiently pick up an unused 3173 * port as long as 4-tuple is unique. Passing non-zero port might 3174 * lead to degraded performance. 3175 * Return 3176 * 0 on success, or a negative error in case of failure. 3177 * 3178 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3179 * Description 3180 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3181 * possible to both shrink and grow the packet tail. 3182 * Shrink done via *delta* being a negative integer. 3183 * 3184 * A call to this helper is susceptible to change the underlying 3185 * packet buffer. Therefore, at load time, all checks on pointers 3186 * previously done by the verifier are invalidated and must be 3187 * performed again, if the helper is used in combination with 3188 * direct packet access. 3189 * Return 3190 * 0 on success, or a negative error in case of failure. 3191 * 3192 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3193 * Description 3194 * Retrieve the XFRM state (IP transform framework, see also 3195 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3196 * 3197 * The retrieved value is stored in the **struct bpf_xfrm_state** 3198 * pointed by *xfrm_state* and of length *size*. 3199 * 3200 * All values for *flags* are reserved for future usage, and must 3201 * be left at zero. 3202 * 3203 * This helper is available only if the kernel was compiled with 3204 * **CONFIG_XFRM** configuration option. 3205 * Return 3206 * 0 on success, or a negative error in case of failure. 3207 * 3208 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3209 * Description 3210 * Return a user or a kernel stack in bpf program provided buffer. 3211 * To achieve this, the helper needs *ctx*, which is a pointer 3212 * to the context on which the tracing program is executed. 3213 * To store the stacktrace, the bpf program provides *buf* with 3214 * a nonnegative *size*. 3215 * 3216 * The last argument, *flags*, holds the number of stack frames to 3217 * skip (from 0 to 255), masked with 3218 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3219 * the following flags: 3220 * 3221 * **BPF_F_USER_STACK** 3222 * Collect a user space stack instead of a kernel stack. 3223 * **BPF_F_USER_BUILD_ID** 3224 * Collect (build_id, file_offset) instead of ips for user 3225 * stack, only valid if **BPF_F_USER_STACK** is also 3226 * specified. 3227 * 3228 * *file_offset* is an offset relative to the beginning 3229 * of the executable or shared object file backing the vma 3230 * which the *ip* falls in. It is *not* an offset relative 3231 * to that object's base address. Accordingly, it must be 3232 * adjusted by adding (sh_addr - sh_offset), where 3233 * sh_{addr,offset} correspond to the executable section 3234 * containing *file_offset* in the object, for comparisons 3235 * to symbols' st_value to be valid. 3236 * 3237 * **bpf_get_stack**\ () can collect up to 3238 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3239 * to sufficient large buffer size. Note that 3240 * this limit can be controlled with the **sysctl** program, and 3241 * that it should be manually increased in order to profile long 3242 * user stacks (such as stacks for Java programs). To do so, use: 3243 * 3244 * :: 3245 * 3246 * # sysctl kernel.perf_event_max_stack=
3247 * Return 3248 * The non-negative copied *buf* length equal to or less than 3249 * *size* on success, or a negative error in case of failure. 3250 * 3251 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3252 * Description 3253 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3254 * it provides an easy way to load *len* bytes from *offset* 3255 * from the packet associated to *skb*, into the buffer pointed 3256 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3257 * a fifth argument *start_header* exists in order to select a 3258 * base offset to start from. *start_header* can be one of: 3259 * 3260 * **BPF_HDR_START_MAC** 3261 * Base offset to load data from is *skb*'s mac header. 3262 * **BPF_HDR_START_NET** 3263 * Base offset to load data from is *skb*'s network header. 3264 * 3265 * In general, "direct packet access" is the preferred method to 3266 * access packet data, however, this helper is in particular useful 3267 * in socket filters where *skb*\ **->data** does not always point 3268 * to the start of the mac header and where "direct packet access" 3269 * is not available. 3270 * Return 3271 * 0 on success, or a negative error in case of failure. 3272 * 3273 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3274 * Description 3275 * Do FIB lookup in kernel tables using parameters in *params*. 3276 * If lookup is successful and result shows packet is to be 3277 * forwarded, the neighbor tables are searched for the nexthop. 3278 * If successful (ie., FIB lookup shows forwarding and nexthop 3279 * is resolved), the nexthop address is returned in ipv4_dst 3280 * or ipv6_dst based on family, smac is set to mac address of 3281 * egress device, dmac is set to nexthop mac address, rt_metric 3282 * is set to metric from route (IPv4/IPv6 only), and ifindex 3283 * is set to the device index of the nexthop from the FIB lookup. 3284 * 3285 * *plen* argument is the size of the passed in struct. 3286 * *flags* argument can be a combination of one or more of the 3287 * following values: 3288 * 3289 * **BPF_FIB_LOOKUP_DIRECT** 3290 * Do a direct table lookup vs full lookup using FIB 3291 * rules. 3292 * **BPF_FIB_LOOKUP_TBID** 3293 * Used with BPF_FIB_LOOKUP_DIRECT. 3294 * Use the routing table ID present in *params*->tbid 3295 * for the fib lookup. 3296 * **BPF_FIB_LOOKUP_OUTPUT** 3297 * Perform lookup from an egress perspective (default is 3298 * ingress). 3299 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3300 * Skip the neighbour table lookup. *params*->dmac 3301 * and *params*->smac will not be set as output. A common 3302 * use case is to call **bpf_redirect_neigh**\ () after 3303 * doing **bpf_fib_lookup**\ (). 3304 * **BPF_FIB_LOOKUP_SRC** 3305 * Derive and set source IP addr in *params*->ipv{4,6}_src 3306 * for the nexthop. If the src addr cannot be derived, 3307 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3308 * case, *params*->dmac and *params*->smac are not set either. 3309 * 3310 * *ctx* is either **struct xdp_md** for XDP programs or 3311 * **struct sk_buff** tc cls_act programs. 3312 * Return 3313 * * < 0 if any input argument is invalid 3314 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3315 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3316 * packet is not forwarded or needs assist from full stack 3317 * 3318 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3319 * was exceeded and output params->mtu_result contains the MTU. 3320 * 3321 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3322 * Description 3323 * Add an entry to, or update a sockhash *map* referencing sockets. 3324 * The *skops* is used as a new value for the entry associated to 3325 * *key*. *flags* is one of: 3326 * 3327 * **BPF_NOEXIST** 3328 * The entry for *key* must not exist in the map. 3329 * **BPF_EXIST** 3330 * The entry for *key* must already exist in the map. 3331 * **BPF_ANY** 3332 * No condition on the existence of the entry for *key*. 3333 * 3334 * If the *map* has eBPF programs (parser and verdict), those will 3335 * be inherited by the socket being added. If the socket is 3336 * already attached to eBPF programs, this results in an error. 3337 * Return 3338 * 0 on success, or a negative error in case of failure. 3339 * 3340 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3341 * Description 3342 * This helper is used in programs implementing policies at the 3343 * socket level. If the message *msg* is allowed to pass (i.e. if 3344 * the verdict eBPF program returns **SK_PASS**), redirect it to 3345 * the socket referenced by *map* (of type 3346 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3347 * egress interfaces can be used for redirection. The 3348 * **BPF_F_INGRESS** value in *flags* is used to make the 3349 * distinction (ingress path is selected if the flag is present, 3350 * egress path otherwise). This is the only flag supported for now. 3351 * Return 3352 * **SK_PASS** on success, or **SK_DROP** on error. 3353 * 3354 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3355 * Description 3356 * This helper is used in programs implementing policies at the 3357 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3358 * if the verdict eBPF program returns **SK_PASS**), redirect it 3359 * to the socket referenced by *map* (of type 3360 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3361 * egress interfaces can be used for redirection. The 3362 * **BPF_F_INGRESS** value in *flags* is used to make the 3363 * distinction (ingress path is selected if the flag is present, 3364 * egress otherwise). This is the only flag supported for now. 3365 * Return 3366 * **SK_PASS** on success, or **SK_DROP** on error. 3367 * 3368 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3369 * Description 3370 * Encapsulate the packet associated to *skb* within a Layer 3 3371 * protocol header. This header is provided in the buffer at 3372 * address *hdr*, with *len* its size in bytes. *type* indicates 3373 * the protocol of the header and can be one of: 3374 * 3375 * **BPF_LWT_ENCAP_SEG6** 3376 * IPv6 encapsulation with Segment Routing Header 3377 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3378 * the IPv6 header is computed by the kernel. 3379 * **BPF_LWT_ENCAP_SEG6_INLINE** 3380 * Only works if *skb* contains an IPv6 packet. Insert a 3381 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3382 * the IPv6 header. 3383 * **BPF_LWT_ENCAP_IP** 3384 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3385 * must be IPv4 or IPv6, followed by zero or more 3386 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3387 * total bytes in all prepended headers. Please note that 3388 * if **skb_is_gso**\ (*skb*) is true, no more than two 3389 * headers can be prepended, and the inner header, if 3390 * present, should be either GRE or UDP/GUE. 3391 * 3392 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3393 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3394 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3395 * **BPF_PROG_TYPE_LWT_XMIT**. 3396 * 3397 * A call to this helper is susceptible to change the underlying 3398 * packet buffer. Therefore, at load time, all checks on pointers 3399 * previously done by the verifier are invalidated and must be 3400 * performed again, if the helper is used in combination with 3401 * direct packet access. 3402 * Return 3403 * 0 on success, or a negative error in case of failure. 3404 * 3405 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3406 * Description 3407 * Store *len* bytes from address *from* into the packet 3408 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3409 * inside the outermost IPv6 Segment Routing Header can be 3410 * modified through this helper. 3411 * 3412 * A call to this helper is susceptible to change the underlying 3413 * packet buffer. Therefore, at load time, all checks on pointers 3414 * previously done by the verifier are invalidated and must be 3415 * performed again, if the helper is used in combination with 3416 * direct packet access. 3417 * Return 3418 * 0 on success, or a negative error in case of failure. 3419 * 3420 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3421 * Description 3422 * Adjust the size allocated to TLVs in the outermost IPv6 3423 * Segment Routing Header contained in the packet associated to 3424 * *skb*, at position *offset* by *delta* bytes. Only offsets 3425 * after the segments are accepted. *delta* can be as well 3426 * positive (growing) as negative (shrinking). 3427 * 3428 * A call to this helper is susceptible to change the underlying 3429 * packet buffer. Therefore, at load time, all checks on pointers 3430 * previously done by the verifier are invalidated and must be 3431 * performed again, if the helper is used in combination with 3432 * direct packet access. 3433 * Return 3434 * 0 on success, or a negative error in case of failure. 3435 * 3436 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3437 * Description 3438 * Apply an IPv6 Segment Routing action of type *action* to the 3439 * packet associated to *skb*. Each action takes a parameter 3440 * contained at address *param*, and of length *param_len* bytes. 3441 * *action* can be one of: 3442 * 3443 * **SEG6_LOCAL_ACTION_END_X** 3444 * End.X action: Endpoint with Layer-3 cross-connect. 3445 * Type of *param*: **struct in6_addr**. 3446 * **SEG6_LOCAL_ACTION_END_T** 3447 * End.T action: Endpoint with specific IPv6 table lookup. 3448 * Type of *param*: **int**. 3449 * **SEG6_LOCAL_ACTION_END_B6** 3450 * End.B6 action: Endpoint bound to an SRv6 policy. 3451 * Type of *param*: **struct ipv6_sr_hdr**. 3452 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3453 * End.B6.Encap action: Endpoint bound to an SRv6 3454 * encapsulation policy. 3455 * Type of *param*: **struct ipv6_sr_hdr**. 3456 * 3457 * A call to this helper is susceptible to change the underlying 3458 * packet buffer. Therefore, at load time, all checks on pointers 3459 * previously done by the verifier are invalidated and must be 3460 * performed again, if the helper is used in combination with 3461 * direct packet access. 3462 * Return 3463 * 0 on success, or a negative error in case of failure. 3464 * 3465 * long bpf_rc_repeat(void *ctx) 3466 * Description 3467 * This helper is used in programs implementing IR decoding, to 3468 * report a successfully decoded repeat key message. This delays 3469 * the generation of a key up event for previously generated 3470 * key down event. 3471 * 3472 * Some IR protocols like NEC have a special IR message for 3473 * repeating last button, for when a button is held down. 3474 * 3475 * The *ctx* should point to the lirc sample as passed into 3476 * the program. 3477 * 3478 * This helper is only available is the kernel was compiled with 3479 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3480 * "**y**". 3481 * Return 3482 * 0 3483 * 3484 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3485 * Description 3486 * This helper is used in programs implementing IR decoding, to 3487 * report a successfully decoded key press with *scancode*, 3488 * *toggle* value in the given *protocol*. The scancode will be 3489 * translated to a keycode using the rc keymap, and reported as 3490 * an input key down event. After a period a key up event is 3491 * generated. This period can be extended by calling either 3492 * **bpf_rc_keydown**\ () again with the same values, or calling 3493 * **bpf_rc_repeat**\ (). 3494 * 3495 * Some protocols include a toggle bit, in case the button was 3496 * released and pressed again between consecutive scancodes. 3497 * 3498 * The *ctx* should point to the lirc sample as passed into 3499 * the program. 3500 * 3501 * The *protocol* is the decoded protocol number (see 3502 * **enum rc_proto** for some predefined values). 3503 * 3504 * This helper is only available is the kernel was compiled with 3505 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3506 * "**y**". 3507 * Return 3508 * 0 3509 * 3510 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3511 * Description 3512 * Return the cgroup v2 id of the socket associated with the *skb*. 3513 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3514 * helper for cgroup v1 by providing a tag resp. identifier that 3515 * can be matched on or used for map lookups e.g. to implement 3516 * policy. The cgroup v2 id of a given path in the hierarchy is 3517 * exposed in user space through the f_handle API in order to get 3518 * to the same 64-bit id. 3519 * 3520 * This helper can be used on TC egress path, but not on ingress, 3521 * and is available only if the kernel was compiled with the 3522 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3523 * Return 3524 * The id is returned or 0 in case the id could not be retrieved. 3525 * 3526 * u64 bpf_get_current_cgroup_id(void) 3527 * Description 3528 * Get the current cgroup id based on the cgroup within which 3529 * the current task is running. 3530 * Return 3531 * A 64-bit integer containing the current cgroup id based 3532 * on the cgroup within which the current task is running. 3533 * 3534 * void *bpf_get_local_storage(void *map, u64 flags) 3535 * Description 3536 * Get the pointer to the local storage area. 3537 * The type and the size of the local storage is defined 3538 * by the *map* argument. 3539 * The *flags* meaning is specific for each map type, 3540 * and has to be 0 for cgroup local storage. 3541 * 3542 * Depending on the BPF program type, a local storage area 3543 * can be shared between multiple instances of the BPF program, 3544 * running simultaneously. 3545 * 3546 * A user should care about the synchronization by himself. 3547 * For example, by using the **BPF_ATOMIC** instructions to alter 3548 * the shared data. 3549 * Return 3550 * A pointer to the local storage area. 3551 * 3552 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3553 * Description 3554 * Select a **SO_REUSEPORT** socket from a 3555 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3556 * It checks the selected socket is matching the incoming 3557 * request in the socket buffer. 3558 * Return 3559 * 0 on success, or a negative error in case of failure. 3560 * 3561 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3562 * Description 3563 * Return id of cgroup v2 that is ancestor of cgroup associated 3564 * with the *skb* at the *ancestor_level*. The root cgroup is at 3565 * *ancestor_level* zero and each step down the hierarchy 3566 * increments the level. If *ancestor_level* == level of cgroup 3567 * associated with *skb*, then return value will be same as that 3568 * of **bpf_skb_cgroup_id**\ (). 3569 * 3570 * The helper is useful to implement policies based on cgroups 3571 * that are upper in hierarchy than immediate cgroup associated 3572 * with *skb*. 3573 * 3574 * The format of returned id and helper limitations are same as in 3575 * **bpf_skb_cgroup_id**\ (). 3576 * Return 3577 * The id is returned or 0 in case the id could not be retrieved. 3578 * 3579 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3580 * Description 3581 * Look for TCP socket matching *tuple*, optionally in a child 3582 * network namespace *netns*. The return value must be checked, 3583 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3584 * 3585 * The *ctx* should point to the context of the program, such as 3586 * the skb or socket (depending on the hook in use). This is used 3587 * to determine the base network namespace for the lookup. 3588 * 3589 * *tuple_size* must be one of: 3590 * 3591 * **sizeof**\ (*tuple*\ **->ipv4**) 3592 * Look for an IPv4 socket. 3593 * **sizeof**\ (*tuple*\ **->ipv6**) 3594 * Look for an IPv6 socket. 3595 * 3596 * If the *netns* is a negative signed 32-bit integer, then the 3597 * socket lookup table in the netns associated with the *ctx* 3598 * will be used. For the TC hooks, this is the netns of the device 3599 * in the skb. For socket hooks, this is the netns of the socket. 3600 * If *netns* is any other signed 32-bit value greater than or 3601 * equal to zero then it specifies the ID of the netns relative to 3602 * the netns associated with the *ctx*. *netns* values beyond the 3603 * range of 32-bit integers are reserved for future use. 3604 * 3605 * All values for *flags* are reserved for future usage, and must 3606 * be left at zero. 3607 * 3608 * This helper is available only if the kernel was compiled with 3609 * **CONFIG_NET** configuration option. 3610 * Return 3611 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3612 * For sockets with reuseport option, the **struct bpf_sock** 3613 * result is from *reuse*\ **->socks**\ [] using the hash of the 3614 * tuple. 3615 * 3616 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3617 * Description 3618 * Look for UDP socket matching *tuple*, optionally in a child 3619 * network namespace *netns*. The return value must be checked, 3620 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3621 * 3622 * The *ctx* should point to the context of the program, such as 3623 * the skb or socket (depending on the hook in use). This is used 3624 * to determine the base network namespace for the lookup. 3625 * 3626 * *tuple_size* must be one of: 3627 * 3628 * **sizeof**\ (*tuple*\ **->ipv4**) 3629 * Look for an IPv4 socket. 3630 * **sizeof**\ (*tuple*\ **->ipv6**) 3631 * Look for an IPv6 socket. 3632 * 3633 * If the *netns* is a negative signed 32-bit integer, then the 3634 * socket lookup table in the netns associated with the *ctx* 3635 * will be used. For the TC hooks, this is the netns of the device 3636 * in the skb. For socket hooks, this is the netns of the socket. 3637 * If *netns* is any other signed 32-bit value greater than or 3638 * equal to zero then it specifies the ID of the netns relative to 3639 * the netns associated with the *ctx*. *netns* values beyond the 3640 * range of 32-bit integers are reserved for future use. 3641 * 3642 * All values for *flags* are reserved for future usage, and must 3643 * be left at zero. 3644 * 3645 * This helper is available only if the kernel was compiled with 3646 * **CONFIG_NET** configuration option. 3647 * Return 3648 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3649 * For sockets with reuseport option, the **struct bpf_sock** 3650 * result is from *reuse*\ **->socks**\ [] using the hash of the 3651 * tuple. 3652 * 3653 * long bpf_sk_release(void *sock) 3654 * Description 3655 * Release the reference held by *sock*. *sock* must be a 3656 * non-**NULL** pointer that was returned from 3657 * **bpf_sk_lookup_xxx**\ (). 3658 * Return 3659 * 0 on success, or a negative error in case of failure. 3660 * 3661 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3662 * Description 3663 * Push an element *value* in *map*. *flags* is one of: 3664 * 3665 * **BPF_EXIST** 3666 * If the queue/stack is full, the oldest element is 3667 * removed to make room for this. 3668 * Return 3669 * 0 on success, or a negative error in case of failure. 3670 * 3671 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3672 * Description 3673 * Pop an element from *map*. 3674 * Return 3675 * 0 on success, or a negative error in case of failure. 3676 * 3677 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3678 * Description 3679 * Get an element from *map* without removing it. 3680 * Return 3681 * 0 on success, or a negative error in case of failure. 3682 * 3683 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3684 * Description 3685 * For socket policies, insert *len* bytes into *msg* at offset 3686 * *start*. 3687 * 3688 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3689 * *msg* it may want to insert metadata or options into the *msg*. 3690 * This can later be read and used by any of the lower layer BPF 3691 * hooks. 3692 * 3693 * This helper may fail if under memory pressure (a malloc 3694 * fails) in these cases BPF programs will get an appropriate 3695 * error and BPF programs will need to handle them. 3696 * Return 3697 * 0 on success, or a negative error in case of failure. 3698 * 3699 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3700 * Description 3701 * Will remove *len* bytes from a *msg* starting at byte *start*. 3702 * This may result in **ENOMEM** errors under certain situations if 3703 * an allocation and copy are required due to a full ring buffer. 3704 * However, the helper will try to avoid doing the allocation 3705 * if possible. Other errors can occur if input parameters are 3706 * invalid either due to *start* byte not being valid part of *msg* 3707 * payload and/or *pop* value being to large. 3708 * Return 3709 * 0 on success, or a negative error in case of failure. 3710 * 3711 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3712 * Description 3713 * This helper is used in programs implementing IR decoding, to 3714 * report a successfully decoded pointer movement. 3715 * 3716 * The *ctx* should point to the lirc sample as passed into 3717 * the program. 3718 * 3719 * This helper is only available is the kernel was compiled with 3720 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3721 * "**y**". 3722 * Return 3723 * 0 3724 * 3725 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3726 * Description 3727 * Acquire a spinlock represented by the pointer *lock*, which is 3728 * stored as part of a value of a map. Taking the lock allows to 3729 * safely update the rest of the fields in that value. The 3730 * spinlock can (and must) later be released with a call to 3731 * **bpf_spin_unlock**\ (\ *lock*\ ). 3732 * 3733 * Spinlocks in BPF programs come with a number of restrictions 3734 * and constraints: 3735 * 3736 * * **bpf_spin_lock** objects are only allowed inside maps of 3737 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3738 * list could be extended in the future). 3739 * * BTF description of the map is mandatory. 3740 * * The BPF program can take ONE lock at a time, since taking two 3741 * or more could cause dead locks. 3742 * * Only one **struct bpf_spin_lock** is allowed per map element. 3743 * * When the lock is taken, calls (either BPF to BPF or helpers) 3744 * are not allowed. 3745 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3746 * allowed inside a spinlock-ed region. 3747 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3748 * the lock, on all execution paths, before it returns. 3749 * * The BPF program can access **struct bpf_spin_lock** only via 3750 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3751 * helpers. Loading or storing data into the **struct 3752 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3753 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3754 * of the map value must be a struct and have **struct 3755 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3756 * Nested lock inside another struct is not allowed. 3757 * * The **struct bpf_spin_lock** *lock* field in a map value must 3758 * be aligned on a multiple of 4 bytes in that value. 3759 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3760 * the **bpf_spin_lock** field to user space. 3761 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3762 * a BPF program, do not update the **bpf_spin_lock** field. 3763 * * **bpf_spin_lock** cannot be on the stack or inside a 3764 * networking packet (it can only be inside of a map values). 3765 * * **bpf_spin_lock** is available to root only. 3766 * * Tracing programs and socket filter programs cannot use 3767 * **bpf_spin_lock**\ () due to insufficient preemption checks 3768 * (but this may change in the future). 3769 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3770 * Return 3771 * 0 3772 * 3773 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3774 * Description 3775 * Release the *lock* previously locked by a call to 3776 * **bpf_spin_lock**\ (\ *lock*\ ). 3777 * Return 3778 * 0 3779 * 3780 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3781 * Description 3782 * This helper gets a **struct bpf_sock** pointer such 3783 * that all the fields in this **bpf_sock** can be accessed. 3784 * Return 3785 * A **struct bpf_sock** pointer on success, or **NULL** in 3786 * case of failure. 3787 * 3788 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3789 * Description 3790 * This helper gets a **struct bpf_tcp_sock** pointer from a 3791 * **struct bpf_sock** pointer. 3792 * Return 3793 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3794 * case of failure. 3795 * 3796 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3797 * Description 3798 * Set ECN (Explicit Congestion Notification) field of IP header 3799 * to **CE** (Congestion Encountered) if current value is **ECT** 3800 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3801 * and IPv4. 3802 * Return 3803 * 1 if the **CE** flag is set (either by the current helper call 3804 * or because it was already present), 0 if it is not set. 3805 * 3806 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3807 * Description 3808 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3809 * **bpf_sk_release**\ () is unnecessary and not allowed. 3810 * Return 3811 * A **struct bpf_sock** pointer on success, or **NULL** in 3812 * case of failure. 3813 * 3814 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3815 * Description 3816 * Look for TCP socket matching *tuple*, optionally in a child 3817 * network namespace *netns*. The return value must be checked, 3818 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3819 * 3820 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3821 * that it also returns timewait or request sockets. Use 3822 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3823 * full structure. 3824 * 3825 * This helper is available only if the kernel was compiled with 3826 * **CONFIG_NET** configuration option. 3827 * Return 3828 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3829 * For sockets with reuseport option, the **struct bpf_sock** 3830 * result is from *reuse*\ **->socks**\ [] using the hash of the 3831 * tuple. 3832 * 3833 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3834 * Description 3835 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3836 * the listening socket in *sk*. 3837 * 3838 * *iph* points to the start of the IPv4 or IPv6 header, while 3839 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3840 * **sizeof**\ (**struct ipv6hdr**). 3841 * 3842 * *th* points to the start of the TCP header, while *th_len* 3843 * contains the length of the TCP header (at least 3844 * **sizeof**\ (**struct tcphdr**)). 3845 * Return 3846 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3847 * error otherwise. 3848 * 3849 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3850 * Description 3851 * Get name of sysctl in /proc/sys/ and copy it into provided by 3852 * program buffer *buf* of size *buf_len*. 3853 * 3854 * The buffer is always NUL terminated, unless it's zero-sized. 3855 * 3856 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3857 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3858 * only (e.g. "tcp_mem"). 3859 * Return 3860 * Number of character copied (not including the trailing NUL). 3861 * 3862 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3863 * truncated name in this case). 3864 * 3865 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3866 * Description 3867 * Get current value of sysctl as it is presented in /proc/sys 3868 * (incl. newline, etc), and copy it as a string into provided 3869 * by program buffer *buf* of size *buf_len*. 3870 * 3871 * The whole value is copied, no matter what file position user 3872 * space issued e.g. sys_read at. 3873 * 3874 * The buffer is always NUL terminated, unless it's zero-sized. 3875 * Return 3876 * Number of character copied (not including the trailing NUL). 3877 * 3878 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3879 * truncated name in this case). 3880 * 3881 * **-EINVAL** if current value was unavailable, e.g. because 3882 * sysctl is uninitialized and read returns -EIO for it. 3883 * 3884 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3885 * Description 3886 * Get new value being written by user space to sysctl (before 3887 * the actual write happens) and copy it as a string into 3888 * provided by program buffer *buf* of size *buf_len*. 3889 * 3890 * User space may write new value at file position > 0. 3891 * 3892 * The buffer is always NUL terminated, unless it's zero-sized. 3893 * Return 3894 * Number of character copied (not including the trailing NUL). 3895 * 3896 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3897 * truncated name in this case). 3898 * 3899 * **-EINVAL** if sysctl is being read. 3900 * 3901 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3902 * Description 3903 * Override new value being written by user space to sysctl with 3904 * value provided by program in buffer *buf* of size *buf_len*. 3905 * 3906 * *buf* should contain a string in same form as provided by user 3907 * space on sysctl write. 3908 * 3909 * User space may write new value at file position > 0. To override 3910 * the whole sysctl value file position should be set to zero. 3911 * Return 3912 * 0 on success. 3913 * 3914 * **-E2BIG** if the *buf_len* is too big. 3915 * 3916 * **-EINVAL** if sysctl is being read. 3917 * 3918 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3919 * Description 3920 * Convert the initial part of the string from buffer *buf* of 3921 * size *buf_len* to a long integer according to the given base 3922 * and save the result in *res*. 3923 * 3924 * The string may begin with an arbitrary amount of white space 3925 * (as determined by **isspace**\ (3)) followed by a single 3926 * optional '**-**' sign. 3927 * 3928 * Five least significant bits of *flags* encode base, other bits 3929 * are currently unused. 3930 * 3931 * Base must be either 8, 10, 16 or 0 to detect it automatically 3932 * similar to user space **strtol**\ (3). 3933 * Return 3934 * Number of characters consumed on success. Must be positive but 3935 * no more than *buf_len*. 3936 * 3937 * **-EINVAL** if no valid digits were found or unsupported base 3938 * was provided. 3939 * 3940 * **-ERANGE** if resulting value was out of range. 3941 * 3942 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3943 * Description 3944 * Convert the initial part of the string from buffer *buf* of 3945 * size *buf_len* to an unsigned long integer according to the 3946 * given base and save the result in *res*. 3947 * 3948 * The string may begin with an arbitrary amount of white space 3949 * (as determined by **isspace**\ (3)). 3950 * 3951 * Five least significant bits of *flags* encode base, other bits 3952 * are currently unused. 3953 * 3954 * Base must be either 8, 10, 16 or 0 to detect it automatically 3955 * similar to user space **strtoul**\ (3). 3956 * Return 3957 * Number of characters consumed on success. Must be positive but 3958 * no more than *buf_len*. 3959 * 3960 * **-EINVAL** if no valid digits were found or unsupported base 3961 * was provided. 3962 * 3963 * **-ERANGE** if resulting value was out of range. 3964 * 3965 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3966 * Description 3967 * Get a bpf-local-storage from a *sk*. 3968 * 3969 * Logically, it could be thought of getting the value from 3970 * a *map* with *sk* as the **key**. From this 3971 * perspective, the usage is not much different from 3972 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3973 * helper enforces the key must be a full socket and the map must 3974 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3975 * 3976 * Underneath, the value is stored locally at *sk* instead of 3977 * the *map*. The *map* is used as the bpf-local-storage 3978 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3979 * searched against all bpf-local-storages residing at *sk*. 3980 * 3981 * *sk* is a kernel **struct sock** pointer for LSM program. 3982 * *sk* is a **struct bpf_sock** pointer for other program types. 3983 * 3984 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3985 * used such that a new bpf-local-storage will be 3986 * created if one does not exist. *value* can be used 3987 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3988 * the initial value of a bpf-local-storage. If *value* is 3989 * **NULL**, the new bpf-local-storage will be zero initialized. 3990 * Return 3991 * A bpf-local-storage pointer is returned on success. 3992 * 3993 * **NULL** if not found or there was an error in adding 3994 * a new bpf-local-storage. 3995 * 3996 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3997 * Description 3998 * Delete a bpf-local-storage from a *sk*. 3999 * Return 4000 * 0 on success. 4001 * 4002 * **-ENOENT** if the bpf-local-storage cannot be found. 4003 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4004 * 4005 * long bpf_send_signal(u32 sig) 4006 * Description 4007 * Send signal *sig* to the process of the current task. 4008 * The signal may be delivered to any of this process's threads. 4009 * Return 4010 * 0 on success or successfully queued. 4011 * 4012 * **-EBUSY** if work queue under nmi is full. 4013 * 4014 * **-EINVAL** if *sig* is invalid. 4015 * 4016 * **-EPERM** if no permission to send the *sig*. 4017 * 4018 * **-EAGAIN** if bpf program can try again. 4019 * 4020 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4021 * Description 4022 * Try to issue a SYN cookie for the packet with corresponding 4023 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4024 * 4025 * *iph* points to the start of the IPv4 or IPv6 header, while 4026 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4027 * **sizeof**\ (**struct ipv6hdr**). 4028 * 4029 * *th* points to the start of the TCP header, while *th_len* 4030 * contains the length of the TCP header with options (at least 4031 * **sizeof**\ (**struct tcphdr**)). 4032 * Return 4033 * On success, lower 32 bits hold the generated SYN cookie in 4034 * followed by 16 bits which hold the MSS value for that cookie, 4035 * and the top 16 bits are unused. 4036 * 4037 * On failure, the returned value is one of the following: 4038 * 4039 * **-EINVAL** SYN cookie cannot be issued due to error 4040 * 4041 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4042 * 4043 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4044 * 4045 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4046 * 4047 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4048 * Description 4049 * Write raw *data* blob into a special BPF perf event held by 4050 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4051 * event must have the following attributes: **PERF_SAMPLE_RAW** 4052 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4053 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4054 * 4055 * The *flags* are used to indicate the index in *map* for which 4056 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4057 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4058 * to indicate that the index of the current CPU core should be 4059 * used. 4060 * 4061 * The value to write, of *size*, is passed through eBPF stack and 4062 * pointed by *data*. 4063 * 4064 * *ctx* is a pointer to in-kernel struct sk_buff. 4065 * 4066 * This helper is similar to **bpf_perf_event_output**\ () but 4067 * restricted to raw_tracepoint bpf programs. 4068 * Return 4069 * 0 on success, or a negative error in case of failure. 4070 * 4071 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4072 * Description 4073 * Safely attempt to read *size* bytes from user space address 4074 * *unsafe_ptr* and store the data in *dst*. 4075 * Return 4076 * 0 on success, or a negative error in case of failure. 4077 * 4078 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4079 * Description 4080 * Safely attempt to read *size* bytes from kernel space address 4081 * *unsafe_ptr* and store the data in *dst*. 4082 * Return 4083 * 0 on success, or a negative error in case of failure. 4084 * 4085 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4086 * Description 4087 * Copy a NUL terminated string from an unsafe user address 4088 * *unsafe_ptr* to *dst*. The *size* should include the 4089 * terminating NUL byte. In case the string length is smaller than 4090 * *size*, the target is not padded with further NUL bytes. If the 4091 * string length is larger than *size*, just *size*-1 bytes are 4092 * copied and the last byte is set to NUL. 4093 * 4094 * On success, returns the number of bytes that were written, 4095 * including the terminal NUL. This makes this helper useful in 4096 * tracing programs for reading strings, and more importantly to 4097 * get its length at runtime. See the following snippet: 4098 * 4099 * :: 4100 * 4101 * SEC("kprobe/sys_open") 4102 * void bpf_sys_open(struct pt_regs *ctx) 4103 * { 4104 * char buf[PATHLEN]; // PATHLEN is defined to 256 4105 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4106 * ctx->di); 4107 * 4108 * // Consume buf, for example push it to 4109 * // userspace via bpf_perf_event_output(); we 4110 * // can use res (the string length) as event 4111 * // size, after checking its boundaries. 4112 * } 4113 * 4114 * In comparison, using **bpf_probe_read_user**\ () helper here 4115 * instead to read the string would require to estimate the length 4116 * at compile time, and would often result in copying more memory 4117 * than necessary. 4118 * 4119 * Another useful use case is when parsing individual process 4120 * arguments or individual environment variables navigating 4121 * *current*\ **->mm->arg_start** and *current*\ 4122 * **->mm->env_start**: using this helper and the return value, 4123 * one can quickly iterate at the right offset of the memory area. 4124 * Return 4125 * On success, the strictly positive length of the output string, 4126 * including the trailing NUL character. On error, a negative 4127 * value. 4128 * 4129 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4130 * Description 4131 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4132 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4133 * Return 4134 * On success, the strictly positive length of the string, including 4135 * the trailing NUL character. On error, a negative value. 4136 * 4137 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4138 * Description 4139 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4140 * *rcv_nxt* is the ack_seq to be sent out. 4141 * Return 4142 * 0 on success, or a negative error in case of failure. 4143 * 4144 * long bpf_send_signal_thread(u32 sig) 4145 * Description 4146 * Send signal *sig* to the thread corresponding to the current task. 4147 * Return 4148 * 0 on success or successfully queued. 4149 * 4150 * **-EBUSY** if work queue under nmi is full. 4151 * 4152 * **-EINVAL** if *sig* is invalid. 4153 * 4154 * **-EPERM** if no permission to send the *sig*. 4155 * 4156 * **-EAGAIN** if bpf program can try again. 4157 * 4158 * u64 bpf_jiffies64(void) 4159 * Description 4160 * Obtain the 64bit jiffies 4161 * Return 4162 * The 64 bit jiffies 4163 * 4164 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4165 * Description 4166 * For an eBPF program attached to a perf event, retrieve the 4167 * branch records (**struct perf_branch_entry**) associated to *ctx* 4168 * and store it in the buffer pointed by *buf* up to size 4169 * *size* bytes. 4170 * Return 4171 * On success, number of bytes written to *buf*. On error, a 4172 * negative value. 4173 * 4174 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4175 * instead return the number of bytes required to store all the 4176 * branch entries. If this flag is set, *buf* may be NULL. 4177 * 4178 * **-EINVAL** if arguments invalid or **size** not a multiple 4179 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4180 * 4181 * **-ENOENT** if architecture does not support branch records. 4182 * 4183 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4184 * Description 4185 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4186 * *namespace* will be returned in *nsdata*. 4187 * Return 4188 * 0 on success, or one of the following in case of failure: 4189 * 4190 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4191 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4192 * 4193 * **-ENOENT** if pidns does not exists for the current task. 4194 * 4195 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4196 * Description 4197 * Write raw *data* blob into a special BPF perf event held by 4198 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4199 * event must have the following attributes: **PERF_SAMPLE_RAW** 4200 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4201 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4202 * 4203 * The *flags* are used to indicate the index in *map* for which 4204 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4205 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4206 * to indicate that the index of the current CPU core should be 4207 * used. 4208 * 4209 * The value to write, of *size*, is passed through eBPF stack and 4210 * pointed by *data*. 4211 * 4212 * *ctx* is a pointer to in-kernel struct xdp_buff. 4213 * 4214 * This helper is similar to **bpf_perf_eventoutput**\ () but 4215 * restricted to raw_tracepoint bpf programs. 4216 * Return 4217 * 0 on success, or a negative error in case of failure. 4218 * 4219 * u64 bpf_get_netns_cookie(void *ctx) 4220 * Description 4221 * Retrieve the cookie (generated by the kernel) of the network 4222 * namespace the input *ctx* is associated with. The network 4223 * namespace cookie remains stable for its lifetime and provides 4224 * a global identifier that can be assumed unique. If *ctx* is 4225 * NULL, then the helper returns the cookie for the initial 4226 * network namespace. The cookie itself is very similar to that 4227 * of **bpf_get_socket_cookie**\ () helper, but for network 4228 * namespaces instead of sockets. 4229 * Return 4230 * A 8-byte long opaque number. 4231 * 4232 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4233 * Description 4234 * Return id of cgroup v2 that is ancestor of the cgroup associated 4235 * with the current task at the *ancestor_level*. The root cgroup 4236 * is at *ancestor_level* zero and each step down the hierarchy 4237 * increments the level. If *ancestor_level* == level of cgroup 4238 * associated with the current task, then return value will be the 4239 * same as that of **bpf_get_current_cgroup_id**\ (). 4240 * 4241 * The helper is useful to implement policies based on cgroups 4242 * that are upper in hierarchy than immediate cgroup associated 4243 * with the current task. 4244 * 4245 * The format of returned id and helper limitations are same as in 4246 * **bpf_get_current_cgroup_id**\ (). 4247 * Return 4248 * The id is returned or 0 in case the id could not be retrieved. 4249 * 4250 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4251 * Description 4252 * Helper is overloaded depending on BPF program type. This 4253 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4254 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4255 * 4256 * Assign the *sk* to the *skb*. When combined with appropriate 4257 * routing configuration to receive the packet towards the socket, 4258 * will cause *skb* to be delivered to the specified socket. 4259 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4260 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4261 * interfere with successful delivery to the socket. 4262 * 4263 * This operation is only valid from TC ingress path. 4264 * 4265 * The *flags* argument must be zero. 4266 * Return 4267 * 0 on success, or a negative error in case of failure: 4268 * 4269 * **-EINVAL** if specified *flags* are not supported. 4270 * 4271 * **-ENOENT** if the socket is unavailable for assignment. 4272 * 4273 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4274 * 4275 * **-EOPNOTSUPP** if the operation is not supported, for example 4276 * a call from outside of TC ingress. 4277 * 4278 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4279 * Description 4280 * Helper is overloaded depending on BPF program type. This 4281 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4282 * 4283 * Select the *sk* as a result of a socket lookup. 4284 * 4285 * For the operation to succeed passed socket must be compatible 4286 * with the packet description provided by the *ctx* object. 4287 * 4288 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4289 * be an exact match. While IP family (**AF_INET** or 4290 * **AF_INET6**) must be compatible, that is IPv6 sockets 4291 * that are not v6-only can be selected for IPv4 packets. 4292 * 4293 * Only TCP listeners and UDP unconnected sockets can be 4294 * selected. *sk* can also be NULL to reset any previous 4295 * selection. 4296 * 4297 * *flags* argument can combination of following values: 4298 * 4299 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4300 * socket selection, potentially done by a BPF program 4301 * that ran before us. 4302 * 4303 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4304 * load-balancing within reuseport group for the socket 4305 * being selected. 4306 * 4307 * On success *ctx->sk* will point to the selected socket. 4308 * 4309 * Return 4310 * 0 on success, or a negative errno in case of failure. 4311 * 4312 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4313 * not compatible with packet family (*ctx->family*). 4314 * 4315 * * **-EEXIST** if socket has been already selected, 4316 * potentially by another program, and 4317 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4318 * 4319 * * **-EINVAL** if unsupported flags were specified. 4320 * 4321 * * **-EPROTOTYPE** if socket L4 protocol 4322 * (*sk->protocol*) doesn't match packet protocol 4323 * (*ctx->protocol*). 4324 * 4325 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4326 * state (TCP listening or UDP unconnected). 4327 * 4328 * u64 bpf_ktime_get_boot_ns(void) 4329 * Description 4330 * Return the time elapsed since system boot, in nanoseconds. 4331 * Does include the time the system was suspended. 4332 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4333 * Return 4334 * Current *ktime*. 4335 * 4336 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4337 * Description 4338 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4339 * out the format string. 4340 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4341 * the format string itself. The *data* and *data_len* are format string 4342 * arguments. The *data* are a **u64** array and corresponding format string 4343 * values are stored in the array. For strings and pointers where pointees 4344 * are accessed, only the pointer values are stored in the *data* array. 4345 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4346 * 4347 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4348 * Reading kernel memory may fail due to either invalid address or 4349 * valid address but requiring a major memory fault. If reading kernel memory 4350 * fails, the string for **%s** will be an empty string, and the ip 4351 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4352 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4353 * Return 4354 * 0 on success, or a negative error in case of failure: 4355 * 4356 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4357 * by returning 1 from bpf program. 4358 * 4359 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4360 * 4361 * **-E2BIG** if *fmt* contains too many format specifiers. 4362 * 4363 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4364 * 4365 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4366 * Description 4367 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4368 * The *m* represents the seq_file. The *data* and *len* represent the 4369 * data to write in bytes. 4370 * Return 4371 * 0 on success, or a negative error in case of failure: 4372 * 4373 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4374 * 4375 * u64 bpf_sk_cgroup_id(void *sk) 4376 * Description 4377 * Return the cgroup v2 id of the socket *sk*. 4378 * 4379 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4380 * returned from **bpf_sk_lookup_xxx**\ (), 4381 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4382 * same as in **bpf_skb_cgroup_id**\ (). 4383 * 4384 * This helper is available only if the kernel was compiled with 4385 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4386 * Return 4387 * The id is returned or 0 in case the id could not be retrieved. 4388 * 4389 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4390 * Description 4391 * Return id of cgroup v2 that is ancestor of cgroup associated 4392 * with the *sk* at the *ancestor_level*. The root cgroup is at 4393 * *ancestor_level* zero and each step down the hierarchy 4394 * increments the level. If *ancestor_level* == level of cgroup 4395 * associated with *sk*, then return value will be same as that 4396 * of **bpf_sk_cgroup_id**\ (). 4397 * 4398 * The helper is useful to implement policies based on cgroups 4399 * that are upper in hierarchy than immediate cgroup associated 4400 * with *sk*. 4401 * 4402 * The format of returned id and helper limitations are same as in 4403 * **bpf_sk_cgroup_id**\ (). 4404 * Return 4405 * The id is returned or 0 in case the id could not be retrieved. 4406 * 4407 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4408 * Description 4409 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4410 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4411 * of new data availability is sent. 4412 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4413 * of new data availability is sent unconditionally. 4414 * If **0** is specified in *flags*, an adaptive notification 4415 * of new data availability is sent. 4416 * 4417 * An adaptive notification is a notification sent whenever the user-space 4418 * process has caught up and consumed all available payloads. In case the user-space 4419 * process is still processing a previous payload, then no notification is needed 4420 * as it will process the newly added payload automatically. 4421 * Return 4422 * 0 on success, or a negative error in case of failure. 4423 * 4424 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4425 * Description 4426 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4427 * *flags* must be 0. 4428 * Return 4429 * Valid pointer with *size* bytes of memory available; NULL, 4430 * otherwise. 4431 * 4432 * void bpf_ringbuf_submit(void *data, u64 flags) 4433 * Description 4434 * Submit reserved ring buffer sample, pointed to by *data*. 4435 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4436 * of new data availability is sent. 4437 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4438 * of new data availability is sent unconditionally. 4439 * If **0** is specified in *flags*, an adaptive notification 4440 * of new data availability is sent. 4441 * 4442 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4443 * Return 4444 * Nothing. Always succeeds. 4445 * 4446 * void bpf_ringbuf_discard(void *data, u64 flags) 4447 * Description 4448 * Discard reserved ring buffer sample, pointed to by *data*. 4449 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4450 * of new data availability is sent. 4451 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4452 * of new data availability is sent unconditionally. 4453 * If **0** is specified in *flags*, an adaptive notification 4454 * of new data availability is sent. 4455 * 4456 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4457 * Return 4458 * Nothing. Always succeeds. 4459 * 4460 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4461 * Description 4462 * Query various characteristics of provided ring buffer. What 4463 * exactly is queries is determined by *flags*: 4464 * 4465 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4466 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4467 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4468 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4469 * 4470 * Data returned is just a momentary snapshot of actual values 4471 * and could be inaccurate, so this facility should be used to 4472 * power heuristics and for reporting, not to make 100% correct 4473 * calculation. 4474 * Return 4475 * Requested value, or 0, if *flags* are not recognized. 4476 * 4477 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4478 * Description 4479 * Change the skbs checksum level by one layer up or down, or 4480 * reset it entirely to none in order to have the stack perform 4481 * checksum validation. The level is applicable to the following 4482 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4483 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4484 * through **bpf_skb_adjust_room**\ () helper with passing in 4485 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4486 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4487 * the UDP header is removed. Similarly, an encap of the latter 4488 * into the former could be accompanied by a helper call to 4489 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4490 * skb is still intended to be processed in higher layers of the 4491 * stack instead of just egressing at tc. 4492 * 4493 * There are three supported level settings at this time: 4494 * 4495 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4496 * with CHECKSUM_UNNECESSARY. 4497 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4498 * with CHECKSUM_UNNECESSARY. 4499 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4500 * sets CHECKSUM_NONE to force checksum validation by the stack. 4501 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4502 * skb->csum_level. 4503 * Return 4504 * 0 on success, or a negative error in case of failure. In the 4505 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4506 * is returned or the error code -EACCES in case the skb is not 4507 * subject to CHECKSUM_UNNECESSARY. 4508 * 4509 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4510 * Description 4511 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4512 * Return 4513 * *sk* if casting is valid, or **NULL** otherwise. 4514 * 4515 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4516 * Description 4517 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4518 * Return 4519 * *sk* if casting is valid, or **NULL** otherwise. 4520 * 4521 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4522 * Description 4523 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4524 * Return 4525 * *sk* if casting is valid, or **NULL** otherwise. 4526 * 4527 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4528 * Description 4529 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4530 * Return 4531 * *sk* if casting is valid, or **NULL** otherwise. 4532 * 4533 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4534 * Description 4535 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4536 * Return 4537 * *sk* if casting is valid, or **NULL** otherwise. 4538 * 4539 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4540 * Description 4541 * Return a user or a kernel stack in bpf program provided buffer. 4542 * Note: the user stack will only be populated if the *task* is 4543 * the current task; all other tasks will return -EOPNOTSUPP. 4544 * To achieve this, the helper needs *task*, which is a valid 4545 * pointer to **struct task_struct**. To store the stacktrace, the 4546 * bpf program provides *buf* with a nonnegative *size*. 4547 * 4548 * The last argument, *flags*, holds the number of stack frames to 4549 * skip (from 0 to 255), masked with 4550 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4551 * the following flags: 4552 * 4553 * **BPF_F_USER_STACK** 4554 * Collect a user space stack instead of a kernel stack. 4555 * The *task* must be the current task. 4556 * **BPF_F_USER_BUILD_ID** 4557 * Collect buildid+offset instead of ips for user stack, 4558 * only valid if **BPF_F_USER_STACK** is also specified. 4559 * 4560 * **bpf_get_task_stack**\ () can collect up to 4561 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4562 * to sufficient large buffer size. Note that 4563 * this limit can be controlled with the **sysctl** program, and 4564 * that it should be manually increased in order to profile long 4565 * user stacks (such as stacks for Java programs). To do so, use: 4566 * 4567 * :: 4568 * 4569 * # sysctl kernel.perf_event_max_stack=
4570 * Return 4571 * The non-negative copied *buf* length equal to or less than 4572 * *size* on success, or a negative error in case of failure. 4573 * 4574 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4575 * Description 4576 * Load header option. Support reading a particular TCP header 4577 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4578 * 4579 * If *flags* is 0, it will search the option from the 4580 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4581 * has details on what skb_data contains under different 4582 * *skops*\ **->op**. 4583 * 4584 * The first byte of the *searchby_res* specifies the 4585 * kind that it wants to search. 4586 * 4587 * If the searching kind is an experimental kind 4588 * (i.e. 253 or 254 according to RFC6994). It also 4589 * needs to specify the "magic" which is either 4590 * 2 bytes or 4 bytes. It then also needs to 4591 * specify the size of the magic by using 4592 * the 2nd byte which is "kind-length" of a TCP 4593 * header option and the "kind-length" also 4594 * includes the first 2 bytes "kind" and "kind-length" 4595 * itself as a normal TCP header option also does. 4596 * 4597 * For example, to search experimental kind 254 with 4598 * 2 byte magic 0xeB9F, the searchby_res should be 4599 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4600 * 4601 * To search for the standard window scale option (3), 4602 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4603 * Note, kind-length must be 0 for regular option. 4604 * 4605 * Searching for No-Op (0) and End-of-Option-List (1) are 4606 * not supported. 4607 * 4608 * *len* must be at least 2 bytes which is the minimal size 4609 * of a header option. 4610 * 4611 * Supported flags: 4612 * 4613 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4614 * saved_syn packet or the just-received syn packet. 4615 * 4616 * Return 4617 * > 0 when found, the header option is copied to *searchby_res*. 4618 * The return value is the total length copied. On failure, a 4619 * negative error code is returned: 4620 * 4621 * **-EINVAL** if a parameter is invalid. 4622 * 4623 * **-ENOMSG** if the option is not found. 4624 * 4625 * **-ENOENT** if no syn packet is available when 4626 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4627 * 4628 * **-ENOSPC** if there is not enough space. Only *len* number of 4629 * bytes are copied. 4630 * 4631 * **-EFAULT** on failure to parse the header options in the 4632 * packet. 4633 * 4634 * **-EPERM** if the helper cannot be used under the current 4635 * *skops*\ **->op**. 4636 * 4637 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4638 * Description 4639 * Store header option. The data will be copied 4640 * from buffer *from* with length *len* to the TCP header. 4641 * 4642 * The buffer *from* should have the whole option that 4643 * includes the kind, kind-length, and the actual 4644 * option data. The *len* must be at least kind-length 4645 * long. The kind-length does not have to be 4 byte 4646 * aligned. The kernel will take care of the padding 4647 * and setting the 4 bytes aligned value to th->doff. 4648 * 4649 * This helper will check for duplicated option 4650 * by searching the same option in the outgoing skb. 4651 * 4652 * This helper can only be called during 4653 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4654 * 4655 * Return 4656 * 0 on success, or negative error in case of failure: 4657 * 4658 * **-EINVAL** If param is invalid. 4659 * 4660 * **-ENOSPC** if there is not enough space in the header. 4661 * Nothing has been written 4662 * 4663 * **-EEXIST** if the option already exists. 4664 * 4665 * **-EFAULT** on failure to parse the existing header options. 4666 * 4667 * **-EPERM** if the helper cannot be used under the current 4668 * *skops*\ **->op**. 4669 * 4670 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4671 * Description 4672 * Reserve *len* bytes for the bpf header option. The 4673 * space will be used by **bpf_store_hdr_opt**\ () later in 4674 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4675 * 4676 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4677 * the total number of bytes will be reserved. 4678 * 4679 * This helper can only be called during 4680 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4681 * 4682 * Return 4683 * 0 on success, or negative error in case of failure: 4684 * 4685 * **-EINVAL** if a parameter is invalid. 4686 * 4687 * **-ENOSPC** if there is not enough space in the header. 4688 * 4689 * **-EPERM** if the helper cannot be used under the current 4690 * *skops*\ **->op**. 4691 * 4692 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4693 * Description 4694 * Get a bpf_local_storage from an *inode*. 4695 * 4696 * Logically, it could be thought of as getting the value from 4697 * a *map* with *inode* as the **key**. From this 4698 * perspective, the usage is not much different from 4699 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4700 * helper enforces the key must be an inode and the map must also 4701 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4702 * 4703 * Underneath, the value is stored locally at *inode* instead of 4704 * the *map*. The *map* is used as the bpf-local-storage 4705 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4706 * searched against all bpf_local_storage residing at *inode*. 4707 * 4708 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4709 * used such that a new bpf_local_storage will be 4710 * created if one does not exist. *value* can be used 4711 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4712 * the initial value of a bpf_local_storage. If *value* is 4713 * **NULL**, the new bpf_local_storage will be zero initialized. 4714 * Return 4715 * A bpf_local_storage pointer is returned on success. 4716 * 4717 * **NULL** if not found or there was an error in adding 4718 * a new bpf_local_storage. 4719 * 4720 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4721 * Description 4722 * Delete a bpf_local_storage from an *inode*. 4723 * Return 4724 * 0 on success. 4725 * 4726 * **-ENOENT** if the bpf_local_storage cannot be found. 4727 * 4728 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4729 * Description 4730 * Return full path for given **struct path** object, which 4731 * needs to be the kernel BTF *path* object. The path is 4732 * returned in the provided buffer *buf* of size *sz* and 4733 * is zero terminated. 4734 * 4735 * Return 4736 * On success, the strictly positive length of the string, 4737 * including the trailing NUL character. On error, a negative 4738 * value. 4739 * 4740 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4741 * Description 4742 * Read *size* bytes from user space address *user_ptr* and store 4743 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4744 * Return 4745 * 0 on success, or a negative error in case of failure. 4746 * 4747 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4748 * Description 4749 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4750 * using *ptr*->type_id. This value should specify the type 4751 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4752 * can be used to look up vmlinux BTF type ids. Traversing the 4753 * data structure using BTF, the type information and values are 4754 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4755 * the pointer data is carried out to avoid kernel crashes during 4756 * operation. Smaller types can use string space on the stack; 4757 * larger programs can use map data to store the string 4758 * representation. 4759 * 4760 * The string can be subsequently shared with userspace via 4761 * bpf_perf_event_output() or ring buffer interfaces. 4762 * bpf_trace_printk() is to be avoided as it places too small 4763 * a limit on string size to be useful. 4764 * 4765 * *flags* is a combination of 4766 * 4767 * **BTF_F_COMPACT** 4768 * no formatting around type information 4769 * **BTF_F_NONAME** 4770 * no struct/union member names/types 4771 * **BTF_F_PTR_RAW** 4772 * show raw (unobfuscated) pointer values; 4773 * equivalent to printk specifier %px. 4774 * **BTF_F_ZERO** 4775 * show zero-valued struct/union members; they 4776 * are not displayed by default 4777 * 4778 * Return 4779 * The number of bytes that were written (or would have been 4780 * written if output had to be truncated due to string size), 4781 * or a negative error in cases of failure. 4782 * 4783 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4784 * Description 4785 * Use BTF to write to seq_write a string representation of 4786 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4787 * *flags* are identical to those used for bpf_snprintf_btf. 4788 * Return 4789 * 0 on success or a negative error in case of failure. 4790 * 4791 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4792 * Description 4793 * See **bpf_get_cgroup_classid**\ () for the main description. 4794 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4795 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4796 * associated socket instead of the current process. 4797 * Return 4798 * The id is returned or 0 in case the id could not be retrieved. 4799 * 4800 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4801 * Description 4802 * Redirect the packet to another net device of index *ifindex* 4803 * and fill in L2 addresses from neighboring subsystem. This helper 4804 * is somewhat similar to **bpf_redirect**\ (), except that it 4805 * populates L2 addresses as well, meaning, internally, the helper 4806 * relies on the neighbor lookup for the L2 address of the nexthop. 4807 * 4808 * The helper will perform a FIB lookup based on the skb's 4809 * networking header to get the address of the next hop, unless 4810 * this is supplied by the caller in the *params* argument. The 4811 * *plen* argument indicates the len of *params* and should be set 4812 * to 0 if *params* is NULL. 4813 * 4814 * The *flags* argument is reserved and must be 0. The helper is 4815 * currently only supported for tc BPF program types, and enabled 4816 * for IPv4 and IPv6 protocols. 4817 * Return 4818 * The helper returns **TC_ACT_REDIRECT** on success or 4819 * **TC_ACT_SHOT** on error. 4820 * 4821 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4822 * Description 4823 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4824 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4825 * extern variable decorated with '__ksym'. For ksym, there is a 4826 * global var (either static or global) defined of the same name 4827 * in the kernel. The ksym is percpu if the global var is percpu. 4828 * The returned pointer points to the global percpu var on *cpu*. 4829 * 4830 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4831 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4832 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4833 * bpf_per_cpu_ptr() must check the returned value. 4834 * Return 4835 * A pointer pointing to the kernel percpu variable on *cpu*, or 4836 * NULL, if *cpu* is invalid. 4837 * 4838 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4839 * Description 4840 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4841 * pointer to the percpu kernel variable on this cpu. See the 4842 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4843 * 4844 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4845 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4846 * never return NULL. 4847 * Return 4848 * A pointer pointing to the kernel percpu variable on this cpu. 4849 * 4850 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4851 * Description 4852 * Redirect the packet to another net device of index *ifindex*. 4853 * This helper is somewhat similar to **bpf_redirect**\ (), except 4854 * that the redirection happens to the *ifindex*' peer device and 4855 * the netns switch takes place from ingress to ingress without 4856 * going through the CPU's backlog queue. 4857 * 4858 * The *flags* argument is reserved and must be 0. The helper is 4859 * currently only supported for tc BPF program types at the ingress 4860 * hook and for veth device types. The peer device must reside in a 4861 * different network namespace. 4862 * Return 4863 * The helper returns **TC_ACT_REDIRECT** on success or 4864 * **TC_ACT_SHOT** on error. 4865 * 4866 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4867 * Description 4868 * Get a bpf_local_storage from the *task*. 4869 * 4870 * Logically, it could be thought of as getting the value from 4871 * a *map* with *task* as the **key**. From this 4872 * perspective, the usage is not much different from 4873 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4874 * helper enforces the key must be a task_struct and the map must also 4875 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4876 * 4877 * Underneath, the value is stored locally at *task* instead of 4878 * the *map*. The *map* is used as the bpf-local-storage 4879 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4880 * searched against all bpf_local_storage residing at *task*. 4881 * 4882 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4883 * used such that a new bpf_local_storage will be 4884 * created if one does not exist. *value* can be used 4885 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4886 * the initial value of a bpf_local_storage. If *value* is 4887 * **NULL**, the new bpf_local_storage will be zero initialized. 4888 * Return 4889 * A bpf_local_storage pointer is returned on success. 4890 * 4891 * **NULL** if not found or there was an error in adding 4892 * a new bpf_local_storage. 4893 * 4894 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4895 * Description 4896 * Delete a bpf_local_storage from a *task*. 4897 * Return 4898 * 0 on success. 4899 * 4900 * **-ENOENT** if the bpf_local_storage cannot be found. 4901 * 4902 * struct task_struct *bpf_get_current_task_btf(void) 4903 * Description 4904 * Return a BTF pointer to the "current" task. 4905 * This pointer can also be used in helpers that accept an 4906 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4907 * Return 4908 * Pointer to the current task. 4909 * 4910 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4911 * Description 4912 * Set or clear certain options on *bprm*: 4913 * 4914 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4915 * which sets the **AT_SECURE** auxv for glibc. The bit 4916 * is cleared if the flag is not specified. 4917 * Return 4918 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4919 * 4920 * u64 bpf_ktime_get_coarse_ns(void) 4921 * Description 4922 * Return a coarse-grained version of the time elapsed since 4923 * system boot, in nanoseconds. Does not include time the system 4924 * was suspended. 4925 * 4926 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4927 * Return 4928 * Current *ktime*. 4929 * 4930 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4931 * Description 4932 * Returns the stored IMA hash of the *inode* (if it's available). 4933 * If the hash is larger than *size*, then only *size* 4934 * bytes will be copied to *dst* 4935 * Return 4936 * The **hash_algo** is returned on success, 4937 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4938 * invalid arguments are passed. 4939 * 4940 * struct socket *bpf_sock_from_file(struct file *file) 4941 * Description 4942 * If the given file represents a socket, returns the associated 4943 * socket. 4944 * Return 4945 * A pointer to a struct socket on success or NULL if the file is 4946 * not a socket. 4947 * 4948 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4949 * Description 4950 * Check packet size against exceeding MTU of net device (based 4951 * on *ifindex*). This helper will likely be used in combination 4952 * with helpers that adjust/change the packet size. 4953 * 4954 * The argument *len_diff* can be used for querying with a planned 4955 * size change. This allows to check MTU prior to changing packet 4956 * ctx. Providing a *len_diff* adjustment that is larger than the 4957 * actual packet size (resulting in negative packet size) will in 4958 * principle not exceed the MTU, which is why it is not considered 4959 * a failure. Other BPF helpers are needed for performing the 4960 * planned size change; therefore the responsibility for catching 4961 * a negative packet size belongs in those helpers. 4962 * 4963 * Specifying *ifindex* zero means the MTU check is performed 4964 * against the current net device. This is practical if this isn't 4965 * used prior to redirect. 4966 * 4967 * On input *mtu_len* must be a valid pointer, else verifier will 4968 * reject BPF program. If the value *mtu_len* is initialized to 4969 * zero then the ctx packet size is use. When value *mtu_len* is 4970 * provided as input this specify the L3 length that the MTU check 4971 * is done against. Remember XDP and TC length operate at L2, but 4972 * this value is L3 as this correlate to MTU and IP-header tot_len 4973 * values which are L3 (similar behavior as bpf_fib_lookup). 4974 * 4975 * The Linux kernel route table can configure MTUs on a more 4976 * specific per route level, which is not provided by this helper. 4977 * For route level MTU checks use the **bpf_fib_lookup**\ () 4978 * helper. 4979 * 4980 * *ctx* is either **struct xdp_md** for XDP programs or 4981 * **struct sk_buff** for tc cls_act programs. 4982 * 4983 * The *flags* argument can be a combination of one or more of the 4984 * following values: 4985 * 4986 * **BPF_MTU_CHK_SEGS** 4987 * This flag will only works for *ctx* **struct sk_buff**. 4988 * If packet context contains extra packet segment buffers 4989 * (often knows as GSO skb), then MTU check is harder to 4990 * check at this point, because in transmit path it is 4991 * possible for the skb packet to get re-segmented 4992 * (depending on net device features). This could still be 4993 * a MTU violation, so this flag enables performing MTU 4994 * check against segments, with a different violation 4995 * return code to tell it apart. Check cannot use len_diff. 4996 * 4997 * On return *mtu_len* pointer contains the MTU value of the net 4998 * device. Remember the net device configured MTU is the L3 size, 4999 * which is returned here and XDP and TC length operate at L2. 5000 * Helper take this into account for you, but remember when using 5001 * MTU value in your BPF-code. 5002 * 5003 * Return 5004 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5005 * 5006 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5007 * 5008 * MTU violations return positive values, but also populate MTU 5009 * value in *mtu_len* pointer, as this can be needed for 5010 * implementing PMTU handing: 5011 * 5012 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5013 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5014 * 5015 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5016 * Description 5017 * For each element in **map**, call **callback_fn** function with 5018 * **map**, **callback_ctx** and other map-specific parameters. 5019 * The **callback_fn** should be a static function and 5020 * the **callback_ctx** should be a pointer to the stack. 5021 * The **flags** is used to control certain aspects of the helper. 5022 * Currently, the **flags** must be 0. 5023 * 5024 * The following are a list of supported map types and their 5025 * respective expected callback signatures: 5026 * 5027 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5028 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5029 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5030 * 5031 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5032 * 5033 * For per_cpu maps, the map_value is the value on the cpu where the 5034 * bpf_prog is running. 5035 * 5036 * If **callback_fn** return 0, the helper will continue to the next 5037 * element. If return value is 1, the helper will skip the rest of 5038 * elements and return. Other return values are not used now. 5039 * 5040 * Return 5041 * The number of traversed map elements for success, **-EINVAL** for 5042 * invalid **flags**. 5043 * 5044 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5045 * Description 5046 * Outputs a string into the **str** buffer of size **str_size** 5047 * based on a format string stored in a read-only map pointed by 5048 * **fmt**. 5049 * 5050 * Each format specifier in **fmt** corresponds to one u64 element 5051 * in the **data** array. For strings and pointers where pointees 5052 * are accessed, only the pointer values are stored in the *data* 5053 * array. The *data_len* is the size of *data* in bytes - must be 5054 * a multiple of 8. 5055 * 5056 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5057 * memory. Reading kernel memory may fail due to either invalid 5058 * address or valid address but requiring a major memory fault. If 5059 * reading kernel memory fails, the string for **%s** will be an 5060 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5061 * Not returning error to bpf program is consistent with what 5062 * **bpf_trace_printk**\ () does for now. 5063 * 5064 * Return 5065 * The strictly positive length of the formatted string, including 5066 * the trailing zero character. If the return value is greater than 5067 * **str_size**, **str** contains a truncated string, guaranteed to 5068 * be zero-terminated except when **str_size** is 0. 5069 * 5070 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5071 * 5072 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5073 * Description 5074 * Execute bpf syscall with given arguments. 5075 * Return 5076 * A syscall result. 5077 * 5078 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5079 * Description 5080 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5081 * Return 5082 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5083 * 5084 * long bpf_sys_close(u32 fd) 5085 * Description 5086 * Execute close syscall for given FD. 5087 * Return 5088 * A syscall result. 5089 * 5090 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5091 * Description 5092 * Initialize the timer. 5093 * First 4 bits of *flags* specify clockid. 5094 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5095 * All other bits of *flags* are reserved. 5096 * The verifier will reject the program if *timer* is not from 5097 * the same *map*. 5098 * Return 5099 * 0 on success. 5100 * **-EBUSY** if *timer* is already initialized. 5101 * **-EINVAL** if invalid *flags* are passed. 5102 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5103 * The user space should either hold a file descriptor to a map with timers 5104 * or pin such map in bpffs. When map is unpinned or file descriptor is 5105 * closed all timers in the map will be cancelled and freed. 5106 * 5107 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5108 * Description 5109 * Configure the timer to call *callback_fn* static function. 5110 * Return 5111 * 0 on success. 5112 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5113 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5114 * The user space should either hold a file descriptor to a map with timers 5115 * or pin such map in bpffs. When map is unpinned or file descriptor is 5116 * closed all timers in the map will be cancelled and freed. 5117 * 5118 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5119 * Description 5120 * Set timer expiration N nanoseconds from the current time. The 5121 * configured callback will be invoked in soft irq context on some cpu 5122 * and will not repeat unless another bpf_timer_start() is made. 5123 * In such case the next invocation can migrate to a different cpu. 5124 * Since struct bpf_timer is a field inside map element the map 5125 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5126 * of BPF program to make sure that callback_fn code stays valid. 5127 * When user space reference to a map reaches zero all timers 5128 * in a map are cancelled and corresponding program's refcnts are 5129 * decremented. This is done to make sure that Ctrl-C of a user 5130 * process doesn't leave any timers running. If map is pinned in 5131 * bpffs the callback_fn can re-arm itself indefinitely. 5132 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5133 * cancel and free the timer in the given map element. 5134 * The map can contain timers that invoke callback_fn-s from different 5135 * programs. The same callback_fn can serve different timers from 5136 * different maps if key/value layout matches across maps. 5137 * Every bpf_timer_set_callback() can have different callback_fn. 5138 * 5139 * *flags* can be one of: 5140 * 5141 * **BPF_F_TIMER_ABS** 5142 * Start the timer in absolute expire value instead of the 5143 * default relative one. 5144 * **BPF_F_TIMER_CPU_PIN** 5145 * Timer will be pinned to the CPU of the caller. 5146 * 5147 * Return 5148 * 0 on success. 5149 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5150 * or invalid *flags* are passed. 5151 * 5152 * long bpf_timer_cancel(struct bpf_timer *timer) 5153 * Description 5154 * Cancel the timer and wait for callback_fn to finish if it was running. 5155 * Return 5156 * 0 if the timer was not active. 5157 * 1 if the timer was active. 5158 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5159 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5160 * own timer which would have led to a deadlock otherwise. 5161 * 5162 * u64 bpf_get_func_ip(void *ctx) 5163 * Description 5164 * Get address of the traced function (for tracing and kprobe programs). 5165 * 5166 * When called for kprobe program attached as uprobe it returns 5167 * probe address for both entry and return uprobe. 5168 * 5169 * Return 5170 * Address of the traced function for kprobe. 5171 * 0 for kprobes placed within the function (not at the entry). 5172 * Address of the probe for uprobe and return uprobe. 5173 * 5174 * u64 bpf_get_attach_cookie(void *ctx) 5175 * Description 5176 * Get bpf_cookie value provided (optionally) during the program 5177 * attachment. It might be different for each individual 5178 * attachment, even if BPF program itself is the same. 5179 * Expects BPF program context *ctx* as a first argument. 5180 * 5181 * Supported for the following program types: 5182 * - kprobe/uprobe; 5183 * - tracepoint; 5184 * - perf_event. 5185 * Return 5186 * Value specified by user at BPF link creation/attachment time 5187 * or 0, if it was not specified. 5188 * 5189 * long bpf_task_pt_regs(struct task_struct *task) 5190 * Description 5191 * Get the struct pt_regs associated with **task**. 5192 * Return 5193 * A pointer to struct pt_regs. 5194 * 5195 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5196 * Description 5197 * Get branch trace from hardware engines like Intel LBR. The 5198 * hardware engine is stopped shortly after the helper is 5199 * called. Therefore, the user need to filter branch entries 5200 * based on the actual use case. To capture branch trace 5201 * before the trigger point of the BPF program, the helper 5202 * should be called at the beginning of the BPF program. 5203 * 5204 * The data is stored as struct perf_branch_entry into output 5205 * buffer *entries*. *size* is the size of *entries* in bytes. 5206 * *flags* is reserved for now and must be zero. 5207 * 5208 * Return 5209 * On success, number of bytes written to *buf*. On error, a 5210 * negative value. 5211 * 5212 * **-EINVAL** if *flags* is not zero. 5213 * 5214 * **-ENOENT** if architecture does not support branch records. 5215 * 5216 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5217 * Description 5218 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5219 * to format and can handle more format args as a result. 5220 * 5221 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5222 * Return 5223 * The number of bytes written to the buffer, or a negative error 5224 * in case of failure. 5225 * 5226 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5227 * Description 5228 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5229 * Return 5230 * *sk* if casting is valid, or **NULL** otherwise. 5231 * 5232 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5233 * Description 5234 * Get the address of a kernel symbol, returned in *res*. *res* is 5235 * set to 0 if the symbol is not found. 5236 * Return 5237 * On success, zero. On error, a negative value. 5238 * 5239 * **-EINVAL** if *flags* is not zero. 5240 * 5241 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5242 * 5243 * **-ENOENT** if symbol is not found. 5244 * 5245 * **-EPERM** if caller does not have permission to obtain kernel address. 5246 * 5247 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5248 * Description 5249 * Find vma of *task* that contains *addr*, call *callback_fn* 5250 * function with *task*, *vma*, and *callback_ctx*. 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. 5255 * 5256 * The expected callback signature is 5257 * 5258 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5259 * 5260 * Return 5261 * 0 on success. 5262 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5263 * **-EBUSY** if failed to try lock mmap_lock. 5264 * **-EINVAL** for invalid **flags**. 5265 * 5266 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5267 * Description 5268 * For **nr_loops**, call **callback_fn** function 5269 * with **callback_ctx** as the context parameter. 5270 * The **callback_fn** should be a static function and 5271 * the **callback_ctx** should be a pointer to the stack. 5272 * The **flags** is used to control certain aspects of the helper. 5273 * Currently, the **flags** must be 0. Currently, nr_loops is 5274 * limited to 1 << 23 (~8 million) loops. 5275 * 5276 * long (\*callback_fn)(u32 index, void \*ctx); 5277 * 5278 * where **index** is the current index in the loop. The index 5279 * is zero-indexed. 5280 * 5281 * If **callback_fn** returns 0, the helper will continue to the next 5282 * loop. If return value is 1, the helper will skip the rest of 5283 * the loops and return. Other return values are not used now, 5284 * and will be rejected by the verifier. 5285 * 5286 * Return 5287 * The number of loops performed, **-EINVAL** for invalid **flags**, 5288 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5289 * 5290 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5291 * Description 5292 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5293 * to be null-terminated and **s1_sz** is the maximum storage 5294 * size of **s1**. **s2** must be a read-only string. 5295 * Return 5296 * An integer less than, equal to, or greater than zero 5297 * if the first **s1_sz** bytes of **s1** is found to be 5298 * less than, to match, or be greater than **s2**. 5299 * 5300 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5301 * Description 5302 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5303 * returned in **value**. 5304 * 5305 * Return 5306 * 0 on success. 5307 * **-EINVAL** if n >= argument register count of traced function. 5308 * 5309 * long bpf_get_func_ret(void *ctx, u64 *value) 5310 * Description 5311 * Get return value of the traced function (for tracing programs) 5312 * in **value**. 5313 * 5314 * Return 5315 * 0 on success. 5316 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5317 * 5318 * long bpf_get_func_arg_cnt(void *ctx) 5319 * Description 5320 * Get number of registers of the traced function (for tracing programs) where 5321 * function arguments are stored in these registers. 5322 * 5323 * Return 5324 * The number of argument registers of the traced function. 5325 * 5326 * int bpf_get_retval(void) 5327 * Description 5328 * Get the BPF program's return value that will be returned to the upper layers. 5329 * 5330 * This helper is currently supported by cgroup programs and only by the hooks 5331 * where BPF program's return value is returned to the userspace via errno. 5332 * Return 5333 * The BPF program's return value. 5334 * 5335 * int bpf_set_retval(int retval) 5336 * Description 5337 * Set the BPF program's return value that will be returned to the upper layers. 5338 * 5339 * This helper is currently supported by cgroup programs and only by the hooks 5340 * where BPF program's return value is returned to the userspace via errno. 5341 * 5342 * Note that there is the following corner case where the program exports an error 5343 * via bpf_set_retval but signals success via 'return 1': 5344 * 5345 * bpf_set_retval(-EPERM); 5346 * return 1; 5347 * 5348 * In this case, the BPF program's return value will use helper's -EPERM. This 5349 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5350 * 5351 * Return 5352 * 0 on success, or a negative error in case of failure. 5353 * 5354 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5355 * Description 5356 * Get the total size of a given xdp buff (linear and paged area) 5357 * Return 5358 * The total size of a given xdp buffer. 5359 * 5360 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5361 * Description 5362 * This helper is provided as an easy way to load data from a 5363 * xdp buffer. It can be used to load *len* bytes from *offset* from 5364 * the frame associated to *xdp_md*, into the buffer pointed by 5365 * *buf*. 5366 * Return 5367 * 0 on success, or a negative error in case of failure. 5368 * 5369 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5370 * Description 5371 * Store *len* bytes from buffer *buf* into the frame 5372 * associated to *xdp_md*, at *offset*. 5373 * Return 5374 * 0 on success, or a negative error in case of failure. 5375 * 5376 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5377 * Description 5378 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5379 * address space, and stores the data in *dst*. *flags* is not 5380 * used yet and is provided for future extensibility. This helper 5381 * can only be used by sleepable programs. 5382 * Return 5383 * 0 on success, or a negative error in case of failure. On error 5384 * *dst* buffer is zeroed out. 5385 * 5386 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5387 * Description 5388 * Change the __sk_buff->tstamp_type to *tstamp_type* 5389 * and set *tstamp* to the __sk_buff->tstamp together. 5390 * 5391 * If there is no need to change the __sk_buff->tstamp_type, 5392 * the tstamp value can be directly written to __sk_buff->tstamp 5393 * instead. 5394 * 5395 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5396 * will be kept during bpf_redirect_*(). A non zero 5397 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5398 * *tstamp_type*. 5399 * 5400 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5401 * with a zero *tstamp*. 5402 * 5403 * Only IPv4 and IPv6 skb->protocol are supported. 5404 * 5405 * This function is most useful when it needs to set a 5406 * mono delivery time to __sk_buff->tstamp and then 5407 * bpf_redirect_*() to the egress of an iface. For example, 5408 * changing the (rcv) timestamp in __sk_buff->tstamp at 5409 * ingress to a mono delivery time and then bpf_redirect_*() 5410 * to sch_fq@phy-dev. 5411 * Return 5412 * 0 on success. 5413 * **-EINVAL** for invalid input 5414 * **-EOPNOTSUPP** for unsupported protocol 5415 * 5416 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5417 * Description 5418 * Returns a calculated IMA hash of the *file*. 5419 * If the hash is larger than *size*, then only *size* 5420 * bytes will be copied to *dst* 5421 * Return 5422 * The **hash_algo** is returned on success, 5423 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5424 * invalid arguments are passed. 5425 * 5426 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5427 * Description 5428 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5429 * old value. *ptr* can be NULL, otherwise it must be a referenced 5430 * pointer which will be released when this helper is called. 5431 * Return 5432 * The old value of kptr (which can be NULL). The returned pointer 5433 * if not NULL, is a reference which must be released using its 5434 * corresponding release function, or moved into a BPF map before 5435 * program exit. 5436 * 5437 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5438 * Description 5439 * Perform a lookup in *percpu map* for an entry associated to 5440 * *key* on *cpu*. 5441 * Return 5442 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5443 * was found or *cpu* is invalid. 5444 * 5445 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5446 * Description 5447 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5448 * Return 5449 * *sk* if casting is valid, or **NULL** otherwise. 5450 * 5451 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5452 * Description 5453 * Get a dynptr to local memory *data*. 5454 * 5455 * *data* must be a ptr to a map value. 5456 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5457 * *flags* is currently unused. 5458 * Return 5459 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5460 * -EINVAL if flags is not 0. 5461 * 5462 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5463 * Description 5464 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5465 * through the dynptr interface. *flags* must be 0. 5466 * 5467 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5468 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5469 * reservation fails. This is enforced by the verifier. 5470 * Return 5471 * 0 on success, or a negative error in case of failure. 5472 * 5473 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5474 * Description 5475 * Submit reserved ring buffer sample, pointed to by *data*, 5476 * through the dynptr interface. This is a no-op if the dynptr is 5477 * invalid/null. 5478 * 5479 * For more information on *flags*, please see 5480 * 'bpf_ringbuf_submit'. 5481 * Return 5482 * Nothing. Always succeeds. 5483 * 5484 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5485 * Description 5486 * Discard reserved ring buffer sample through the dynptr 5487 * interface. This is a no-op if the dynptr is invalid/null. 5488 * 5489 * For more information on *flags*, please see 5490 * 'bpf_ringbuf_discard'. 5491 * Return 5492 * Nothing. Always succeeds. 5493 * 5494 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5495 * Description 5496 * Read *len* bytes from *src* into *dst*, starting from *offset* 5497 * into *src*. 5498 * *flags* is currently unused. 5499 * Return 5500 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5501 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5502 * *flags* is not 0. 5503 * 5504 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5505 * Description 5506 * Write *len* bytes from *src* into *dst*, starting from *offset* 5507 * into *dst*. 5508 * 5509 * *flags* must be 0 except for skb-type dynptrs. 5510 * 5511 * For skb-type dynptrs: 5512 * * All data slices of the dynptr are automatically 5513 * invalidated after **bpf_dynptr_write**\ (). This is 5514 * because writing may pull the skb and change the 5515 * underlying packet buffer. 5516 * 5517 * * For *flags*, please see the flags accepted by 5518 * **bpf_skb_store_bytes**\ (). 5519 * Return 5520 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5521 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5522 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5523 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5524 * 5525 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5526 * Description 5527 * Get a pointer to the underlying dynptr data. 5528 * 5529 * *len* must be a statically known value. The returned data slice 5530 * is invalidated whenever the dynptr is invalidated. 5531 * 5532 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5533 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5534 * Return 5535 * Pointer to the underlying dynptr data, NULL if the dynptr is 5536 * read-only, if the dynptr is invalid, or if the offset and length 5537 * is out of bounds. 5538 * 5539 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5540 * Description 5541 * Try to issue a SYN cookie for the packet with corresponding 5542 * IPv4/TCP headers, *iph* and *th*, without depending on a 5543 * listening socket. 5544 * 5545 * *iph* points to the IPv4 header. 5546 * 5547 * *th* points to the start of the TCP header, while *th_len* 5548 * contains the length of the TCP header (at least 5549 * **sizeof**\ (**struct tcphdr**)). 5550 * Return 5551 * On success, lower 32 bits hold the generated SYN cookie in 5552 * followed by 16 bits which hold the MSS value for that cookie, 5553 * and the top 16 bits are unused. 5554 * 5555 * On failure, the returned value is one of the following: 5556 * 5557 * **-EINVAL** if *th_len* is invalid. 5558 * 5559 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5560 * Description 5561 * Try to issue a SYN cookie for the packet with corresponding 5562 * IPv6/TCP headers, *iph* and *th*, without depending on a 5563 * listening socket. 5564 * 5565 * *iph* points to the IPv6 header. 5566 * 5567 * *th* points to the start of the TCP header, while *th_len* 5568 * contains the length of the TCP header (at least 5569 * **sizeof**\ (**struct tcphdr**)). 5570 * Return 5571 * On success, lower 32 bits hold the generated SYN cookie in 5572 * followed by 16 bits which hold the MSS value for that cookie, 5573 * and the top 16 bits are unused. 5574 * 5575 * On failure, the returned value is one of the following: 5576 * 5577 * **-EINVAL** if *th_len* is invalid. 5578 * 5579 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5580 * 5581 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5582 * Description 5583 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5584 * without depending on a listening socket. 5585 * 5586 * *iph* points to the IPv4 header. 5587 * 5588 * *th* points to the TCP header. 5589 * Return 5590 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5591 * 5592 * On failure, the returned value is one of the following: 5593 * 5594 * **-EACCES** if the SYN cookie is not valid. 5595 * 5596 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5597 * Description 5598 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5599 * without depending on a listening socket. 5600 * 5601 * *iph* points to the IPv6 header. 5602 * 5603 * *th* points to the TCP header. 5604 * Return 5605 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5606 * 5607 * On failure, the returned value is one of the following: 5608 * 5609 * **-EACCES** if the SYN cookie is not valid. 5610 * 5611 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5612 * 5613 * u64 bpf_ktime_get_tai_ns(void) 5614 * Description 5615 * A nonsettable system-wide clock derived from wall-clock time but 5616 * ignoring leap seconds. This clock does not experience 5617 * discontinuities and backwards jumps caused by NTP inserting leap 5618 * seconds as CLOCK_REALTIME does. 5619 * 5620 * See: **clock_gettime**\ (**CLOCK_TAI**) 5621 * Return 5622 * Current *ktime*. 5623 * 5624 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5625 * Description 5626 * Drain samples from the specified user ring buffer, and invoke 5627 * the provided callback for each such sample: 5628 * 5629 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5630 * 5631 * If **callback_fn** returns 0, the helper will continue to try 5632 * and drain the next sample, up to a maximum of 5633 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5634 * the helper will skip the rest of the samples and return. Other 5635 * return values are not used now, and will be rejected by the 5636 * verifier. 5637 * Return 5638 * The number of drained samples if no error was encountered while 5639 * draining samples, or 0 if no samples were present in the ring 5640 * buffer. If a user-space producer was epoll-waiting on this map, 5641 * and at least one sample was drained, they will receive an event 5642 * notification notifying them of available space in the ring 5643 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5644 * function, no wakeup notification will be sent. If the 5645 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5646 * be sent even if no sample was drained. 5647 * 5648 * On failure, the returned value is one of the following: 5649 * 5650 * **-EBUSY** if the ring buffer is contended, and another calling 5651 * context was concurrently draining the ring buffer. 5652 * 5653 * **-EINVAL** if user-space is not properly tracking the ring 5654 * buffer due to the producer position not being aligned to 8 5655 * bytes, a sample not being aligned to 8 bytes, or the producer 5656 * position not matching the advertised length of a sample. 5657 * 5658 * **-E2BIG** if user-space has tried to publish a sample which is 5659 * larger than the size of the ring buffer, or which cannot fit 5660 * within a struct bpf_dynptr. 5661 * 5662 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5663 * Description 5664 * Get a bpf_local_storage from the *cgroup*. 5665 * 5666 * Logically, it could be thought of as getting the value from 5667 * a *map* with *cgroup* as the **key**. From this 5668 * perspective, the usage is not much different from 5669 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5670 * helper enforces the key must be a cgroup struct and the map must also 5671 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5672 * 5673 * In reality, the local-storage value is embedded directly inside of the 5674 * *cgroup* object itself, rather than being located in the 5675 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5676 * queried for some *map* on a *cgroup* object, the kernel will perform an 5677 * O(n) iteration over all of the live local-storage values for that 5678 * *cgroup* object until the local-storage value for the *map* is found. 5679 * 5680 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5681 * used such that a new bpf_local_storage will be 5682 * created if one does not exist. *value* can be used 5683 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5684 * the initial value of a bpf_local_storage. If *value* is 5685 * **NULL**, the new bpf_local_storage will be zero initialized. 5686 * Return 5687 * A bpf_local_storage pointer is returned on success. 5688 * 5689 * **NULL** if not found or there was an error in adding 5690 * a new bpf_local_storage. 5691 * 5692 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5693 * Description 5694 * Delete a bpf_local_storage from a *cgroup*. 5695 * Return 5696 * 0 on success. 5697 * 5698 * **-ENOENT** if the bpf_local_storage cannot be found. 5699 */ 5700 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5701 FN(unspec, 0, ##ctx) \ 5702 FN(map_lookup_elem, 1, ##ctx) \ 5703 FN(map_update_elem, 2, ##ctx) \ 5704 FN(map_delete_elem, 3, ##ctx) \ 5705 FN(probe_read, 4, ##ctx) \ 5706 FN(ktime_get_ns, 5, ##ctx) \ 5707 FN(trace_printk, 6, ##ctx) \ 5708 FN(get_prandom_u32, 7, ##ctx) \ 5709 FN(get_smp_processor_id, 8, ##ctx) \ 5710 FN(skb_store_bytes, 9, ##ctx) \ 5711 FN(l3_csum_replace, 10, ##ctx) \ 5712 FN(l4_csum_replace, 11, ##ctx) \ 5713 FN(tail_call, 12, ##ctx) \ 5714 FN(clone_redirect, 13, ##ctx) \ 5715 FN(get_current_pid_tgid, 14, ##ctx) \ 5716 FN(get_current_uid_gid, 15, ##ctx) \ 5717 FN(get_current_comm, 16, ##ctx) \ 5718 FN(get_cgroup_classid, 17, ##ctx) \ 5719 FN(skb_vlan_push, 18, ##ctx) \ 5720 FN(skb_vlan_pop, 19, ##ctx) \ 5721 FN(skb_get_tunnel_key, 20, ##ctx) \ 5722 FN(skb_set_tunnel_key, 21, ##ctx) \ 5723 FN(perf_event_read, 22, ##ctx) \ 5724 FN(redirect, 23, ##ctx) \ 5725 FN(get_route_realm, 24, ##ctx) \ 5726 FN(perf_event_output, 25, ##ctx) \ 5727 FN(skb_load_bytes, 26, ##ctx) \ 5728 FN(get_stackid, 27, ##ctx) \ 5729 FN(csum_diff, 28, ##ctx) \ 5730 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5731 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5732 FN(skb_change_proto, 31, ##ctx) \ 5733 FN(skb_change_type, 32, ##ctx) \ 5734 FN(skb_under_cgroup, 33, ##ctx) \ 5735 FN(get_hash_recalc, 34, ##ctx) \ 5736 FN(get_current_task, 35, ##ctx) \ 5737 FN(probe_write_user, 36, ##ctx) \ 5738 FN(current_task_under_cgroup, 37, ##ctx) \ 5739 FN(skb_change_tail, 38, ##ctx) \ 5740 FN(skb_pull_data, 39, ##ctx) \ 5741 FN(csum_update, 40, ##ctx) \ 5742 FN(set_hash_invalid, 41, ##ctx) \ 5743 FN(get_numa_node_id, 42, ##ctx) \ 5744 FN(skb_change_head, 43, ##ctx) \ 5745 FN(xdp_adjust_head, 44, ##ctx) \ 5746 FN(probe_read_str, 45, ##ctx) \ 5747 FN(get_socket_cookie, 46, ##ctx) \ 5748 FN(get_socket_uid, 47, ##ctx) \ 5749 FN(set_hash, 48, ##ctx) \ 5750 FN(setsockopt, 49, ##ctx) \ 5751 FN(skb_adjust_room, 50, ##ctx) \ 5752 FN(redirect_map, 51, ##ctx) \ 5753 FN(sk_redirect_map, 52, ##ctx) \ 5754 FN(sock_map_update, 53, ##ctx) \ 5755 FN(xdp_adjust_meta, 54, ##ctx) \ 5756 FN(perf_event_read_value, 55, ##ctx) \ 5757 FN(perf_prog_read_value, 56, ##ctx) \ 5758 FN(getsockopt, 57, ##ctx) \ 5759 FN(override_return, 58, ##ctx) \ 5760 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5761 FN(msg_redirect_map, 60, ##ctx) \ 5762 FN(msg_apply_bytes, 61, ##ctx) \ 5763 FN(msg_cork_bytes, 62, ##ctx) \ 5764 FN(msg_pull_data, 63, ##ctx) \ 5765 FN(bind, 64, ##ctx) \ 5766 FN(xdp_adjust_tail, 65, ##ctx) \ 5767 FN(skb_get_xfrm_state, 66, ##ctx) \ 5768 FN(get_stack, 67, ##ctx) \ 5769 FN(skb_load_bytes_relative, 68, ##ctx) \ 5770 FN(fib_lookup, 69, ##ctx) \ 5771 FN(sock_hash_update, 70, ##ctx) \ 5772 FN(msg_redirect_hash, 71, ##ctx) \ 5773 FN(sk_redirect_hash, 72, ##ctx) \ 5774 FN(lwt_push_encap, 73, ##ctx) \ 5775 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5776 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5777 FN(lwt_seg6_action, 76, ##ctx) \ 5778 FN(rc_repeat, 77, ##ctx) \ 5779 FN(rc_keydown, 78, ##ctx) \ 5780 FN(skb_cgroup_id, 79, ##ctx) \ 5781 FN(get_current_cgroup_id, 80, ##ctx) \ 5782 FN(get_local_storage, 81, ##ctx) \ 5783 FN(sk_select_reuseport, 82, ##ctx) \ 5784 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5785 FN(sk_lookup_tcp, 84, ##ctx) \ 5786 FN(sk_lookup_udp, 85, ##ctx) \ 5787 FN(sk_release, 86, ##ctx) \ 5788 FN(map_push_elem, 87, ##ctx) \ 5789 FN(map_pop_elem, 88, ##ctx) \ 5790 FN(map_peek_elem, 89, ##ctx) \ 5791 FN(msg_push_data, 90, ##ctx) \ 5792 FN(msg_pop_data, 91, ##ctx) \ 5793 FN(rc_pointer_rel, 92, ##ctx) \ 5794 FN(spin_lock, 93, ##ctx) \ 5795 FN(spin_unlock, 94, ##ctx) \ 5796 FN(sk_fullsock, 95, ##ctx) \ 5797 FN(tcp_sock, 96, ##ctx) \ 5798 FN(skb_ecn_set_ce, 97, ##ctx) \ 5799 FN(get_listener_sock, 98, ##ctx) \ 5800 FN(skc_lookup_tcp, 99, ##ctx) \ 5801 FN(tcp_check_syncookie, 100, ##ctx) \ 5802 FN(sysctl_get_name, 101, ##ctx) \ 5803 FN(sysctl_get_current_value, 102, ##ctx) \ 5804 FN(sysctl_get_new_value, 103, ##ctx) \ 5805 FN(sysctl_set_new_value, 104, ##ctx) \ 5806 FN(strtol, 105, ##ctx) \ 5807 FN(strtoul, 106, ##ctx) \ 5808 FN(sk_storage_get, 107, ##ctx) \ 5809 FN(sk_storage_delete, 108, ##ctx) \ 5810 FN(send_signal, 109, ##ctx) \ 5811 FN(tcp_gen_syncookie, 110, ##ctx) \ 5812 FN(skb_output, 111, ##ctx) \ 5813 FN(probe_read_user, 112, ##ctx) \ 5814 FN(probe_read_kernel, 113, ##ctx) \ 5815 FN(probe_read_user_str, 114, ##ctx) \ 5816 FN(probe_read_kernel_str, 115, ##ctx) \ 5817 FN(tcp_send_ack, 116, ##ctx) \ 5818 FN(send_signal_thread, 117, ##ctx) \ 5819 FN(jiffies64, 118, ##ctx) \ 5820 FN(read_branch_records, 119, ##ctx) \ 5821 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5822 FN(xdp_output, 121, ##ctx) \ 5823 FN(get_netns_cookie, 122, ##ctx) \ 5824 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5825 FN(sk_assign, 124, ##ctx) \ 5826 FN(ktime_get_boot_ns, 125, ##ctx) \ 5827 FN(seq_printf, 126, ##ctx) \ 5828 FN(seq_write, 127, ##ctx) \ 5829 FN(sk_cgroup_id, 128, ##ctx) \ 5830 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5831 FN(ringbuf_output, 130, ##ctx) \ 5832 FN(ringbuf_reserve, 131, ##ctx) \ 5833 FN(ringbuf_submit, 132, ##ctx) \ 5834 FN(ringbuf_discard, 133, ##ctx) \ 5835 FN(ringbuf_query, 134, ##ctx) \ 5836 FN(csum_level, 135, ##ctx) \ 5837 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5838 FN(skc_to_tcp_sock, 137, ##ctx) \ 5839 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5840 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5841 FN(skc_to_udp6_sock, 140, ##ctx) \ 5842 FN(get_task_stack, 141, ##ctx) \ 5843 FN(load_hdr_opt, 142, ##ctx) \ 5844 FN(store_hdr_opt, 143, ##ctx) \ 5845 FN(reserve_hdr_opt, 144, ##ctx) \ 5846 FN(inode_storage_get, 145, ##ctx) \ 5847 FN(inode_storage_delete, 146, ##ctx) \ 5848 FN(d_path, 147, ##ctx) \ 5849 FN(copy_from_user, 148, ##ctx) \ 5850 FN(snprintf_btf, 149, ##ctx) \ 5851 FN(seq_printf_btf, 150, ##ctx) \ 5852 FN(skb_cgroup_classid, 151, ##ctx) \ 5853 FN(redirect_neigh, 152, ##ctx) \ 5854 FN(per_cpu_ptr, 153, ##ctx) \ 5855 FN(this_cpu_ptr, 154, ##ctx) \ 5856 FN(redirect_peer, 155, ##ctx) \ 5857 FN(task_storage_get, 156, ##ctx) \ 5858 FN(task_storage_delete, 157, ##ctx) \ 5859 FN(get_current_task_btf, 158, ##ctx) \ 5860 FN(bprm_opts_set, 159, ##ctx) \ 5861 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5862 FN(ima_inode_hash, 161, ##ctx) \ 5863 FN(sock_from_file, 162, ##ctx) \ 5864 FN(check_mtu, 163, ##ctx) \ 5865 FN(for_each_map_elem, 164, ##ctx) \ 5866 FN(snprintf, 165, ##ctx) \ 5867 FN(sys_bpf, 166, ##ctx) \ 5868 FN(btf_find_by_name_kind, 167, ##ctx) \ 5869 FN(sys_close, 168, ##ctx) \ 5870 FN(timer_init, 169, ##ctx) \ 5871 FN(timer_set_callback, 170, ##ctx) \ 5872 FN(timer_start, 171, ##ctx) \ 5873 FN(timer_cancel, 172, ##ctx) \ 5874 FN(get_func_ip, 173, ##ctx) \ 5875 FN(get_attach_cookie, 174, ##ctx) \ 5876 FN(task_pt_regs, 175, ##ctx) \ 5877 FN(get_branch_snapshot, 176, ##ctx) \ 5878 FN(trace_vprintk, 177, ##ctx) \ 5879 FN(skc_to_unix_sock, 178, ##ctx) \ 5880 FN(kallsyms_lookup_name, 179, ##ctx) \ 5881 FN(find_vma, 180, ##ctx) \ 5882 FN(loop, 181, ##ctx) \ 5883 FN(strncmp, 182, ##ctx) \ 5884 FN(get_func_arg, 183, ##ctx) \ 5885 FN(get_func_ret, 184, ##ctx) \ 5886 FN(get_func_arg_cnt, 185, ##ctx) \ 5887 FN(get_retval, 186, ##ctx) \ 5888 FN(set_retval, 187, ##ctx) \ 5889 FN(xdp_get_buff_len, 188, ##ctx) \ 5890 FN(xdp_load_bytes, 189, ##ctx) \ 5891 FN(xdp_store_bytes, 190, ##ctx) \ 5892 FN(copy_from_user_task, 191, ##ctx) \ 5893 FN(skb_set_tstamp, 192, ##ctx) \ 5894 FN(ima_file_hash, 193, ##ctx) \ 5895 FN(kptr_xchg, 194, ##ctx) \ 5896 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5897 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5898 FN(dynptr_from_mem, 197, ##ctx) \ 5899 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5900 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5901 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5902 FN(dynptr_read, 201, ##ctx) \ 5903 FN(dynptr_write, 202, ##ctx) \ 5904 FN(dynptr_data, 203, ##ctx) \ 5905 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5906 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5907 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5908 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5909 FN(ktime_get_tai_ns, 208, ##ctx) \ 5910 FN(user_ringbuf_drain, 209, ##ctx) \ 5911 FN(cgrp_storage_get, 210, ##ctx) \ 5912 FN(cgrp_storage_delete, 211, ##ctx) \ 5913 /* */ 5914 5915 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 5916 * know or care about integer value that is now passed as second argument 5917 */ 5918 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 5919 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 5920 5921 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5922 * function eBPF program intends to call 5923 */ 5924 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 5925 enum bpf_func_id { 5926 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5927 __BPF_FUNC_MAX_ID, 5928 }; 5929 #undef __BPF_ENUM_FN 5930 5931 /* All flags used by eBPF helper functions, placed here. */ 5932 5933 /* BPF_FUNC_skb_store_bytes flags. */ 5934 enum { 5935 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5936 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5937 }; 5938 5939 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5940 * First 4 bits are for passing the header field size. 5941 */ 5942 enum { 5943 BPF_F_HDR_FIELD_MASK = 0xfULL, 5944 }; 5945 5946 /* BPF_FUNC_l4_csum_replace flags. */ 5947 enum { 5948 BPF_F_PSEUDO_HDR = (1ULL << 4), 5949 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5950 BPF_F_MARK_ENFORCE = (1ULL << 6), 5951 }; 5952 5953 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5954 enum { 5955 BPF_F_INGRESS = (1ULL << 0), 5956 }; 5957 5958 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5959 enum { 5960 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5961 }; 5962 5963 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5964 enum { 5965 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5966 BPF_F_USER_STACK = (1ULL << 8), 5967 /* flags used by BPF_FUNC_get_stackid only. */ 5968 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5969 BPF_F_REUSE_STACKID = (1ULL << 10), 5970 /* flags used by BPF_FUNC_get_stack only. */ 5971 BPF_F_USER_BUILD_ID = (1ULL << 11), 5972 }; 5973 5974 /* BPF_FUNC_skb_set_tunnel_key flags. */ 5975 enum { 5976 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5977 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5978 BPF_F_SEQ_NUMBER = (1ULL << 3), 5979 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 5980 }; 5981 5982 /* BPF_FUNC_skb_get_tunnel_key flags. */ 5983 enum { 5984 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 5985 }; 5986 5987 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5988 * BPF_FUNC_perf_event_read_value flags. 5989 */ 5990 enum { 5991 BPF_F_INDEX_MASK = 0xffffffffULL, 5992 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5993 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 5994 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5995 }; 5996 5997 /* Current network namespace */ 5998 enum { 5999 BPF_F_CURRENT_NETNS = (-1L), 6000 }; 6001 6002 /* BPF_FUNC_csum_level level values. */ 6003 enum { 6004 BPF_CSUM_LEVEL_QUERY, 6005 BPF_CSUM_LEVEL_INC, 6006 BPF_CSUM_LEVEL_DEC, 6007 BPF_CSUM_LEVEL_RESET, 6008 }; 6009 6010 /* BPF_FUNC_skb_adjust_room flags. */ 6011 enum { 6012 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6013 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6014 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6015 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6016 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6017 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6018 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6019 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6020 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6021 }; 6022 6023 enum { 6024 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6025 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6026 }; 6027 6028 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6029 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6030 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6031 6032 /* BPF_FUNC_sysctl_get_name flags. */ 6033 enum { 6034 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6035 }; 6036 6037 /* BPF_FUNC_
_storage_get flags */ 6038 enum { 6039 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6040 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6041 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6042 */ 6043 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6044 }; 6045 6046 /* BPF_FUNC_read_branch_records flags. */ 6047 enum { 6048 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6049 }; 6050 6051 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6052 * BPF_FUNC_bpf_ringbuf_output flags. 6053 */ 6054 enum { 6055 BPF_RB_NO_WAKEUP = (1ULL << 0), 6056 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6057 }; 6058 6059 /* BPF_FUNC_bpf_ringbuf_query flags */ 6060 enum { 6061 BPF_RB_AVAIL_DATA = 0, 6062 BPF_RB_RING_SIZE = 1, 6063 BPF_RB_CONS_POS = 2, 6064 BPF_RB_PROD_POS = 3, 6065 }; 6066 6067 /* BPF ring buffer constants */ 6068 enum { 6069 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6070 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6071 BPF_RINGBUF_HDR_SZ = 8, 6072 }; 6073 6074 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6075 enum { 6076 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6077 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6078 }; 6079 6080 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6081 enum bpf_adj_room_mode { 6082 BPF_ADJ_ROOM_NET, 6083 BPF_ADJ_ROOM_MAC, 6084 }; 6085 6086 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6087 enum bpf_hdr_start_off { 6088 BPF_HDR_START_MAC, 6089 BPF_HDR_START_NET, 6090 }; 6091 6092 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6093 enum bpf_lwt_encap_mode { 6094 BPF_LWT_ENCAP_SEG6, 6095 BPF_LWT_ENCAP_SEG6_INLINE, 6096 BPF_LWT_ENCAP_IP, 6097 }; 6098 6099 /* Flags for bpf_bprm_opts_set helper */ 6100 enum { 6101 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6102 }; 6103 6104 /* Flags for bpf_redirect_map helper */ 6105 enum { 6106 BPF_F_BROADCAST = (1ULL << 3), 6107 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 6108 }; 6109 6110 #define __bpf_md_ptr(type, name) \ 6111 union { \ 6112 type name; \ 6113 __u64 :64; \ 6114 } __attribute__((aligned(8))) 6115 6116 enum { 6117 BPF_SKB_TSTAMP_UNSPEC, 6118 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 6119 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 6120 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 6121 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 6122 */ 6123 }; 6124 6125 /* user accessible mirror of in-kernel sk_buff. 6126 * new fields can only be added to the end of this structure 6127 */ 6128 struct __sk_buff { 6129 __u32 len; 6130 __u32 pkt_type; 6131 __u32 mark; 6132 __u32 queue_mapping; 6133 __u32 protocol; 6134 __u32 vlan_present; 6135 __u32 vlan_tci; 6136 __u32 vlan_proto; 6137 __u32 priority; 6138 __u32 ingress_ifindex; 6139 __u32 ifindex; 6140 __u32 tc_index; 6141 __u32 cb[5]; 6142 __u32 hash; 6143 __u32 tc_classid; 6144 __u32 data; 6145 __u32 data_end; 6146 __u32 napi_id; 6147 6148 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6149 __u32 family; 6150 __u32 remote_ip4; /* Stored in network byte order */ 6151 __u32 local_ip4; /* Stored in network byte order */ 6152 __u32 remote_ip6[4]; /* Stored in network byte order */ 6153 __u32 local_ip6[4]; /* Stored in network byte order */ 6154 __u32 remote_port; /* Stored in network byte order */ 6155 __u32 local_port; /* stored in host byte order */ 6156 /* ... here. */ 6157 6158 __u32 data_meta; 6159 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6160 __u64 tstamp; 6161 __u32 wire_len; 6162 __u32 gso_segs; 6163 __bpf_md_ptr(struct bpf_sock *, sk); 6164 __u32 gso_size; 6165 __u8 tstamp_type; 6166 __u32 :24; /* Padding, future use. */ 6167 __u64 hwtstamp; 6168 }; 6169 6170 struct bpf_tunnel_key { 6171 __u32 tunnel_id; 6172 union { 6173 __u32 remote_ipv4; 6174 __u32 remote_ipv6[4]; 6175 }; 6176 __u8 tunnel_tos; 6177 __u8 tunnel_ttl; 6178 union { 6179 __u16 tunnel_ext; /* compat */ 6180 __be16 tunnel_flags; 6181 }; 6182 __u32 tunnel_label; 6183 union { 6184 __u32 local_ipv4; 6185 __u32 local_ipv6[4]; 6186 }; 6187 }; 6188 6189 /* user accessible mirror of in-kernel xfrm_state. 6190 * new fields can only be added to the end of this structure 6191 */ 6192 struct bpf_xfrm_state { 6193 __u32 reqid; 6194 __u32 spi; /* Stored in network byte order */ 6195 __u16 family; 6196 __u16 ext; /* Padding, future use. */ 6197 union { 6198 __u32 remote_ipv4; /* Stored in network byte order */ 6199 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6200 }; 6201 }; 6202 6203 /* Generic BPF return codes which all BPF program types may support. 6204 * The values are binary compatible with their TC_ACT_* counter-part to 6205 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6206 * programs. 6207 * 6208 * XDP is handled seprately, see XDP_*. 6209 */ 6210 enum bpf_ret_code { 6211 BPF_OK = 0, 6212 /* 1 reserved */ 6213 BPF_DROP = 2, 6214 /* 3-6 reserved */ 6215 BPF_REDIRECT = 7, 6216 /* >127 are reserved for prog type specific return codes. 6217 * 6218 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6219 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6220 * changed and should be routed based on its new L3 header. 6221 * (This is an L3 redirect, as opposed to L2 redirect 6222 * represented by BPF_REDIRECT above). 6223 */ 6224 BPF_LWT_REROUTE = 128, 6225 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6226 * to indicate that no custom dissection was performed, and 6227 * fallback to standard dissector is requested. 6228 */ 6229 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6230 }; 6231 6232 struct bpf_sock { 6233 __u32 bound_dev_if; 6234 __u32 family; 6235 __u32 type; 6236 __u32 protocol; 6237 __u32 mark; 6238 __u32 priority; 6239 /* IP address also allows 1 and 2 bytes access */ 6240 __u32 src_ip4; 6241 __u32 src_ip6[4]; 6242 __u32 src_port; /* host byte order */ 6243 __be16 dst_port; /* network byte order */ 6244 __u16 :16; /* zero padding */ 6245 __u32 dst_ip4; 6246 __u32 dst_ip6[4]; 6247 __u32 state; 6248 __s32 rx_queue_mapping; 6249 }; 6250 6251 struct bpf_tcp_sock { 6252 __u32 snd_cwnd; /* Sending congestion window */ 6253 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6254 __u32 rtt_min; 6255 __u32 snd_ssthresh; /* Slow start size threshold */ 6256 __u32 rcv_nxt; /* What we want to receive next */ 6257 __u32 snd_nxt; /* Next sequence we send */ 6258 __u32 snd_una; /* First byte we want an ack for */ 6259 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6260 __u32 ecn_flags; /* ECN status bits. */ 6261 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6262 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6263 __u32 packets_out; /* Packets which are "in flight" */ 6264 __u32 retrans_out; /* Retransmitted packets out */ 6265 __u32 total_retrans; /* Total retransmits for entire connection */ 6266 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6267 * total number of segments in. 6268 */ 6269 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6270 * total number of data segments in. 6271 */ 6272 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6273 * The total number of segments sent. 6274 */ 6275 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6276 * total number of data segments sent. 6277 */ 6278 __u32 lost_out; /* Lost packets */ 6279 __u32 sacked_out; /* SACK'd packets */ 6280 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6281 * sum(delta(rcv_nxt)), or how many bytes 6282 * were acked. 6283 */ 6284 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6285 * sum(delta(snd_una)), or how many bytes 6286 * were acked. 6287 */ 6288 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6289 * total number of DSACK blocks received 6290 */ 6291 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6292 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6293 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6294 }; 6295 6296 struct bpf_sock_tuple { 6297 union { 6298 struct { 6299 __be32 saddr; 6300 __be32 daddr; 6301 __be16 sport; 6302 __be16 dport; 6303 } ipv4; 6304 struct { 6305 __be32 saddr[4]; 6306 __be32 daddr[4]; 6307 __be16 sport; 6308 __be16 dport; 6309 } ipv6; 6310 }; 6311 }; 6312 6313 /* (Simplified) user return codes for tcx prog type. 6314 * A valid tcx program must return one of these defined values. All other 6315 * return codes are reserved for future use. Must remain compatible with 6316 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6317 * return codes are mapped to TCX_NEXT. 6318 */ 6319 enum tcx_action_base { 6320 TCX_NEXT = -1, 6321 TCX_PASS = 0, 6322 TCX_DROP = 2, 6323 TCX_REDIRECT = 7, 6324 }; 6325 6326 struct bpf_xdp_sock { 6327 __u32 queue_id; 6328 }; 6329 6330 #define XDP_PACKET_HEADROOM 256 6331 6332 /* User return codes for XDP prog type. 6333 * A valid XDP program must return one of these defined values. All other 6334 * return codes are reserved for future use. Unknown return codes will 6335 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6336 */ 6337 enum xdp_action { 6338 XDP_ABORTED = 0, 6339 XDP_DROP, 6340 XDP_PASS, 6341 XDP_TX, 6342 XDP_REDIRECT, 6343 }; 6344 6345 /* user accessible metadata for XDP packet hook 6346 * new fields must be added to the end of this structure 6347 */ 6348 struct xdp_md { 6349 __u32 data; 6350 __u32 data_end; 6351 __u32 data_meta; 6352 /* Below access go through struct xdp_rxq_info */ 6353 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6354 __u32 rx_queue_index; /* rxq->queue_index */ 6355 6356 __u32 egress_ifindex; /* txq->dev->ifindex */ 6357 }; 6358 6359 /* DEVMAP map-value layout 6360 * 6361 * The struct data-layout of map-value is a configuration interface. 6362 * New members can only be added to the end of this structure. 6363 */ 6364 struct bpf_devmap_val { 6365 __u32 ifindex; /* device index */ 6366 union { 6367 int fd; /* prog fd on map write */ 6368 __u32 id; /* prog id on map read */ 6369 } bpf_prog; 6370 }; 6371 6372 /* CPUMAP map-value layout 6373 * 6374 * The struct data-layout of map-value is a configuration interface. 6375 * New members can only be added to the end of this structure. 6376 */ 6377 struct bpf_cpumap_val { 6378 __u32 qsize; /* queue size to remote target CPU */ 6379 union { 6380 int fd; /* prog fd on map write */ 6381 __u32 id; /* prog id on map read */ 6382 } bpf_prog; 6383 }; 6384 6385 enum sk_action { 6386 SK_DROP = 0, 6387 SK_PASS, 6388 }; 6389 6390 /* user accessible metadata for SK_MSG packet hook, new fields must 6391 * be added to the end of this structure 6392 */ 6393 struct sk_msg_md { 6394 __bpf_md_ptr(void *, data); 6395 __bpf_md_ptr(void *, data_end); 6396 6397 __u32 family; 6398 __u32 remote_ip4; /* Stored in network byte order */ 6399 __u32 local_ip4; /* Stored in network byte order */ 6400 __u32 remote_ip6[4]; /* Stored in network byte order */ 6401 __u32 local_ip6[4]; /* Stored in network byte order */ 6402 __u32 remote_port; /* Stored in network byte order */ 6403 __u32 local_port; /* stored in host byte order */ 6404 __u32 size; /* Total size of sk_msg */ 6405 6406 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6407 }; 6408 6409 struct sk_reuseport_md { 6410 /* 6411 * Start of directly accessible data. It begins from 6412 * the tcp/udp header. 6413 */ 6414 __bpf_md_ptr(void *, data); 6415 /* End of directly accessible data */ 6416 __bpf_md_ptr(void *, data_end); 6417 /* 6418 * Total length of packet (starting from the tcp/udp header). 6419 * Note that the directly accessible bytes (data_end - data) 6420 * could be less than this "len". Those bytes could be 6421 * indirectly read by a helper "bpf_skb_load_bytes()". 6422 */ 6423 __u32 len; 6424 /* 6425 * Eth protocol in the mac header (network byte order). e.g. 6426 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6427 */ 6428 __u32 eth_protocol; 6429 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6430 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6431 __u32 hash; /* A hash of the packet 4 tuples */ 6432 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6433 * new incoming connection request (e.g. selecting a listen sk for 6434 * the received SYN in the TCP case). reuse->sk is one of the sk 6435 * in the reuseport group. The bpf prog can use reuse->sk to learn 6436 * the local listening ip/port without looking into the skb. 6437 * 6438 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6439 * reuse->migrating_sk is the socket that needs to be migrated 6440 * to another listening socket. migrating_sk could be a fullsock 6441 * sk that is fully established or a reqsk that is in-the-middle 6442 * of 3-way handshake. 6443 */ 6444 __bpf_md_ptr(struct bpf_sock *, sk); 6445 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6446 }; 6447 6448 #define BPF_TAG_SIZE 8 6449 6450 struct bpf_prog_info { 6451 __u32 type; 6452 __u32 id; 6453 __u8 tag[BPF_TAG_SIZE]; 6454 __u32 jited_prog_len; 6455 __u32 xlated_prog_len; 6456 __aligned_u64 jited_prog_insns; 6457 __aligned_u64 xlated_prog_insns; 6458 __u64 load_time; /* ns since boottime */ 6459 __u32 created_by_uid; 6460 __u32 nr_map_ids; 6461 __aligned_u64 map_ids; 6462 char name[BPF_OBJ_NAME_LEN]; 6463 __u32 ifindex; 6464 __u32 gpl_compatible:1; 6465 __u32 :31; /* alignment pad */ 6466 __u64 netns_dev; 6467 __u64 netns_ino; 6468 __u32 nr_jited_ksyms; 6469 __u32 nr_jited_func_lens; 6470 __aligned_u64 jited_ksyms; 6471 __aligned_u64 jited_func_lens; 6472 __u32 btf_id; 6473 __u32 func_info_rec_size; 6474 __aligned_u64 func_info; 6475 __u32 nr_func_info; 6476 __u32 nr_line_info; 6477 __aligned_u64 line_info; 6478 __aligned_u64 jited_line_info; 6479 __u32 nr_jited_line_info; 6480 __u32 line_info_rec_size; 6481 __u32 jited_line_info_rec_size; 6482 __u32 nr_prog_tags; 6483 __aligned_u64 prog_tags; 6484 __u64 run_time_ns; 6485 __u64 run_cnt; 6486 __u64 recursion_misses; 6487 __u32 verified_insns; 6488 __u32 attach_btf_obj_id; 6489 __u32 attach_btf_id; 6490 } __attribute__((aligned(8))); 6491 6492 struct bpf_map_info { 6493 __u32 type; 6494 __u32 id; 6495 __u32 key_size; 6496 __u32 value_size; 6497 __u32 max_entries; 6498 __u32 map_flags; 6499 char name[BPF_OBJ_NAME_LEN]; 6500 __u32 ifindex; 6501 __u32 btf_vmlinux_value_type_id; 6502 __u64 netns_dev; 6503 __u64 netns_ino; 6504 __u32 btf_id; 6505 __u32 btf_key_type_id; 6506 __u32 btf_value_type_id; 6507 __u32 :32; /* alignment pad */ 6508 __u64 map_extra; 6509 } __attribute__((aligned(8))); 6510 6511 struct bpf_btf_info { 6512 __aligned_u64 btf; 6513 __u32 btf_size; 6514 __u32 id; 6515 __aligned_u64 name; 6516 __u32 name_len; 6517 __u32 kernel_btf; 6518 } __attribute__((aligned(8))); 6519 6520 struct bpf_link_info { 6521 __u32 type; 6522 __u32 id; 6523 __u32 prog_id; 6524 union { 6525 struct { 6526 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6527 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6528 } raw_tracepoint; 6529 struct { 6530 __u32 attach_type; 6531 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6532 __u32 target_btf_id; /* BTF type id inside the object */ 6533 } tracing; 6534 struct { 6535 __u64 cgroup_id; 6536 __u32 attach_type; 6537 } cgroup; 6538 struct { 6539 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6540 __u32 target_name_len; /* in/out: target_name buffer len */ 6541 6542 /* If the iter specific field is 32 bits, it can be put 6543 * in the first or second union. Otherwise it should be 6544 * put in the second union. 6545 */ 6546 union { 6547 struct { 6548 __u32 map_id; 6549 } map; 6550 }; 6551 union { 6552 struct { 6553 __u64 cgroup_id; 6554 __u32 order; 6555 } cgroup; 6556 struct { 6557 __u32 tid; 6558 __u32 pid; 6559 } task; 6560 }; 6561 } iter; 6562 struct { 6563 __u32 netns_ino; 6564 __u32 attach_type; 6565 } netns; 6566 struct { 6567 __u32 ifindex; 6568 } xdp; 6569 struct { 6570 __u32 map_id; 6571 } struct_ops; 6572 struct { 6573 __u32 pf; 6574 __u32 hooknum; 6575 __s32 priority; 6576 __u32 flags; 6577 } netfilter; 6578 struct { 6579 __aligned_u64 addrs; 6580 __u32 count; /* in/out: kprobe_multi function count */ 6581 __u32 flags; 6582 __u64 missed; 6583 } kprobe_multi; 6584 struct { 6585 __aligned_u64 path; 6586 __aligned_u64 offsets; 6587 __aligned_u64 ref_ctr_offsets; 6588 __aligned_u64 cookies; 6589 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6590 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6591 __u32 flags; 6592 __u32 pid; 6593 } uprobe_multi; 6594 struct { 6595 __u32 type; /* enum bpf_perf_event_type */ 6596 __u32 :32; 6597 union { 6598 struct { 6599 __aligned_u64 file_name; /* in/out */ 6600 __u32 name_len; 6601 __u32 offset; /* offset from file_name */ 6602 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6603 struct { 6604 __aligned_u64 func_name; /* in/out */ 6605 __u32 name_len; 6606 __u32 offset; /* offset from func_name */ 6607 __u64 addr; 6608 __u64 missed; 6609 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6610 struct { 6611 __aligned_u64 tp_name; /* in/out */ 6612 __u32 name_len; 6613 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6614 struct { 6615 __u64 config; 6616 __u32 type; 6617 } event; /* BPF_PERF_EVENT_EVENT */ 6618 }; 6619 } perf_event; 6620 struct { 6621 __u32 ifindex; 6622 __u32 attach_type; 6623 } tcx; 6624 struct { 6625 __u32 ifindex; 6626 __u32 attach_type; 6627 } netkit; 6628 }; 6629 } __attribute__((aligned(8))); 6630 6631 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6632 * by user and intended to be used by socket (e.g. to bind to, depends on 6633 * attach type). 6634 */ 6635 struct bpf_sock_addr { 6636 __u32 user_family; /* Allows 4-byte read, but no write. */ 6637 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6638 * Stored in network byte order. 6639 */ 6640 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6641 * Stored in network byte order. 6642 */ 6643 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6644 * Stored in network byte order 6645 */ 6646 __u32 family; /* Allows 4-byte read, but no write */ 6647 __u32 type; /* Allows 4-byte read, but no write */ 6648 __u32 protocol; /* Allows 4-byte read, but no write */ 6649 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6650 * Stored in network byte order. 6651 */ 6652 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6653 * Stored in network byte order. 6654 */ 6655 __bpf_md_ptr(struct bpf_sock *, sk); 6656 }; 6657 6658 /* User bpf_sock_ops struct to access socket values and specify request ops 6659 * and their replies. 6660 * Some of this fields are in network (bigendian) byte order and may need 6661 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6662 * New fields can only be added at the end of this structure 6663 */ 6664 struct bpf_sock_ops { 6665 __u32 op; 6666 union { 6667 __u32 args[4]; /* Optionally passed to bpf program */ 6668 __u32 reply; /* Returned by bpf program */ 6669 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6670 }; 6671 __u32 family; 6672 __u32 remote_ip4; /* Stored in network byte order */ 6673 __u32 local_ip4; /* Stored in network byte order */ 6674 __u32 remote_ip6[4]; /* Stored in network byte order */ 6675 __u32 local_ip6[4]; /* Stored in network byte order */ 6676 __u32 remote_port; /* Stored in network byte order */ 6677 __u32 local_port; /* stored in host byte order */ 6678 __u32 is_fullsock; /* Some TCP fields are only valid if 6679 * there is a full socket. If not, the 6680 * fields read as zero. 6681 */ 6682 __u32 snd_cwnd; 6683 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6684 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6685 __u32 state; 6686 __u32 rtt_min; 6687 __u32 snd_ssthresh; 6688 __u32 rcv_nxt; 6689 __u32 snd_nxt; 6690 __u32 snd_una; 6691 __u32 mss_cache; 6692 __u32 ecn_flags; 6693 __u32 rate_delivered; 6694 __u32 rate_interval_us; 6695 __u32 packets_out; 6696 __u32 retrans_out; 6697 __u32 total_retrans; 6698 __u32 segs_in; 6699 __u32 data_segs_in; 6700 __u32 segs_out; 6701 __u32 data_segs_out; 6702 __u32 lost_out; 6703 __u32 sacked_out; 6704 __u32 sk_txhash; 6705 __u64 bytes_received; 6706 __u64 bytes_acked; 6707 __bpf_md_ptr(struct bpf_sock *, sk); 6708 /* [skb_data, skb_data_end) covers the whole TCP header. 6709 * 6710 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6711 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6712 * header has not been written. 6713 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6714 * been written so far. 6715 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6716 * the 3WHS. 6717 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6718 * the 3WHS. 6719 * 6720 * bpf_load_hdr_opt() can also be used to read a particular option. 6721 */ 6722 __bpf_md_ptr(void *, skb_data); 6723 __bpf_md_ptr(void *, skb_data_end); 6724 __u32 skb_len; /* The total length of a packet. 6725 * It includes the header, options, 6726 * and payload. 6727 */ 6728 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6729 * an easy way to check for tcp_flags 6730 * without parsing skb_data. 6731 * 6732 * In particular, the skb_tcp_flags 6733 * will still be available in 6734 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6735 * the outgoing header has not 6736 * been written yet. 6737 */ 6738 __u64 skb_hwtstamp; 6739 }; 6740 6741 /* Definitions for bpf_sock_ops_cb_flags */ 6742 enum { 6743 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6744 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6745 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6746 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6747 /* Call bpf for all received TCP headers. The bpf prog will be 6748 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6749 * 6750 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6751 * for the header option related helpers that will be useful 6752 * to the bpf programs. 6753 * 6754 * It could be used at the client/active side (i.e. connect() side) 6755 * when the server told it that the server was in syncookie 6756 * mode and required the active side to resend the bpf-written 6757 * options. The active side can keep writing the bpf-options until 6758 * it received a valid packet from the server side to confirm 6759 * the earlier packet (and options) has been received. The later 6760 * example patch is using it like this at the active side when the 6761 * server is in syncookie mode. 6762 * 6763 * The bpf prog will usually turn this off in the common cases. 6764 */ 6765 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6766 /* Call bpf when kernel has received a header option that 6767 * the kernel cannot handle. The bpf prog will be called under 6768 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6769 * 6770 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6771 * for the header option related helpers that will be useful 6772 * to the bpf programs. 6773 */ 6774 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6775 /* Call bpf when the kernel is writing header options for the 6776 * outgoing packet. The bpf prog will first be called 6777 * to reserve space in a skb under 6778 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6779 * the bpf prog will be called to write the header option(s) 6780 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6781 * 6782 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6783 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6784 * related helpers that will be useful to the bpf programs. 6785 * 6786 * The kernel gets its chance to reserve space and write 6787 * options first before the BPF program does. 6788 */ 6789 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6790 /* Mask of all currently supported cb flags */ 6791 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6792 }; 6793 6794 /* List of known BPF sock_ops operators. 6795 * New entries can only be added at the end 6796 */ 6797 enum { 6798 BPF_SOCK_OPS_VOID, 6799 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6800 * -1 if default value should be used 6801 */ 6802 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6803 * window (in packets) or -1 if default 6804 * value should be used 6805 */ 6806 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6807 * active connection is initialized 6808 */ 6809 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6810 * active connection is 6811 * established 6812 */ 6813 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6814 * passive connection is 6815 * established 6816 */ 6817 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6818 * needs ECN 6819 */ 6820 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6821 * based on the path and may be 6822 * dependent on the congestion control 6823 * algorithm. In general it indicates 6824 * a congestion threshold. RTTs above 6825 * this indicate congestion 6826 */ 6827 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6828 * Arg1: value of icsk_retransmits 6829 * Arg2: value of icsk_rto 6830 * Arg3: whether RTO has expired 6831 */ 6832 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6833 * Arg1: sequence number of 1st byte 6834 * Arg2: # segments 6835 * Arg3: return value of 6836 * tcp_transmit_skb (0 => success) 6837 */ 6838 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6839 * Arg1: old_state 6840 * Arg2: new_state 6841 */ 6842 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6843 * socket transition to LISTEN state. 6844 */ 6845 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6846 */ 6847 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6848 * It will be called to handle 6849 * the packets received at 6850 * an already established 6851 * connection. 6852 * 6853 * sock_ops->skb_data: 6854 * Referring to the received skb. 6855 * It covers the TCP header only. 6856 * 6857 * bpf_load_hdr_opt() can also 6858 * be used to search for a 6859 * particular option. 6860 */ 6861 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6862 * header option later in 6863 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6864 * Arg1: bool want_cookie. (in 6865 * writing SYNACK only) 6866 * 6867 * sock_ops->skb_data: 6868 * Not available because no header has 6869 * been written yet. 6870 * 6871 * sock_ops->skb_tcp_flags: 6872 * The tcp_flags of the 6873 * outgoing skb. (e.g. SYN, ACK, FIN). 6874 * 6875 * bpf_reserve_hdr_opt() should 6876 * be used to reserve space. 6877 */ 6878 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6879 * Arg1: bool want_cookie. (in 6880 * writing SYNACK only) 6881 * 6882 * sock_ops->skb_data: 6883 * Referring to the outgoing skb. 6884 * It covers the TCP header 6885 * that has already been written 6886 * by the kernel and the 6887 * earlier bpf-progs. 6888 * 6889 * sock_ops->skb_tcp_flags: 6890 * The tcp_flags of the outgoing 6891 * skb. (e.g. SYN, ACK, FIN). 6892 * 6893 * bpf_store_hdr_opt() should 6894 * be used to write the 6895 * option. 6896 * 6897 * bpf_load_hdr_opt() can also 6898 * be used to search for a 6899 * particular option that 6900 * has already been written 6901 * by the kernel or the 6902 * earlier bpf-progs. 6903 */ 6904 }; 6905 6906 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6907 * changes between the TCP and BPF versions. Ideally this should never happen. 6908 * If it does, we need to add code to convert them before calling 6909 * the BPF sock_ops function. 6910 */ 6911 enum { 6912 BPF_TCP_ESTABLISHED = 1, 6913 BPF_TCP_SYN_SENT, 6914 BPF_TCP_SYN_RECV, 6915 BPF_TCP_FIN_WAIT1, 6916 BPF_TCP_FIN_WAIT2, 6917 BPF_TCP_TIME_WAIT, 6918 BPF_TCP_CLOSE, 6919 BPF_TCP_CLOSE_WAIT, 6920 BPF_TCP_LAST_ACK, 6921 BPF_TCP_LISTEN, 6922 BPF_TCP_CLOSING, /* Now a valid state */ 6923 BPF_TCP_NEW_SYN_RECV, 6924 BPF_TCP_BOUND_INACTIVE, 6925 6926 BPF_TCP_MAX_STATES /* Leave at the end! */ 6927 }; 6928 6929 enum { 6930 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6931 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6932 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6933 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6934 /* Copy the SYN pkt to optval 6935 * 6936 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6937 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6938 * to only getting from the saved_syn. It can either get the 6939 * syn packet from: 6940 * 6941 * 1. the just-received SYN packet (only available when writing the 6942 * SYNACK). It will be useful when it is not necessary to 6943 * save the SYN packet for latter use. It is also the only way 6944 * to get the SYN during syncookie mode because the syn 6945 * packet cannot be saved during syncookie. 6946 * 6947 * OR 6948 * 6949 * 2. the earlier saved syn which was done by 6950 * bpf_setsockopt(TCP_SAVE_SYN). 6951 * 6952 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 6953 * SYN packet is obtained. 6954 * 6955 * If the bpf-prog does not need the IP[46] header, the 6956 * bpf-prog can avoid parsing the IP header by using 6957 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 6958 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 6959 * 6960 * >0: Total number of bytes copied 6961 * -ENOSPC: Not enough space in optval. Only optlen number of 6962 * bytes is copied. 6963 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 6964 * is not saved by setsockopt(TCP_SAVE_SYN). 6965 */ 6966 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 6967 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6968 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 6969 }; 6970 6971 enum { 6972 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 6973 }; 6974 6975 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 6976 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6977 */ 6978 enum { 6979 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 6980 * total option spaces 6981 * required for an established 6982 * sk in order to calculate the 6983 * MSS. No skb is actually 6984 * sent. 6985 */ 6986 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 6987 * when sending a SYN. 6988 */ 6989 }; 6990 6991 struct bpf_perf_event_value { 6992 __u64 counter; 6993 __u64 enabled; 6994 __u64 running; 6995 }; 6996 6997 enum { 6998 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 6999 BPF_DEVCG_ACC_READ = (1ULL << 1), 7000 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7001 }; 7002 7003 enum { 7004 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7005 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7006 }; 7007 7008 struct bpf_cgroup_dev_ctx { 7009 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7010 __u32 access_type; 7011 __u32 major; 7012 __u32 minor; 7013 }; 7014 7015 struct bpf_raw_tracepoint_args { 7016 __u64 args[0]; 7017 }; 7018 7019 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7020 * OUTPUT: Do lookup from egress perspective; default is ingress 7021 */ 7022 enum { 7023 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7024 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7025 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7026 BPF_FIB_LOOKUP_TBID = (1U << 3), 7027 BPF_FIB_LOOKUP_SRC = (1U << 4), 7028 }; 7029 7030 enum { 7031 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7032 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7033 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7034 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7035 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7036 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7037 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7038 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7039 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7040 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7041 }; 7042 7043 struct bpf_fib_lookup { 7044 /* input: network family for lookup (AF_INET, AF_INET6) 7045 * output: network family of egress nexthop 7046 */ 7047 __u8 family; 7048 7049 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7050 __u8 l4_protocol; 7051 __be16 sport; 7052 __be16 dport; 7053 7054 union { /* used for MTU check */ 7055 /* input to lookup */ 7056 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7057 7058 /* output: MTU value */ 7059 __u16 mtu_result; 7060 } __attribute__((packed, aligned(2))); 7061 /* input: L3 device index for lookup 7062 * output: device index from FIB lookup 7063 */ 7064 __u32 ifindex; 7065 7066 union { 7067 /* inputs to lookup */ 7068 __u8 tos; /* AF_INET */ 7069 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7070 7071 /* output: metric of fib result (IPv4/IPv6 only) */ 7072 __u32 rt_metric; 7073 }; 7074 7075 /* input: source address to consider for lookup 7076 * output: source address result from lookup 7077 */ 7078 union { 7079 __be32 ipv4_src; 7080 __u32 ipv6_src[4]; /* in6_addr; network order */ 7081 }; 7082 7083 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7084 * network header. output: bpf_fib_lookup sets to gateway address 7085 * if FIB lookup returns gateway route 7086 */ 7087 union { 7088 __be32 ipv4_dst; 7089 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7090 }; 7091 7092 union { 7093 struct { 7094 /* output */ 7095 __be16 h_vlan_proto; 7096 __be16 h_vlan_TCI; 7097 }; 7098 /* input: when accompanied with the 7099 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7100 * specific routing table to use for the fib lookup. 7101 */ 7102 __u32 tbid; 7103 }; 7104 7105 __u8 smac[6]; /* ETH_ALEN */ 7106 __u8 dmac[6]; /* ETH_ALEN */ 7107 }; 7108 7109 struct bpf_redir_neigh { 7110 /* network family for lookup (AF_INET, AF_INET6) */ 7111 __u32 nh_family; 7112 /* network address of nexthop; skips fib lookup to find gateway */ 7113 union { 7114 __be32 ipv4_nh; 7115 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7116 }; 7117 }; 7118 7119 /* bpf_check_mtu flags*/ 7120 enum bpf_check_mtu_flags { 7121 BPF_MTU_CHK_SEGS = (1U << 0), 7122 }; 7123 7124 enum bpf_check_mtu_ret { 7125 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7126 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7127 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7128 }; 7129 7130 enum bpf_task_fd_type { 7131 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7132 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7133 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7134 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7135 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7136 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7137 }; 7138 7139 enum { 7140 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7141 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7142 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7143 }; 7144 7145 struct bpf_flow_keys { 7146 __u16 nhoff; 7147 __u16 thoff; 7148 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7149 __u8 is_frag; 7150 __u8 is_first_frag; 7151 __u8 is_encap; 7152 __u8 ip_proto; 7153 __be16 n_proto; 7154 __be16 sport; 7155 __be16 dport; 7156 union { 7157 struct { 7158 __be32 ipv4_src; 7159 __be32 ipv4_dst; 7160 }; 7161 struct { 7162 __u32 ipv6_src[4]; /* in6_addr; network order */ 7163 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7164 }; 7165 }; 7166 __u32 flags; 7167 __be32 flow_label; 7168 }; 7169 7170 struct bpf_func_info { 7171 __u32 insn_off; 7172 __u32 type_id; 7173 }; 7174 7175 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7176 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7177 7178 struct bpf_line_info { 7179 __u32 insn_off; 7180 __u32 file_name_off; 7181 __u32 line_off; 7182 __u32 line_col; 7183 }; 7184 7185 struct bpf_spin_lock { 7186 __u32 val; 7187 }; 7188 7189 struct bpf_timer { 7190 __u64 __opaque[2]; 7191 } __attribute__((aligned(8))); 7192 7193 struct bpf_dynptr { 7194 __u64 __opaque[2]; 7195 } __attribute__((aligned(8))); 7196 7197 struct bpf_list_head { 7198 __u64 __opaque[2]; 7199 } __attribute__((aligned(8))); 7200 7201 struct bpf_list_node { 7202 __u64 __opaque[3]; 7203 } __attribute__((aligned(8))); 7204 7205 struct bpf_rb_root { 7206 __u64 __opaque[2]; 7207 } __attribute__((aligned(8))); 7208 7209 struct bpf_rb_node { 7210 __u64 __opaque[4]; 7211 } __attribute__((aligned(8))); 7212 7213 struct bpf_refcount { 7214 __u32 __opaque[1]; 7215 } __attribute__((aligned(4))); 7216 7217 struct bpf_sysctl { 7218 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7219 * Allows 1,2,4-byte read, but no write. 7220 */ 7221 __u32 file_pos; /* Sysctl file position to read from, write to. 7222 * Allows 1,2,4-byte read an 4-byte write. 7223 */ 7224 }; 7225 7226 struct bpf_sockopt { 7227 __bpf_md_ptr(struct bpf_sock *, sk); 7228 __bpf_md_ptr(void *, optval); 7229 __bpf_md_ptr(void *, optval_end); 7230 7231 __s32 level; 7232 __s32 optname; 7233 __s32 optlen; 7234 __s32 retval; 7235 }; 7236 7237 struct bpf_pidns_info { 7238 __u32 pid; 7239 __u32 tgid; 7240 }; 7241 7242 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7243 struct bpf_sk_lookup { 7244 union { 7245 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7246 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7247 }; 7248 7249 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7250 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7251 __u32 remote_ip4; /* Network byte order */ 7252 __u32 remote_ip6[4]; /* Network byte order */ 7253 __be16 remote_port; /* Network byte order */ 7254 __u16 :16; /* Zero padding */ 7255 __u32 local_ip4; /* Network byte order */ 7256 __u32 local_ip6[4]; /* Network byte order */ 7257 __u32 local_port; /* Host byte order */ 7258 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7259 }; 7260 7261 /* 7262 * struct btf_ptr is used for typed pointer representation; the 7263 * type id is used to render the pointer data as the appropriate type 7264 * via the bpf_snprintf_btf() helper described above. A flags field - 7265 * potentially to specify additional details about the BTF pointer 7266 * (rather than its mode of display) - is included for future use. 7267 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7268 */ 7269 struct btf_ptr { 7270 void *ptr; 7271 __u32 type_id; 7272 __u32 flags; /* BTF ptr flags; unused at present. */ 7273 }; 7274 7275 /* 7276 * Flags to control bpf_snprintf_btf() behaviour. 7277 * - BTF_F_COMPACT: no formatting around type information 7278 * - BTF_F_NONAME: no struct/union member names/types 7279 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7280 * equivalent to %px. 7281 * - BTF_F_ZERO: show zero-valued struct/union members; they 7282 * are not displayed by default 7283 */ 7284 enum { 7285 BTF_F_COMPACT = (1ULL << 0), 7286 BTF_F_NONAME = (1ULL << 1), 7287 BTF_F_PTR_RAW = (1ULL << 2), 7288 BTF_F_ZERO = (1ULL << 3), 7289 }; 7290 7291 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7292 * has to be adjusted by relocations. It is emitted by llvm and passed to 7293 * libbpf and later to the kernel. 7294 */ 7295 enum bpf_core_relo_kind { 7296 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7297 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7298 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7299 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7300 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7301 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7302 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7303 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7304 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7305 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7306 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7307 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7308 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7309 }; 7310 7311 /* 7312 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7313 * and from libbpf to the kernel. 7314 * 7315 * CO-RE relocation captures the following data: 7316 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7317 * its insn->imm field to be relocated with actual field info; 7318 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7319 * type or field; 7320 * - access_str_off - offset into corresponding .BTF string section. String 7321 * interpretation depends on specific relocation kind: 7322 * - for field-based relocations, string encodes an accessed field using 7323 * a sequence of field and array indices, separated by colon (:). It's 7324 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7325 * arguments for identifying offset to a field. 7326 * - for type-based relocations, strings is expected to be just "0"; 7327 * - for enum value-based relocations, string contains an index of enum 7328 * value within its enum type; 7329 * - kind - one of enum bpf_core_relo_kind; 7330 * 7331 * Example: 7332 * struct sample { 7333 * int a; 7334 * struct { 7335 * int b[10]; 7336 * }; 7337 * }; 7338 * 7339 * struct sample *s = ...; 7340 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7341 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7342 * // b is field #0 inside anon struct, accessing elem #5) 7343 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7344 * 7345 * type_id for all relocs in this example will capture BTF type id of 7346 * `struct sample`. 7347 * 7348 * Such relocation is emitted when using __builtin_preserve_access_index() 7349 * Clang built-in, passing expression that captures field address, e.g.: 7350 * 7351 * bpf_probe_read(&dst, sizeof(dst), 7352 * __builtin_preserve_access_index(&src->a.b.c)); 7353 * 7354 * In this case Clang will emit field relocation recording necessary data to 7355 * be able to find offset of embedded `a.b.c` field within `src` struct. 7356 * 7357 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7358 */ 7359 struct bpf_core_relo { 7360 __u32 insn_off; 7361 __u32 type_id; 7362 __u32 access_str_off; 7363 enum bpf_core_relo_kind kind; 7364 }; 7365 7366 /* 7367 * Flags to control bpf_timer_start() behaviour. 7368 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7369 * relative to current time. 7370 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7371 */ 7372 enum { 7373 BPF_F_TIMER_ABS = (1ULL << 0), 7374 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7375 }; 7376 7377 /* BPF numbers iterator state */ 7378 struct bpf_iter_num { 7379 /* opaque iterator state; having __u64 here allows to preserve correct 7380 * alignment requirements in vmlinux.h, generated from BTF 7381 */ 7382 __u64 __opaque[1]; 7383 } __attribute__((aligned(8))); 7384 7385 #endif /* __LINUX_BPF_H__ */
Contact us
|
About us
|
Term of use
|
Copyright © 2000-2025 MyWebUniversity.com ™