1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_MEMSX 0x80 /* load with sign extension */ 23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 25 26 /* alu/jmp fields */ 27 #define BPF_MOV 0xb0 /* mov reg to reg */ 28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 29 30 /* change endianness of a register */ 31 #define BPF_END 0xd0 /* flags for endianness conversion: */ 32 #define BPF_TO_LE 0x00 /* convert to little-endian */ 33 #define BPF_TO_BE 0x08 /* convert to big-endian */ 34 #define BPF_FROM_LE BPF_TO_LE 35 #define BPF_FROM_BE BPF_TO_BE 36 37 /* jmp encodings */ 38 #define BPF_JNE 0x50 /* jump != */ 39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 45 #define BPF_JCOND 0xe0 /* conditional pseudo jumps: may_goto, goto_or_nop */ 46 #define BPF_CALL 0x80 /* function call */ 47 #define BPF_EXIT 0x90 /* function return */ 48 49 /* atomic op type fields (stored in immediate) */ 50 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 51 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 52 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 53 54 enum bpf_cond_pseudo_jmp { 55 BPF_MAY_GOTO = 0, 56 }; 57 58 /* Register numbers */ 59 enum { 60 BPF_REG_0 = 0, 61 BPF_REG_1, 62 BPF_REG_2, 63 BPF_REG_3, 64 BPF_REG_4, 65 BPF_REG_5, 66 BPF_REG_6, 67 BPF_REG_7, 68 BPF_REG_8, 69 BPF_REG_9, 70 BPF_REG_10, 71 __MAX_BPF_REG, 72 }; 73 74 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 75 #define MAX_BPF_REG __MAX_BPF_REG 76 77 struct bpf_insn { 78 __u8 code; /* opcode */ 79 __u8 dst_reg:4; /* dest register */ 80 __u8 src_reg:4; /* source register */ 81 __s16 off; /* signed offset */ 82 __s32 imm; /* signed immediate constant */ 83 }; 84 85 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 86 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 87 * the trailing flexible array member) instead. 88 */ 89 struct bpf_lpm_trie_key { 90 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 91 __u8 data[0]; /* Arbitrary size */ 92 }; 93 94 /* Header for bpf_lpm_trie_key structs */ 95 struct bpf_lpm_trie_key_hdr { 96 __u32 prefixlen; 97 }; 98 99 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 100 struct bpf_lpm_trie_key_u8 { 101 union { 102 struct bpf_lpm_trie_key_hdr hdr; 103 __u32 prefixlen; 104 }; 105 __u8 data[]; /* Arbitrary size */ 106 }; 107 108 struct bpf_cgroup_storage_key { 109 __u64 cgroup_inode_id; /* cgroup inode id */ 110 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 111 }; 112 113 enum bpf_cgroup_iter_order { 114 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 115 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 116 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 117 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 118 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 119 }; 120 121 union bpf_iter_link_info { 122 struct { 123 __u32 map_fd; 124 } map; 125 struct { 126 enum bpf_cgroup_iter_order order; 127 128 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 129 * both are zero, the walk starts from the default cgroup v2 130 * root. For walking v1 hierarchy, one should always explicitly 131 * specify cgroup_fd. 132 */ 133 __u32 cgroup_fd; 134 __u64 cgroup_id; 135 } cgroup; 136 /* Parameters of task iterators. */ 137 struct { 138 __u32 tid; 139 __u32 pid; 140 __u32 pid_fd; 141 } task; 142 }; 143 144 /* BPF syscall commands, see bpf(2) man-page for more details. */ 145 /** 146 * DOC: eBPF Syscall Preamble 147 * 148 * The operation to be performed by the **bpf**\ () system call is determined 149 * by the *cmd* argument. Each operation takes an accompanying argument, 150 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 151 * below). The size argument is the size of the union pointed to by *attr*. 152 */ 153 /** 154 * DOC: eBPF Syscall Commands 155 * 156 * BPF_MAP_CREATE 157 * Description 158 * Create a map and return a file descriptor that refers to the 159 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 160 * is automatically enabled for the new file descriptor. 161 * 162 * Applying **close**\ (2) to the file descriptor returned by 163 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 164 * 165 * Return 166 * A new file descriptor (a nonnegative integer), or -1 if an 167 * error occurred (in which case, *errno* is set appropriately). 168 * 169 * BPF_MAP_LOOKUP_ELEM 170 * Description 171 * Look up an element with a given *key* in the map referred to 172 * by the file descriptor *map_fd*. 173 * 174 * The *flags* argument may be specified as one of the 175 * following: 176 * 177 * **BPF_F_LOCK** 178 * Look up the value of a spin-locked map without 179 * returning the lock. This must be specified if the 180 * elements contain a spinlock. 181 * 182 * Return 183 * Returns zero on success. On error, -1 is returned and *errno* 184 * is set appropriately. 185 * 186 * BPF_MAP_UPDATE_ELEM 187 * Description 188 * Create or update an element (key/value pair) in a specified map. 189 * 190 * The *flags* argument should be specified as one of the 191 * following: 192 * 193 * **BPF_ANY** 194 * Create a new element or update an existing element. 195 * **BPF_NOEXIST** 196 * Create a new element only if it did not exist. 197 * **BPF_EXIST** 198 * Update an existing element. 199 * **BPF_F_LOCK** 200 * Update a spin_lock-ed map element. 201 * 202 * Return 203 * Returns zero on success. On error, -1 is returned and *errno* 204 * is set appropriately. 205 * 206 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 207 * **E2BIG**, **EEXIST**, or **ENOENT**. 208 * 209 * **E2BIG** 210 * The number of elements in the map reached the 211 * *max_entries* limit specified at map creation time. 212 * **EEXIST** 213 * If *flags* specifies **BPF_NOEXIST** and the element 214 * with *key* already exists in the map. 215 * **ENOENT** 216 * If *flags* specifies **BPF_EXIST** and the element with 217 * *key* does not exist in the map. 218 * 219 * BPF_MAP_DELETE_ELEM 220 * Description 221 * Look up and delete an element by key in a specified map. 222 * 223 * Return 224 * Returns zero on success. On error, -1 is returned and *errno* 225 * is set appropriately. 226 * 227 * BPF_MAP_GET_NEXT_KEY 228 * Description 229 * Look up an element by key in a specified map and return the key 230 * of the next element. Can be used to iterate over all elements 231 * in the map. 232 * 233 * Return 234 * Returns zero on success. On error, -1 is returned and *errno* 235 * is set appropriately. 236 * 237 * The following cases can be used to iterate over all elements of 238 * the map: 239 * 240 * * If *key* is not found, the operation returns zero and sets 241 * the *next_key* pointer to the key of the first element. 242 * * If *key* is found, the operation returns zero and sets the 243 * *next_key* pointer to the key of the next element. 244 * * If *key* is the last element, returns -1 and *errno* is set 245 * to **ENOENT**. 246 * 247 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 248 * **EINVAL** on error. 249 * 250 * BPF_PROG_LOAD 251 * Description 252 * Verify and load an eBPF program, returning a new file 253 * descriptor associated with the program. 254 * 255 * Applying **close**\ (2) to the file descriptor returned by 256 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 257 * 258 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 259 * automatically enabled for the new file descriptor. 260 * 261 * Return 262 * A new file descriptor (a nonnegative integer), or -1 if an 263 * error occurred (in which case, *errno* is set appropriately). 264 * 265 * BPF_OBJ_PIN 266 * Description 267 * Pin an eBPF program or map referred by the specified *bpf_fd* 268 * to the provided *pathname* on the filesystem. 269 * 270 * The *pathname* argument must not contain a dot ("."). 271 * 272 * On success, *pathname* retains a reference to the eBPF object, 273 * preventing deallocation of the object when the original 274 * *bpf_fd* is closed. This allow the eBPF object to live beyond 275 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 276 * process. 277 * 278 * Applying **unlink**\ (2) or similar calls to the *pathname* 279 * unpins the object from the filesystem, removing the reference. 280 * If no other file descriptors or filesystem nodes refer to the 281 * same object, it will be deallocated (see NOTES). 282 * 283 * The filesystem type for the parent directory of *pathname* must 284 * be **BPF_FS_MAGIC**. 285 * 286 * Return 287 * Returns zero on success. On error, -1 is returned and *errno* 288 * is set appropriately. 289 * 290 * BPF_OBJ_GET 291 * Description 292 * Open a file descriptor for the eBPF object pinned to the 293 * specified *pathname*. 294 * 295 * Return 296 * A new file descriptor (a nonnegative integer), or -1 if an 297 * error occurred (in which case, *errno* is set appropriately). 298 * 299 * BPF_PROG_ATTACH 300 * Description 301 * Attach an eBPF program to a *target_fd* at the specified 302 * *attach_type* hook. 303 * 304 * The *attach_type* specifies the eBPF attachment point to 305 * attach the program to, and must be one of *bpf_attach_type* 306 * (see below). 307 * 308 * The *attach_bpf_fd* must be a valid file descriptor for a 309 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 310 * or sock_ops type corresponding to the specified *attach_type*. 311 * 312 * The *target_fd* must be a valid file descriptor for a kernel 313 * object which depends on the attach type of *attach_bpf_fd*: 314 * 315 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 316 * **BPF_PROG_TYPE_CGROUP_SKB**, 317 * **BPF_PROG_TYPE_CGROUP_SOCK**, 318 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 319 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 320 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 321 * **BPF_PROG_TYPE_SOCK_OPS** 322 * 323 * Control Group v2 hierarchy with the eBPF controller 324 * enabled. Requires the kernel to be compiled with 325 * **CONFIG_CGROUP_BPF**. 326 * 327 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 328 * 329 * Network namespace (eg /proc/self/ns/net). 330 * 331 * **BPF_PROG_TYPE_LIRC_MODE2** 332 * 333 * LIRC device path (eg /dev/lircN). Requires the kernel 334 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 335 * 336 * **BPF_PROG_TYPE_SK_SKB**, 337 * **BPF_PROG_TYPE_SK_MSG** 338 * 339 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 340 * 341 * Return 342 * Returns zero on success. On error, -1 is returned and *errno* 343 * is set appropriately. 344 * 345 * BPF_PROG_DETACH 346 * Description 347 * Detach the eBPF program associated with the *target_fd* at the 348 * hook specified by *attach_type*. The program must have been 349 * previously attached using **BPF_PROG_ATTACH**. 350 * 351 * Return 352 * Returns zero on success. On error, -1 is returned and *errno* 353 * is set appropriately. 354 * 355 * BPF_PROG_TEST_RUN 356 * Description 357 * Run the eBPF program associated with the *prog_fd* a *repeat* 358 * number of times against a provided program context *ctx_in* and 359 * data *data_in*, and return the modified program context 360 * *ctx_out*, *data_out* (for example, packet data), result of the 361 * execution *retval*, and *duration* of the test run. 362 * 363 * The sizes of the buffers provided as input and output 364 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 365 * be provided in the corresponding variables *ctx_size_in*, 366 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 367 * of these parameters are not provided (ie set to NULL), the 368 * corresponding size field must be zero. 369 * 370 * Some program types have particular requirements: 371 * 372 * **BPF_PROG_TYPE_SK_LOOKUP** 373 * *data_in* and *data_out* must be NULL. 374 * 375 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 376 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 377 * 378 * *ctx_out*, *data_in* and *data_out* must be NULL. 379 * *repeat* must be zero. 380 * 381 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 382 * 383 * Return 384 * Returns zero on success. On error, -1 is returned and *errno* 385 * is set appropriately. 386 * 387 * **ENOSPC** 388 * Either *data_size_out* or *ctx_size_out* is too small. 389 * **ENOTSUPP** 390 * This command is not supported by the program type of 391 * the program referred to by *prog_fd*. 392 * 393 * BPF_PROG_GET_NEXT_ID 394 * Description 395 * Fetch the next eBPF program currently loaded into the kernel. 396 * 397 * Looks for the eBPF program with an id greater than *start_id* 398 * and updates *next_id* on success. If no other eBPF programs 399 * remain with ids higher than *start_id*, returns -1 and sets 400 * *errno* to **ENOENT**. 401 * 402 * Return 403 * Returns zero on success. On error, or when no id remains, -1 404 * is returned and *errno* is set appropriately. 405 * 406 * BPF_MAP_GET_NEXT_ID 407 * Description 408 * Fetch the next eBPF map currently loaded into the kernel. 409 * 410 * Looks for the eBPF map with an id greater than *start_id* 411 * and updates *next_id* on success. If no other eBPF maps 412 * remain with ids higher than *start_id*, returns -1 and sets 413 * *errno* to **ENOENT**. 414 * 415 * Return 416 * Returns zero on success. On error, or when no id remains, -1 417 * is returned and *errno* is set appropriately. 418 * 419 * BPF_PROG_GET_FD_BY_ID 420 * Description 421 * Open a file descriptor for the eBPF program corresponding to 422 * *prog_id*. 423 * 424 * Return 425 * A new file descriptor (a nonnegative integer), or -1 if an 426 * error occurred (in which case, *errno* is set appropriately). 427 * 428 * BPF_MAP_GET_FD_BY_ID 429 * Description 430 * Open a file descriptor for the eBPF map corresponding to 431 * *map_id*. 432 * 433 * Return 434 * A new file descriptor (a nonnegative integer), or -1 if an 435 * error occurred (in which case, *errno* is set appropriately). 436 * 437 * BPF_OBJ_GET_INFO_BY_FD 438 * Description 439 * Obtain information about the eBPF object corresponding to 440 * *bpf_fd*. 441 * 442 * Populates up to *info_len* bytes of *info*, which will be in 443 * one of the following formats depending on the eBPF object type 444 * of *bpf_fd*: 445 * 446 * * **struct bpf_prog_info** 447 * * **struct bpf_map_info** 448 * * **struct bpf_btf_info** 449 * * **struct bpf_link_info** 450 * 451 * Return 452 * Returns zero on success. On error, -1 is returned and *errno* 453 * is set appropriately. 454 * 455 * BPF_PROG_QUERY 456 * Description 457 * Obtain information about eBPF programs associated with the 458 * specified *attach_type* hook. 459 * 460 * The *target_fd* must be a valid file descriptor for a kernel 461 * object which depends on the attach type of *attach_bpf_fd*: 462 * 463 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 464 * **BPF_PROG_TYPE_CGROUP_SKB**, 465 * **BPF_PROG_TYPE_CGROUP_SOCK**, 466 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 467 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 468 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 469 * **BPF_PROG_TYPE_SOCK_OPS** 470 * 471 * Control Group v2 hierarchy with the eBPF controller 472 * enabled. Requires the kernel to be compiled with 473 * **CONFIG_CGROUP_BPF**. 474 * 475 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 476 * 477 * Network namespace (eg /proc/self/ns/net). 478 * 479 * **BPF_PROG_TYPE_LIRC_MODE2** 480 * 481 * LIRC device path (eg /dev/lircN). Requires the kernel 482 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 483 * 484 * **BPF_PROG_QUERY** always fetches the number of programs 485 * attached and the *attach_flags* which were used to attach those 486 * programs. Additionally, if *prog_ids* is nonzero and the number 487 * of attached programs is less than *prog_cnt*, populates 488 * *prog_ids* with the eBPF program ids of the programs attached 489 * at *target_fd*. 490 * 491 * The following flags may alter the result: 492 * 493 * **BPF_F_QUERY_EFFECTIVE** 494 * Only return information regarding programs which are 495 * currently effective at the specified *target_fd*. 496 * 497 * Return 498 * Returns zero on success. On error, -1 is returned and *errno* 499 * is set appropriately. 500 * 501 * BPF_RAW_TRACEPOINT_OPEN 502 * Description 503 * Attach an eBPF program to a tracepoint *name* to access kernel 504 * internal arguments of the tracepoint in their raw form. 505 * 506 * The *prog_fd* must be a valid file descriptor associated with 507 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 508 * 509 * No ABI guarantees are made about the content of tracepoint 510 * arguments exposed to the corresponding eBPF program. 511 * 512 * Applying **close**\ (2) to the file descriptor returned by 513 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 514 * 515 * Return 516 * A new file descriptor (a nonnegative integer), or -1 if an 517 * error occurred (in which case, *errno* is set appropriately). 518 * 519 * BPF_BTF_LOAD 520 * Description 521 * Verify and load BPF Type Format (BTF) metadata into the kernel, 522 * returning a new file descriptor associated with the metadata. 523 * BTF is described in more detail at 524 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 525 * 526 * The *btf* parameter must point to valid memory providing 527 * *btf_size* bytes of BTF binary metadata. 528 * 529 * The returned file descriptor can be passed to other **bpf**\ () 530 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 531 * associate the BTF with those objects. 532 * 533 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 534 * parameters to specify a *btf_log_buf*, *btf_log_size* and 535 * *btf_log_level* which allow the kernel to return freeform log 536 * output regarding the BTF verification process. 537 * 538 * Return 539 * A new file descriptor (a nonnegative integer), or -1 if an 540 * error occurred (in which case, *errno* is set appropriately). 541 * 542 * BPF_BTF_GET_FD_BY_ID 543 * Description 544 * Open a file descriptor for the BPF Type Format (BTF) 545 * corresponding to *btf_id*. 546 * 547 * Return 548 * A new file descriptor (a nonnegative integer), or -1 if an 549 * error occurred (in which case, *errno* is set appropriately). 550 * 551 * BPF_TASK_FD_QUERY 552 * Description 553 * Obtain information about eBPF programs associated with the 554 * target process identified by *pid* and *fd*. 555 * 556 * If the *pid* and *fd* are associated with a tracepoint, kprobe 557 * or uprobe perf event, then the *prog_id* and *fd_type* will 558 * be populated with the eBPF program id and file descriptor type 559 * of type **bpf_task_fd_type**. If associated with a kprobe or 560 * uprobe, the *probe_offset* and *probe_addr* will also be 561 * populated. Optionally, if *buf* is provided, then up to 562 * *buf_len* bytes of *buf* will be populated with the name of 563 * the tracepoint, kprobe or uprobe. 564 * 565 * The resulting *prog_id* may be introspected in deeper detail 566 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 567 * 568 * Return 569 * Returns zero on success. On error, -1 is returned and *errno* 570 * is set appropriately. 571 * 572 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 573 * Description 574 * Look up an element with the given *key* in the map referred to 575 * by the file descriptor *fd*, and if found, delete the element. 576 * 577 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 578 * types, the *flags* argument needs to be set to 0, but for other 579 * map types, it may be specified as: 580 * 581 * **BPF_F_LOCK** 582 * Look up and delete the value of a spin-locked map 583 * without returning the lock. This must be specified if 584 * the elements contain a spinlock. 585 * 586 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 587 * implement this command as a "pop" operation, deleting the top 588 * element rather than one corresponding to *key*. 589 * The *key* and *key_len* parameters should be zeroed when 590 * issuing this operation for these map types. 591 * 592 * This command is only valid for the following map types: 593 * * **BPF_MAP_TYPE_QUEUE** 594 * * **BPF_MAP_TYPE_STACK** 595 * * **BPF_MAP_TYPE_HASH** 596 * * **BPF_MAP_TYPE_PERCPU_HASH** 597 * * **BPF_MAP_TYPE_LRU_HASH** 598 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 599 * 600 * Return 601 * Returns zero on success. On error, -1 is returned and *errno* 602 * is set appropriately. 603 * 604 * BPF_MAP_FREEZE 605 * Description 606 * Freeze the permissions of the specified map. 607 * 608 * Write permissions may be frozen by passing zero *flags*. 609 * Upon success, no future syscall invocations may alter the 610 * map state of *map_fd*. Write operations from eBPF programs 611 * are still possible for a frozen map. 612 * 613 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 614 * 615 * Return 616 * Returns zero on success. On error, -1 is returned and *errno* 617 * is set appropriately. 618 * 619 * BPF_BTF_GET_NEXT_ID 620 * Description 621 * Fetch the next BPF Type Format (BTF) object currently loaded 622 * into the kernel. 623 * 624 * Looks for the BTF object with an id greater than *start_id* 625 * and updates *next_id* on success. If no other BTF objects 626 * remain with ids higher than *start_id*, returns -1 and sets 627 * *errno* to **ENOENT**. 628 * 629 * Return 630 * Returns zero on success. On error, or when no id remains, -1 631 * is returned and *errno* is set appropriately. 632 * 633 * BPF_MAP_LOOKUP_BATCH 634 * Description 635 * Iterate and fetch multiple elements in a map. 636 * 637 * Two opaque values are used to manage batch operations, 638 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 639 * to NULL to begin the batched operation. After each subsequent 640 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 641 * *out_batch* as the *in_batch* for the next operation to 642 * continue iteration from the current point. Both *in_batch* and 643 * *out_batch* must point to memory large enough to hold a key, 644 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH, 645 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters 646 * must be at least 4 bytes wide regardless of key size. 647 * 648 * The *keys* and *values* are output parameters which must point 649 * to memory large enough to hold *count* items based on the key 650 * and value size of the map *map_fd*. The *keys* buffer must be 651 * of *key_size* * *count*. The *values* buffer must be of 652 * *value_size* * *count*. 653 * 654 * The *elem_flags* argument may be specified as one of the 655 * following: 656 * 657 * **BPF_F_LOCK** 658 * Look up the value of a spin-locked map without 659 * returning the lock. This must be specified if the 660 * elements contain a spinlock. 661 * 662 * On success, *count* elements from the map are copied into the 663 * user buffer, with the keys copied into *keys* and the values 664 * copied into the corresponding indices in *values*. 665 * 666 * If an error is returned and *errno* is not **EFAULT**, *count* 667 * is set to the number of successfully processed elements. 668 * 669 * Return 670 * Returns zero on success. On error, -1 is returned and *errno* 671 * is set appropriately. 672 * 673 * May set *errno* to **ENOSPC** to indicate that *keys* or 674 * *values* is too small to dump an entire bucket during 675 * iteration of a hash-based map type. 676 * 677 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 678 * Description 679 * Iterate and delete all elements in a map. 680 * 681 * This operation has the same behavior as 682 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 683 * 684 * * Every element that is successfully returned is also deleted 685 * from the map. This is at least *count* elements. Note that 686 * *count* is both an input and an output parameter. 687 * * Upon returning with *errno* set to **EFAULT**, up to 688 * *count* elements may be deleted without returning the keys 689 * and values of the deleted elements. 690 * 691 * Return 692 * Returns zero on success. On error, -1 is returned and *errno* 693 * is set appropriately. 694 * 695 * BPF_MAP_UPDATE_BATCH 696 * Description 697 * Update multiple elements in a map by *key*. 698 * 699 * The *keys* and *values* are input parameters which must point 700 * to memory large enough to hold *count* items based on the key 701 * and value size of the map *map_fd*. The *keys* buffer must be 702 * of *key_size* * *count*. The *values* buffer must be of 703 * *value_size* * *count*. 704 * 705 * Each element specified in *keys* is sequentially updated to the 706 * value in the corresponding index in *values*. The *in_batch* 707 * and *out_batch* parameters are ignored and should be zeroed. 708 * 709 * The *elem_flags* argument should be specified as one of the 710 * following: 711 * 712 * **BPF_ANY** 713 * Create new elements or update a existing elements. 714 * **BPF_NOEXIST** 715 * Create new elements only if they do not exist. 716 * **BPF_EXIST** 717 * Update existing elements. 718 * **BPF_F_LOCK** 719 * Update spin_lock-ed map elements. This must be 720 * specified if the map value contains a spinlock. 721 * 722 * On success, *count* elements from the map are updated. 723 * 724 * If an error is returned and *errno* is not **EFAULT**, *count* 725 * is set to the number of successfully processed elements. 726 * 727 * Return 728 * Returns zero on success. On error, -1 is returned and *errno* 729 * is set appropriately. 730 * 731 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 732 * **E2BIG**. **E2BIG** indicates that the number of elements in 733 * the map reached the *max_entries* limit specified at map 734 * creation time. 735 * 736 * May set *errno* to one of the following error codes under 737 * specific circumstances: 738 * 739 * **EEXIST** 740 * If *flags* specifies **BPF_NOEXIST** and the element 741 * with *key* already exists in the map. 742 * **ENOENT** 743 * If *flags* specifies **BPF_EXIST** and the element with 744 * *key* does not exist in the map. 745 * 746 * BPF_MAP_DELETE_BATCH 747 * Description 748 * Delete multiple elements in a map by *key*. 749 * 750 * The *keys* parameter is an input parameter which must point 751 * to memory large enough to hold *count* items based on the key 752 * size of the map *map_fd*, that is, *key_size* * *count*. 753 * 754 * Each element specified in *keys* is sequentially deleted. The 755 * *in_batch*, *out_batch*, and *values* parameters are ignored 756 * and should be zeroed. 757 * 758 * The *elem_flags* argument may be specified as one of the 759 * following: 760 * 761 * **BPF_F_LOCK** 762 * Look up the value of a spin-locked map without 763 * returning the lock. This must be specified if the 764 * elements contain a spinlock. 765 * 766 * On success, *count* elements from the map are updated. 767 * 768 * If an error is returned and *errno* is not **EFAULT**, *count* 769 * is set to the number of successfully processed elements. If 770 * *errno* is **EFAULT**, up to *count* elements may be been 771 * deleted. 772 * 773 * Return 774 * Returns zero on success. On error, -1 is returned and *errno* 775 * is set appropriately. 776 * 777 * BPF_LINK_CREATE 778 * Description 779 * Attach an eBPF program to a *target_fd* at the specified 780 * *attach_type* hook and return a file descriptor handle for 781 * managing the link. 782 * 783 * Return 784 * A new file descriptor (a nonnegative integer), or -1 if an 785 * error occurred (in which case, *errno* is set appropriately). 786 * 787 * BPF_LINK_UPDATE 788 * Description 789 * Update the eBPF program in the specified *link_fd* to 790 * *new_prog_fd*. 791 * 792 * Return 793 * Returns zero on success. On error, -1 is returned and *errno* 794 * is set appropriately. 795 * 796 * BPF_LINK_GET_FD_BY_ID 797 * Description 798 * Open a file descriptor for the eBPF Link corresponding to 799 * *link_id*. 800 * 801 * Return 802 * A new file descriptor (a nonnegative integer), or -1 if an 803 * error occurred (in which case, *errno* is set appropriately). 804 * 805 * BPF_LINK_GET_NEXT_ID 806 * Description 807 * Fetch the next eBPF link currently loaded into the kernel. 808 * 809 * Looks for the eBPF link with an id greater than *start_id* 810 * and updates *next_id* on success. If no other eBPF links 811 * remain with ids higher than *start_id*, returns -1 and sets 812 * *errno* to **ENOENT**. 813 * 814 * Return 815 * Returns zero on success. On error, or when no id remains, -1 816 * is returned and *errno* is set appropriately. 817 * 818 * BPF_ENABLE_STATS 819 * Description 820 * Enable eBPF runtime statistics gathering. 821 * 822 * Runtime statistics gathering for the eBPF runtime is disabled 823 * by default to minimize the corresponding performance overhead. 824 * This command enables statistics globally. 825 * 826 * Multiple programs may independently enable statistics. 827 * After gathering the desired statistics, eBPF runtime statistics 828 * may be disabled again by calling **close**\ (2) for the file 829 * descriptor returned by this function. Statistics will only be 830 * disabled system-wide when all outstanding file descriptors 831 * returned by prior calls for this subcommand are closed. 832 * 833 * Return 834 * A new file descriptor (a nonnegative integer), or -1 if an 835 * error occurred (in which case, *errno* is set appropriately). 836 * 837 * BPF_ITER_CREATE 838 * Description 839 * Create an iterator on top of the specified *link_fd* (as 840 * previously created using **BPF_LINK_CREATE**) and return a 841 * file descriptor that can be used to trigger the iteration. 842 * 843 * If the resulting file descriptor is pinned to the filesystem 844 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 845 * for that path will trigger the iterator to read kernel state 846 * using the eBPF program attached to *link_fd*. 847 * 848 * Return 849 * A new file descriptor (a nonnegative integer), or -1 if an 850 * error occurred (in which case, *errno* is set appropriately). 851 * 852 * BPF_LINK_DETACH 853 * Description 854 * Forcefully detach the specified *link_fd* from its 855 * corresponding attachment point. 856 * 857 * Return 858 * Returns zero on success. On error, -1 is returned and *errno* 859 * is set appropriately. 860 * 861 * BPF_PROG_BIND_MAP 862 * Description 863 * Bind a map to the lifetime of an eBPF program. 864 * 865 * The map identified by *map_fd* is bound to the program 866 * identified by *prog_fd* and only released when *prog_fd* is 867 * released. This may be used in cases where metadata should be 868 * associated with a program which otherwise does not contain any 869 * references to the map (for example, embedded in the eBPF 870 * program instructions). 871 * 872 * Return 873 * Returns zero on success. On error, -1 is returned and *errno* 874 * is set appropriately. 875 * 876 * BPF_TOKEN_CREATE 877 * Description 878 * Create BPF token with embedded information about what 879 * BPF-related functionality it allows: 880 * - a set of allowed bpf() syscall commands; 881 * - a set of allowed BPF map types to be created with 882 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 883 * - a set of allowed BPF program types and BPF program attach 884 * types to be loaded with BPF_PROG_LOAD command, if 885 * BPF_PROG_LOAD itself is allowed. 886 * 887 * BPF token is created (derived) from an instance of BPF FS, 888 * assuming it has necessary delegation mount options specified. 889 * This BPF token can be passed as an extra parameter to various 890 * bpf() syscall commands to grant BPF subsystem functionality to 891 * unprivileged processes. 892 * 893 * When created, BPF token is "associated" with the owning 894 * user namespace of BPF FS instance (super block) that it was 895 * derived from, and subsequent BPF operations performed with 896 * BPF token would be performing capabilities checks (i.e., 897 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 898 * that user namespace. Without BPF token, such capabilities 899 * have to be granted in init user namespace, making bpf() 900 * syscall incompatible with user namespace, for the most part. 901 * 902 * Return 903 * A new file descriptor (a nonnegative integer), or -1 if an 904 * error occurred (in which case, *errno* is set appropriately). 905 * 906 * NOTES 907 * eBPF objects (maps and programs) can be shared between processes. 908 * 909 * * After **fork**\ (2), the child inherits file descriptors 910 * referring to the same eBPF objects. 911 * * File descriptors referring to eBPF objects can be transferred over 912 * **unix**\ (7) domain sockets. 913 * * File descriptors referring to eBPF objects can be duplicated in the 914 * usual way, using **dup**\ (2) and similar calls. 915 * * File descriptors referring to eBPF objects can be pinned to the 916 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 917 * 918 * An eBPF object is deallocated only after all file descriptors referring 919 * to the object have been closed and no references remain pinned to the 920 * filesystem or attached (for example, bound to a program or device). 921 */ 922 enum bpf_cmd { 923 BPF_MAP_CREATE, 924 BPF_MAP_LOOKUP_ELEM, 925 BPF_MAP_UPDATE_ELEM, 926 BPF_MAP_DELETE_ELEM, 927 BPF_MAP_GET_NEXT_KEY, 928 BPF_PROG_LOAD, 929 BPF_OBJ_PIN, 930 BPF_OBJ_GET, 931 BPF_PROG_ATTACH, 932 BPF_PROG_DETACH, 933 BPF_PROG_TEST_RUN, 934 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 935 BPF_PROG_GET_NEXT_ID, 936 BPF_MAP_GET_NEXT_ID, 937 BPF_PROG_GET_FD_BY_ID, 938 BPF_MAP_GET_FD_BY_ID, 939 BPF_OBJ_GET_INFO_BY_FD, 940 BPF_PROG_QUERY, 941 BPF_RAW_TRACEPOINT_OPEN, 942 BPF_BTF_LOAD, 943 BPF_BTF_GET_FD_BY_ID, 944 BPF_TASK_FD_QUERY, 945 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 946 BPF_MAP_FREEZE, 947 BPF_BTF_GET_NEXT_ID, 948 BPF_MAP_LOOKUP_BATCH, 949 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 950 BPF_MAP_UPDATE_BATCH, 951 BPF_MAP_DELETE_BATCH, 952 BPF_LINK_CREATE, 953 BPF_LINK_UPDATE, 954 BPF_LINK_GET_FD_BY_ID, 955 BPF_LINK_GET_NEXT_ID, 956 BPF_ENABLE_STATS, 957 BPF_ITER_CREATE, 958 BPF_LINK_DETACH, 959 BPF_PROG_BIND_MAP, 960 BPF_TOKEN_CREATE, 961 __MAX_BPF_CMD, 962 }; 963 964 enum bpf_map_type { 965 BPF_MAP_TYPE_UNSPEC, 966 BPF_MAP_TYPE_HASH, 967 BPF_MAP_TYPE_ARRAY, 968 BPF_MAP_TYPE_PROG_ARRAY, 969 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 970 BPF_MAP_TYPE_PERCPU_HASH, 971 BPF_MAP_TYPE_PERCPU_ARRAY, 972 BPF_MAP_TYPE_STACK_TRACE, 973 BPF_MAP_TYPE_CGROUP_ARRAY, 974 BPF_MAP_TYPE_LRU_HASH, 975 BPF_MAP_TYPE_LRU_PERCPU_HASH, 976 BPF_MAP_TYPE_LPM_TRIE, 977 BPF_MAP_TYPE_ARRAY_OF_MAPS, 978 BPF_MAP_TYPE_HASH_OF_MAPS, 979 BPF_MAP_TYPE_DEVMAP, 980 BPF_MAP_TYPE_SOCKMAP, 981 BPF_MAP_TYPE_CPUMAP, 982 BPF_MAP_TYPE_XSKMAP, 983 BPF_MAP_TYPE_SOCKHASH, 984 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 985 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 986 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 987 * both cgroup-attached and other progs and supports all functionality 988 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 989 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 990 */ 991 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 992 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 993 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 994 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 995 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 996 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 997 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 998 * deprecated. 999 */ 1000 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 1001 BPF_MAP_TYPE_QUEUE, 1002 BPF_MAP_TYPE_STACK, 1003 BPF_MAP_TYPE_SK_STORAGE, 1004 BPF_MAP_TYPE_DEVMAP_HASH, 1005 BPF_MAP_TYPE_STRUCT_OPS, 1006 BPF_MAP_TYPE_RINGBUF, 1007 BPF_MAP_TYPE_INODE_STORAGE, 1008 BPF_MAP_TYPE_TASK_STORAGE, 1009 BPF_MAP_TYPE_BLOOM_FILTER, 1010 BPF_MAP_TYPE_USER_RINGBUF, 1011 BPF_MAP_TYPE_CGRP_STORAGE, 1012 BPF_MAP_TYPE_ARENA, 1013 __MAX_BPF_MAP_TYPE 1014 }; 1015 1016 /* Note that tracing related programs such as 1017 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1018 * are not subject to a stable API since kernel internal data 1019 * structures can change from release to release and may 1020 * therefore break existing tracing BPF programs. Tracing BPF 1021 * programs correspond to /a/ specific kernel which is to be 1022 * analyzed, and not /a/ specific kernel /and/ all future ones. 1023 */ 1024 enum bpf_prog_type { 1025 BPF_PROG_TYPE_UNSPEC, 1026 BPF_PROG_TYPE_SOCKET_FILTER, 1027 BPF_PROG_TYPE_KPROBE, 1028 BPF_PROG_TYPE_SCHED_CLS, 1029 BPF_PROG_TYPE_SCHED_ACT, 1030 BPF_PROG_TYPE_TRACEPOINT, 1031 BPF_PROG_TYPE_XDP, 1032 BPF_PROG_TYPE_PERF_EVENT, 1033 BPF_PROG_TYPE_CGROUP_SKB, 1034 BPF_PROG_TYPE_CGROUP_SOCK, 1035 BPF_PROG_TYPE_LWT_IN, 1036 BPF_PROG_TYPE_LWT_OUT, 1037 BPF_PROG_TYPE_LWT_XMIT, 1038 BPF_PROG_TYPE_SOCK_OPS, 1039 BPF_PROG_TYPE_SK_SKB, 1040 BPF_PROG_TYPE_CGROUP_DEVICE, 1041 BPF_PROG_TYPE_SK_MSG, 1042 BPF_PROG_TYPE_RAW_TRACEPOINT, 1043 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1044 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1045 BPF_PROG_TYPE_LIRC_MODE2, 1046 BPF_PROG_TYPE_SK_REUSEPORT, 1047 BPF_PROG_TYPE_FLOW_DISSECTOR, 1048 BPF_PROG_TYPE_CGROUP_SYSCTL, 1049 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1050 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1051 BPF_PROG_TYPE_TRACING, 1052 BPF_PROG_TYPE_STRUCT_OPS, 1053 BPF_PROG_TYPE_EXT, 1054 BPF_PROG_TYPE_LSM, 1055 BPF_PROG_TYPE_SK_LOOKUP, 1056 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1057 BPF_PROG_TYPE_NETFILTER, 1058 __MAX_BPF_PROG_TYPE 1059 }; 1060 1061 enum bpf_attach_type { 1062 BPF_CGROUP_INET_INGRESS, 1063 BPF_CGROUP_INET_EGRESS, 1064 BPF_CGROUP_INET_SOCK_CREATE, 1065 BPF_CGROUP_SOCK_OPS, 1066 BPF_SK_SKB_STREAM_PARSER, 1067 BPF_SK_SKB_STREAM_VERDICT, 1068 BPF_CGROUP_DEVICE, 1069 BPF_SK_MSG_VERDICT, 1070 BPF_CGROUP_INET4_BIND, 1071 BPF_CGROUP_INET6_BIND, 1072 BPF_CGROUP_INET4_CONNECT, 1073 BPF_CGROUP_INET6_CONNECT, 1074 BPF_CGROUP_INET4_POST_BIND, 1075 BPF_CGROUP_INET6_POST_BIND, 1076 BPF_CGROUP_UDP4_SENDMSG, 1077 BPF_CGROUP_UDP6_SENDMSG, 1078 BPF_LIRC_MODE2, 1079 BPF_FLOW_DISSECTOR, 1080 BPF_CGROUP_SYSCTL, 1081 BPF_CGROUP_UDP4_RECVMSG, 1082 BPF_CGROUP_UDP6_RECVMSG, 1083 BPF_CGROUP_GETSOCKOPT, 1084 BPF_CGROUP_SETSOCKOPT, 1085 BPF_TRACE_RAW_TP, 1086 BPF_TRACE_FENTRY, 1087 BPF_TRACE_FEXIT, 1088 BPF_MODIFY_RETURN, 1089 BPF_LSM_MAC, 1090 BPF_TRACE_ITER, 1091 BPF_CGROUP_INET4_GETPEERNAME, 1092 BPF_CGROUP_INET6_GETPEERNAME, 1093 BPF_CGROUP_INET4_GETSOCKNAME, 1094 BPF_CGROUP_INET6_GETSOCKNAME, 1095 BPF_XDP_DEVMAP, 1096 BPF_CGROUP_INET_SOCK_RELEASE, 1097 BPF_XDP_CPUMAP, 1098 BPF_SK_LOOKUP, 1099 BPF_XDP, 1100 BPF_SK_SKB_VERDICT, 1101 BPF_SK_REUSEPORT_SELECT, 1102 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1103 BPF_PERF_EVENT, 1104 BPF_TRACE_KPROBE_MULTI, 1105 BPF_LSM_CGROUP, 1106 BPF_STRUCT_OPS, 1107 BPF_NETFILTER, 1108 BPF_TCX_INGRESS, 1109 BPF_TCX_EGRESS, 1110 BPF_TRACE_UPROBE_MULTI, 1111 BPF_CGROUP_UNIX_CONNECT, 1112 BPF_CGROUP_UNIX_SENDMSG, 1113 BPF_CGROUP_UNIX_RECVMSG, 1114 BPF_CGROUP_UNIX_GETPEERNAME, 1115 BPF_CGROUP_UNIX_GETSOCKNAME, 1116 BPF_NETKIT_PRIMARY, 1117 BPF_NETKIT_PEER, 1118 __MAX_BPF_ATTACH_TYPE 1119 }; 1120 1121 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1122 1123 enum bpf_link_type { 1124 BPF_LINK_TYPE_UNSPEC = 0, 1125 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1126 BPF_LINK_TYPE_TRACING = 2, 1127 BPF_LINK_TYPE_CGROUP = 3, 1128 BPF_LINK_TYPE_ITER = 4, 1129 BPF_LINK_TYPE_NETNS = 5, 1130 BPF_LINK_TYPE_XDP = 6, 1131 BPF_LINK_TYPE_PERF_EVENT = 7, 1132 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1133 BPF_LINK_TYPE_STRUCT_OPS = 9, 1134 BPF_LINK_TYPE_NETFILTER = 10, 1135 BPF_LINK_TYPE_TCX = 11, 1136 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1137 BPF_LINK_TYPE_NETKIT = 13, 1138 __MAX_BPF_LINK_TYPE, 1139 }; 1140 1141 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1142 1143 enum bpf_perf_event_type { 1144 BPF_PERF_EVENT_UNSPEC = 0, 1145 BPF_PERF_EVENT_UPROBE = 1, 1146 BPF_PERF_EVENT_URETPROBE = 2, 1147 BPF_PERF_EVENT_KPROBE = 3, 1148 BPF_PERF_EVENT_KRETPROBE = 4, 1149 BPF_PERF_EVENT_TRACEPOINT = 5, 1150 BPF_PERF_EVENT_EVENT = 6, 1151 }; 1152 1153 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1154 * 1155 * NONE(default): No further bpf programs allowed in the subtree. 1156 * 1157 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1158 * the program in this cgroup yields to sub-cgroup program. 1159 * 1160 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1161 * that cgroup program gets run in addition to the program in this cgroup. 1162 * 1163 * Only one program is allowed to be attached to a cgroup with 1164 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1165 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1166 * release old program and attach the new one. Attach flags has to match. 1167 * 1168 * Multiple programs are allowed to be attached to a cgroup with 1169 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1170 * (those that were attached first, run first) 1171 * The programs of sub-cgroup are executed first, then programs of 1172 * this cgroup and then programs of parent cgroup. 1173 * When children program makes decision (like picking TCP CA or sock bind) 1174 * parent program has a chance to override it. 1175 * 1176 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1177 * programs for a cgroup. Though it's possible to replace an old program at 1178 * any position by also specifying BPF_F_REPLACE flag and position itself in 1179 * replace_bpf_fd attribute. Old program at this position will be released. 1180 * 1181 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1182 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1183 * Ex1: 1184 * cgrp1 (MULTI progs A, B) -> 1185 * cgrp2 (OVERRIDE prog C) -> 1186 * cgrp3 (MULTI prog D) -> 1187 * cgrp4 (OVERRIDE prog E) -> 1188 * cgrp5 (NONE prog F) 1189 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1190 * if prog F is detached, the execution is E,D,A,B 1191 * if prog F and D are detached, the execution is E,A,B 1192 * if prog F, E and D are detached, the execution is C,A,B 1193 * 1194 * All eligible programs are executed regardless of return code from 1195 * earlier programs. 1196 */ 1197 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1198 #define BPF_F_ALLOW_MULTI (1U << 1) 1199 /* Generic attachment flags. */ 1200 #define BPF_F_REPLACE (1U << 2) 1201 #define BPF_F_BEFORE (1U << 3) 1202 #define BPF_F_AFTER (1U << 4) 1203 #define BPF_F_ID (1U << 5) 1204 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1205 1206 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1207 * verifier will perform strict alignment checking as if the kernel 1208 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1209 * and NET_IP_ALIGN defined to 2. 1210 */ 1211 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1212 1213 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1214 * verifier will allow any alignment whatsoever. On platforms 1215 * with strict alignment requirements for loads ands stores (such 1216 * as sparc and mips) the verifier validates that all loads and 1217 * stores provably follow this requirement. This flag turns that 1218 * checking and enforcement off. 1219 * 1220 * It is mostly used for testing when we want to validate the 1221 * context and memory access aspects of the verifier, but because 1222 * of an unaligned access the alignment check would trigger before 1223 * the one we are interested in. 1224 */ 1225 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1226 1227 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1228 * Verifier does sub-register def/use analysis and identifies instructions whose 1229 * def only matters for low 32-bit, high 32-bit is never referenced later 1230 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1231 * that it is safe to ignore clearing high 32-bit for these instructions. This 1232 * saves some back-ends a lot of code-gen. However such optimization is not 1233 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1234 * hence hasn't used verifier's analysis result. But, we really want to have a 1235 * way to be able to verify the correctness of the described optimization on 1236 * x86_64 on which testsuites are frequently exercised. 1237 * 1238 * So, this flag is introduced. Once it is set, verifier will randomize high 1239 * 32-bit for those instructions who has been identified as safe to ignore them. 1240 * Then, if verifier is not doing correct analysis, such randomization will 1241 * regress tests to expose bugs. 1242 */ 1243 #define BPF_F_TEST_RND_HI32 (1U << 2) 1244 1245 /* The verifier internal test flag. Behavior is undefined */ 1246 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1247 1248 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1249 * restrict map and helper usage for such programs. Sleepable BPF programs can 1250 * only be attached to hooks where kernel execution context allows sleeping. 1251 * Such programs are allowed to use helpers that may sleep like 1252 * bpf_copy_from_user(). 1253 */ 1254 #define BPF_F_SLEEPABLE (1U << 4) 1255 1256 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1257 * fully support xdp frags. 1258 */ 1259 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1260 1261 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1262 * program becomes device-bound but can access XDP metadata. 1263 */ 1264 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1265 1266 /* The verifier internal test flag. Behavior is undefined */ 1267 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1268 1269 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1270 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1271 */ 1272 enum { 1273 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1274 }; 1275 1276 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1277 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1278 */ 1279 enum { 1280 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1281 }; 1282 1283 /* link_create.netfilter.flags used in LINK_CREATE command for 1284 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1285 */ 1286 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1287 1288 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1289 * the following extensions: 1290 * 1291 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1292 * insn[0].imm: map fd or fd_idx 1293 * insn[1].imm: 0 1294 * insn[0].off: 0 1295 * insn[1].off: 0 1296 * ldimm64 rewrite: address of map 1297 * verifier type: CONST_PTR_TO_MAP 1298 */ 1299 #define BPF_PSEUDO_MAP_FD 1 1300 #define BPF_PSEUDO_MAP_IDX 5 1301 1302 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1303 * insn[0].imm: map fd or fd_idx 1304 * insn[1].imm: offset into value 1305 * insn[0].off: 0 1306 * insn[1].off: 0 1307 * ldimm64 rewrite: address of map[0]+offset 1308 * verifier type: PTR_TO_MAP_VALUE 1309 */ 1310 #define BPF_PSEUDO_MAP_VALUE 2 1311 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1312 1313 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1314 * insn[0].imm: kernel btd id of VAR 1315 * insn[1].imm: 0 1316 * insn[0].off: 0 1317 * insn[1].off: 0 1318 * ldimm64 rewrite: address of the kernel variable 1319 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1320 * is struct/union. 1321 */ 1322 #define BPF_PSEUDO_BTF_ID 3 1323 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1324 * insn[0].imm: insn offset to the func 1325 * insn[1].imm: 0 1326 * insn[0].off: 0 1327 * insn[1].off: 0 1328 * ldimm64 rewrite: address of the function 1329 * verifier type: PTR_TO_FUNC. 1330 */ 1331 #define BPF_PSEUDO_FUNC 4 1332 1333 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1334 * offset to another bpf function 1335 */ 1336 #define BPF_PSEUDO_CALL 1 1337 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1338 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1339 */ 1340 #define BPF_PSEUDO_KFUNC_CALL 2 1341 1342 enum bpf_addr_space_cast { 1343 BPF_ADDR_SPACE_CAST = 1, 1344 }; 1345 1346 /* flags for BPF_MAP_UPDATE_ELEM command */ 1347 enum { 1348 BPF_ANY = 0, /* create new element or update existing */ 1349 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1350 BPF_EXIST = 2, /* update existing element */ 1351 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1352 }; 1353 1354 /* flags for BPF_MAP_CREATE command */ 1355 enum { 1356 BPF_F_NO_PREALLOC = (1U << 0), 1357 /* Instead of having one common LRU list in the 1358 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1359 * which can scale and perform better. 1360 * Note, the LRU nodes (including free nodes) cannot be moved 1361 * across different LRU lists. 1362 */ 1363 BPF_F_NO_COMMON_LRU = (1U << 1), 1364 /* Specify numa node during map creation */ 1365 BPF_F_NUMA_NODE = (1U << 2), 1366 1367 /* Flags for accessing BPF object from syscall side. */ 1368 BPF_F_RDONLY = (1U << 3), 1369 BPF_F_WRONLY = (1U << 4), 1370 1371 /* Flag for stack_map, store build_id+offset instead of pointer */ 1372 BPF_F_STACK_BUILD_ID = (1U << 5), 1373 1374 /* Zero-initialize hash function seed. This should only be used for testing. */ 1375 BPF_F_ZERO_SEED = (1U << 6), 1376 1377 /* Flags for accessing BPF object from program side. */ 1378 BPF_F_RDONLY_PROG = (1U << 7), 1379 BPF_F_WRONLY_PROG = (1U << 8), 1380 1381 /* Clone map from listener for newly accepted socket */ 1382 BPF_F_CLONE = (1U << 9), 1383 1384 /* Enable memory-mapping BPF map */ 1385 BPF_F_MMAPABLE = (1U << 10), 1386 1387 /* Share perf_event among processes */ 1388 BPF_F_PRESERVE_ELEMS = (1U << 11), 1389 1390 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1391 BPF_F_INNER_MAP = (1U << 12), 1392 1393 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1394 BPF_F_LINK = (1U << 13), 1395 1396 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1397 BPF_F_PATH_FD = (1U << 14), 1398 1399 /* Flag for value_type_btf_obj_fd, the fd is available */ 1400 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1401 1402 /* BPF token FD is passed in a corresponding command's token_fd field */ 1403 BPF_F_TOKEN_FD = (1U << 16), 1404 1405 /* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */ 1406 BPF_F_SEGV_ON_FAULT = (1U << 17), 1407 1408 /* Do not translate kernel bpf_arena pointers to user pointers */ 1409 BPF_F_NO_USER_CONV = (1U << 18), 1410 }; 1411 1412 /* Flags for BPF_PROG_QUERY. */ 1413 1414 /* Query effective (directly attached + inherited from ancestor cgroups) 1415 * programs that will be executed for events within a cgroup. 1416 * attach_flags with this flag are always returned 0. 1417 */ 1418 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1419 1420 /* Flags for BPF_PROG_TEST_RUN */ 1421 1422 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1423 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1424 /* If set, XDP frames will be transmitted after processing */ 1425 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1426 1427 /* type for BPF_ENABLE_STATS */ 1428 enum bpf_stats_type { 1429 /* enabled run_time_ns and run_cnt */ 1430 BPF_STATS_RUN_TIME = 0, 1431 }; 1432 1433 enum bpf_stack_build_id_status { 1434 /* user space need an empty entry to identify end of a trace */ 1435 BPF_STACK_BUILD_ID_EMPTY = 0, 1436 /* with valid build_id and offset */ 1437 BPF_STACK_BUILD_ID_VALID = 1, 1438 /* couldn't get build_id, fallback to ip */ 1439 BPF_STACK_BUILD_ID_IP = 2, 1440 }; 1441 1442 #define BPF_BUILD_ID_SIZE 20 1443 struct bpf_stack_build_id { 1444 __s32 status; 1445 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1446 union { 1447 __u64 offset; 1448 __u64 ip; 1449 }; 1450 }; 1451 1452 #define BPF_OBJ_NAME_LEN 16U 1453 1454 union bpf_attr { 1455 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1456 __u32 map_type; /* one of enum bpf_map_type */ 1457 __u32 key_size; /* size of key in bytes */ 1458 __u32 value_size; /* size of value in bytes */ 1459 __u32 max_entries; /* max number of entries in a map */ 1460 __u32 map_flags; /* BPF_MAP_CREATE related 1461 * flags defined above. 1462 */ 1463 __u32 inner_map_fd; /* fd pointing to the inner map */ 1464 __u32 numa_node; /* numa node (effective only if 1465 * BPF_F_NUMA_NODE is set). 1466 */ 1467 char map_name[BPF_OBJ_NAME_LEN]; 1468 __u32 map_ifindex; /* ifindex of netdev to create on */ 1469 __u32 btf_fd; /* fd pointing to a BTF type data */ 1470 __u32 btf_key_type_id; /* BTF type_id of the key */ 1471 __u32 btf_value_type_id; /* BTF type_id of the value */ 1472 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1473 * struct stored as the 1474 * map value 1475 */ 1476 /* Any per-map-type extra fields 1477 * 1478 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1479 * number of hash functions (if 0, the bloom filter will default 1480 * to using 5 hash functions). 1481 * 1482 * BPF_MAP_TYPE_ARENA - contains the address where user space 1483 * is going to mmap() the arena. It has to be page aligned. 1484 */ 1485 __u64 map_extra; 1486 1487 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1488 * type data for 1489 * btf_vmlinux_value_type_id. 1490 */ 1491 /* BPF token FD to use with BPF_MAP_CREATE operation. 1492 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1493 */ 1494 __s32 map_token_fd; 1495 }; 1496 1497 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1498 __u32 map_fd; 1499 __aligned_u64 key; 1500 union { 1501 __aligned_u64 value; 1502 __aligned_u64 next_key; 1503 }; 1504 __u64 flags; 1505 }; 1506 1507 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1508 __aligned_u64 in_batch; /* start batch, 1509 * NULL to start from beginning 1510 */ 1511 __aligned_u64 out_batch; /* output: next start batch */ 1512 __aligned_u64 keys; 1513 __aligned_u64 values; 1514 __u32 count; /* input/output: 1515 * input: # of key/value 1516 * elements 1517 * output: # of filled elements 1518 */ 1519 __u32 map_fd; 1520 __u64 elem_flags; 1521 __u64 flags; 1522 } batch; 1523 1524 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1525 __u32 prog_type; /* one of enum bpf_prog_type */ 1526 __u32 insn_cnt; 1527 __aligned_u64 insns; 1528 __aligned_u64 license; 1529 __u32 log_level; /* verbosity level of verifier */ 1530 __u32 log_size; /* size of user buffer */ 1531 __aligned_u64 log_buf; /* user supplied buffer */ 1532 __u32 kern_version; /* not used */ 1533 __u32 prog_flags; 1534 char prog_name[BPF_OBJ_NAME_LEN]; 1535 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1536 /* For some prog types expected attach type must be known at 1537 * load time to verify attach type specific parts of prog 1538 * (context accesses, allowed helpers, etc). 1539 */ 1540 __u32 expected_attach_type; 1541 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1542 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1543 __aligned_u64 func_info; /* func info */ 1544 __u32 func_info_cnt; /* number of bpf_func_info records */ 1545 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1546 __aligned_u64 line_info; /* line info */ 1547 __u32 line_info_cnt; /* number of bpf_line_info records */ 1548 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1549 union { 1550 /* valid prog_fd to attach to bpf prog */ 1551 __u32 attach_prog_fd; 1552 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1553 __u32 attach_btf_obj_fd; 1554 }; 1555 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1556 __aligned_u64 fd_array; /* array of FDs */ 1557 __aligned_u64 core_relos; 1558 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1559 /* output: actual total log contents size (including termintaing zero). 1560 * It could be both larger than original log_size (if log was 1561 * truncated), or smaller (if log buffer wasn't filled completely). 1562 */ 1563 __u32 log_true_size; 1564 /* BPF token FD to use with BPF_PROG_LOAD operation. 1565 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1566 */ 1567 __s32 prog_token_fd; 1568 }; 1569 1570 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1571 __aligned_u64 pathname; 1572 __u32 bpf_fd; 1573 __u32 file_flags; 1574 /* Same as dirfd in openat() syscall; see openat(2) 1575 * manpage for details of path FD and pathname semantics; 1576 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1577 * file_flags field, otherwise it should be set to zero; 1578 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1579 */ 1580 __s32 path_fd; 1581 }; 1582 1583 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1584 union { 1585 __u32 target_fd; /* target object to attach to or ... */ 1586 __u32 target_ifindex; /* target ifindex */ 1587 }; 1588 __u32 attach_bpf_fd; 1589 __u32 attach_type; 1590 __u32 attach_flags; 1591 __u32 replace_bpf_fd; 1592 union { 1593 __u32 relative_fd; 1594 __u32 relative_id; 1595 }; 1596 __u64 expected_revision; 1597 }; 1598 1599 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1600 __u32 prog_fd; 1601 __u32 retval; 1602 __u32 data_size_in; /* input: len of data_in */ 1603 __u32 data_size_out; /* input/output: len of data_out 1604 * returns ENOSPC if data_out 1605 * is too small. 1606 */ 1607 __aligned_u64 data_in; 1608 __aligned_u64 data_out; 1609 __u32 repeat; 1610 __u32 duration; 1611 __u32 ctx_size_in; /* input: len of ctx_in */ 1612 __u32 ctx_size_out; /* input/output: len of ctx_out 1613 * returns ENOSPC if ctx_out 1614 * is too small. 1615 */ 1616 __aligned_u64 ctx_in; 1617 __aligned_u64 ctx_out; 1618 __u32 flags; 1619 __u32 cpu; 1620 __u32 batch_size; 1621 } test; 1622 1623 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1624 union { 1625 __u32 start_id; 1626 __u32 prog_id; 1627 __u32 map_id; 1628 __u32 btf_id; 1629 __u32 link_id; 1630 }; 1631 __u32 next_id; 1632 __u32 open_flags; 1633 }; 1634 1635 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1636 __u32 bpf_fd; 1637 __u32 info_len; 1638 __aligned_u64 info; 1639 } info; 1640 1641 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1642 union { 1643 __u32 target_fd; /* target object to query or ... */ 1644 __u32 target_ifindex; /* target ifindex */ 1645 }; 1646 __u32 attach_type; 1647 __u32 query_flags; 1648 __u32 attach_flags; 1649 __aligned_u64 prog_ids; 1650 union { 1651 __u32 prog_cnt; 1652 __u32 count; 1653 }; 1654 __u32 :32; 1655 /* output: per-program attach_flags. 1656 * not allowed to be set during effective query. 1657 */ 1658 __aligned_u64 prog_attach_flags; 1659 __aligned_u64 link_ids; 1660 __aligned_u64 link_attach_flags; 1661 __u64 revision; 1662 } query; 1663 1664 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1665 __u64 name; 1666 __u32 prog_fd; 1667 __u32 :32; 1668 __aligned_u64 cookie; 1669 } raw_tracepoint; 1670 1671 struct { /* anonymous struct for BPF_BTF_LOAD */ 1672 __aligned_u64 btf; 1673 __aligned_u64 btf_log_buf; 1674 __u32 btf_size; 1675 __u32 btf_log_size; 1676 __u32 btf_log_level; 1677 /* output: actual total log contents size (including termintaing zero). 1678 * It could be both larger than original log_size (if log was 1679 * truncated), or smaller (if log buffer wasn't filled completely). 1680 */ 1681 __u32 btf_log_true_size; 1682 __u32 btf_flags; 1683 /* BPF token FD to use with BPF_BTF_LOAD operation. 1684 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1685 */ 1686 __s32 btf_token_fd; 1687 }; 1688 1689 struct { 1690 __u32 pid; /* input: pid */ 1691 __u32 fd; /* input: fd */ 1692 __u32 flags; /* input: flags */ 1693 __u32 buf_len; /* input/output: buf len */ 1694 __aligned_u64 buf; /* input/output: 1695 * tp_name for tracepoint 1696 * symbol for kprobe 1697 * filename for uprobe 1698 */ 1699 __u32 prog_id; /* output: prod_id */ 1700 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1701 __u64 probe_offset; /* output: probe_offset */ 1702 __u64 probe_addr; /* output: probe_addr */ 1703 } task_fd_query; 1704 1705 struct { /* struct used by BPF_LINK_CREATE command */ 1706 union { 1707 __u32 prog_fd; /* eBPF program to attach */ 1708 __u32 map_fd; /* struct_ops to attach */ 1709 }; 1710 union { 1711 __u32 target_fd; /* target object to attach to or ... */ 1712 __u32 target_ifindex; /* target ifindex */ 1713 }; 1714 __u32 attach_type; /* attach type */ 1715 __u32 flags; /* extra flags */ 1716 union { 1717 __u32 target_btf_id; /* btf_id of target to attach to */ 1718 struct { 1719 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1720 __u32 iter_info_len; /* iter_info length */ 1721 }; 1722 struct { 1723 /* black box user-provided value passed through 1724 * to BPF program at the execution time and 1725 * accessible through bpf_get_attach_cookie() BPF helper 1726 */ 1727 __u64 bpf_cookie; 1728 } perf_event; 1729 struct { 1730 __u32 flags; 1731 __u32 cnt; 1732 __aligned_u64 syms; 1733 __aligned_u64 addrs; 1734 __aligned_u64 cookies; 1735 } kprobe_multi; 1736 struct { 1737 /* this is overlaid with the target_btf_id above. */ 1738 __u32 target_btf_id; 1739 /* black box user-provided value passed through 1740 * to BPF program at the execution time and 1741 * accessible through bpf_get_attach_cookie() BPF helper 1742 */ 1743 __u64 cookie; 1744 } tracing; 1745 struct { 1746 __u32 pf; 1747 __u32 hooknum; 1748 __s32 priority; 1749 __u32 flags; 1750 } netfilter; 1751 struct { 1752 union { 1753 __u32 relative_fd; 1754 __u32 relative_id; 1755 }; 1756 __u64 expected_revision; 1757 } tcx; 1758 struct { 1759 __aligned_u64 path; 1760 __aligned_u64 offsets; 1761 __aligned_u64 ref_ctr_offsets; 1762 __aligned_u64 cookies; 1763 __u32 cnt; 1764 __u32 flags; 1765 __u32 pid; 1766 } uprobe_multi; 1767 struct { 1768 union { 1769 __u32 relative_fd; 1770 __u32 relative_id; 1771 }; 1772 __u64 expected_revision; 1773 } netkit; 1774 }; 1775 } link_create; 1776 1777 struct { /* struct used by BPF_LINK_UPDATE command */ 1778 __u32 link_fd; /* link fd */ 1779 union { 1780 /* new program fd to update link with */ 1781 __u32 new_prog_fd; 1782 /* new struct_ops map fd to update link with */ 1783 __u32 new_map_fd; 1784 }; 1785 __u32 flags; /* extra flags */ 1786 union { 1787 /* expected link's program fd; is specified only if 1788 * BPF_F_REPLACE flag is set in flags. 1789 */ 1790 __u32 old_prog_fd; 1791 /* expected link's map fd; is specified only 1792 * if BPF_F_REPLACE flag is set. 1793 */ 1794 __u32 old_map_fd; 1795 }; 1796 } link_update; 1797 1798 struct { 1799 __u32 link_fd; 1800 } link_detach; 1801 1802 struct { /* struct used by BPF_ENABLE_STATS command */ 1803 __u32 type; 1804 } enable_stats; 1805 1806 struct { /* struct used by BPF_ITER_CREATE command */ 1807 __u32 link_fd; 1808 __u32 flags; 1809 } iter_create; 1810 1811 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1812 __u32 prog_fd; 1813 __u32 map_fd; 1814 __u32 flags; /* extra flags */ 1815 } prog_bind_map; 1816 1817 struct { /* struct used by BPF_TOKEN_CREATE command */ 1818 __u32 flags; 1819 __u32 bpffs_fd; 1820 } token_create; 1821 1822 } __attribute__((aligned(8))); 1823 1824 /* The description below is an attempt at providing documentation to eBPF 1825 * developers about the multiple available eBPF helper functions. It can be 1826 * parsed and used to produce a manual page. The workflow is the following, 1827 * and requires the rst2man utility: 1828 * 1829 * $ ./scripts/bpf_doc.py \ 1830 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1831 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1832 * $ man /tmp/bpf-helpers.7 1833 * 1834 * Note that in order to produce this external documentation, some RST 1835 * formatting is used in the descriptions to get "bold" and "italics" in 1836 * manual pages. Also note that the few trailing white spaces are 1837 * intentional, removing them would break paragraphs for rst2man. 1838 * 1839 * Start of BPF helper function descriptions: 1840 * 1841 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1842 * Description 1843 * Perform a lookup in *map* for an entry associated to *key*. 1844 * Return 1845 * Map value associated to *key*, or **NULL** if no entry was 1846 * found. 1847 * 1848 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1849 * Description 1850 * Add or update the value of the entry associated to *key* in 1851 * *map* with *value*. *flags* is one of: 1852 * 1853 * **BPF_NOEXIST** 1854 * The entry for *key* must not exist in the map. 1855 * **BPF_EXIST** 1856 * The entry for *key* must already exist in the map. 1857 * **BPF_ANY** 1858 * No condition on the existence of the entry for *key*. 1859 * 1860 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1861 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1862 * elements always exist), the helper would return an error. 1863 * Return 1864 * 0 on success, or a negative error in case of failure. 1865 * 1866 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1867 * Description 1868 * Delete entry with *key* from *map*. 1869 * Return 1870 * 0 on success, or a negative error in case of failure. 1871 * 1872 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1873 * Description 1874 * For tracing programs, safely attempt to read *size* bytes from 1875 * kernel space address *unsafe_ptr* and store the data in *dst*. 1876 * 1877 * Generally, use **bpf_probe_read_user**\ () or 1878 * **bpf_probe_read_kernel**\ () instead. 1879 * Return 1880 * 0 on success, or a negative error in case of failure. 1881 * 1882 * u64 bpf_ktime_get_ns(void) 1883 * Description 1884 * Return the time elapsed since system boot, in nanoseconds. 1885 * Does not include time the system was suspended. 1886 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1887 * Return 1888 * Current *ktime*. 1889 * 1890 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1891 * Description 1892 * This helper is a "printk()-like" facility for debugging. It 1893 * prints a message defined by format *fmt* (of size *fmt_size*) 1894 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1895 * available. It can take up to three additional **u64** 1896 * arguments (as an eBPF helpers, the total number of arguments is 1897 * limited to five). 1898 * 1899 * Each time the helper is called, it appends a line to the trace. 1900 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1901 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1902 * The format of the trace is customizable, and the exact output 1903 * one will get depends on the options set in 1904 * *\/sys/kernel/tracing/trace_options* (see also the 1905 * *README* file under the same directory). However, it usually 1906 * defaults to something like: 1907 * 1908 * :: 1909 * 1910 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1911 * 1912 * In the above: 1913 * 1914 * * ``telnet`` is the name of the current task. 1915 * * ``470`` is the PID of the current task. 1916 * * ``001`` is the CPU number on which the task is 1917 * running. 1918 * * In ``.N..``, each character refers to a set of 1919 * options (whether irqs are enabled, scheduling 1920 * options, whether hard/softirqs are running, level of 1921 * preempt_disabled respectively). **N** means that 1922 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1923 * are set. 1924 * * ``419421.045894`` is a timestamp. 1925 * * ``0x00000001`` is a fake value used by BPF for the 1926 * instruction pointer register. 1927 * * ``<formatted msg>`` is the message formatted with 1928 * *fmt*. 1929 * 1930 * The conversion specifiers supported by *fmt* are similar, but 1931 * more limited than for printk(). They are **%d**, **%i**, 1932 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1933 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1934 * of field, padding with zeroes, etc.) is available, and the 1935 * helper will return **-EINVAL** (but print nothing) if it 1936 * encounters an unknown specifier. 1937 * 1938 * Also, note that **bpf_trace_printk**\ () is slow, and should 1939 * only be used for debugging purposes. For this reason, a notice 1940 * block (spanning several lines) is printed to kernel logs and 1941 * states that the helper should not be used "for production use" 1942 * the first time this helper is used (or more precisely, when 1943 * **trace_printk**\ () buffers are allocated). For passing values 1944 * to user space, perf events should be preferred. 1945 * Return 1946 * The number of bytes written to the buffer, or a negative error 1947 * in case of failure. 1948 * 1949 * u32 bpf_get_prandom_u32(void) 1950 * Description 1951 * Get a pseudo-random number. 1952 * 1953 * From a security point of view, this helper uses its own 1954 * pseudo-random internal state, and cannot be used to infer the 1955 * seed of other random functions in the kernel. However, it is 1956 * essential to note that the generator used by the helper is not 1957 * cryptographically secure. 1958 * Return 1959 * A random 32-bit unsigned value. 1960 * 1961 * u32 bpf_get_smp_processor_id(void) 1962 * Description 1963 * Get the SMP (symmetric multiprocessing) processor id. Note that 1964 * all programs run with migration disabled, which means that the 1965 * SMP processor id is stable during all the execution of the 1966 * program. 1967 * Return 1968 * The SMP id of the processor running the program. 1969 * 1970 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1971 * Description 1972 * Store *len* bytes from address *from* into the packet 1973 * associated to *skb*, at *offset*. *flags* are a combination of 1974 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1975 * checksum for the packet after storing the bytes) and 1976 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1977 * **->swhash** and *skb*\ **->l4hash** to 0). 1978 * 1979 * A call to this helper is susceptible to change the underlying 1980 * packet buffer. Therefore, at load time, all checks on pointers 1981 * previously done by the verifier are invalidated and must be 1982 * performed again, if the helper is used in combination with 1983 * direct packet access. 1984 * Return 1985 * 0 on success, or a negative error in case of failure. 1986 * 1987 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1988 * Description 1989 * Recompute the layer 3 (e.g. IP) checksum for the packet 1990 * associated to *skb*. Computation is incremental, so the helper 1991 * must know the former value of the header field that was 1992 * modified (*from*), the new value of this field (*to*), and the 1993 * number of bytes (2 or 4) for this field, stored in *size*. 1994 * Alternatively, it is possible to store the difference between 1995 * the previous and the new values of the header field in *to*, by 1996 * setting *from* and *size* to 0. For both methods, *offset* 1997 * indicates the location of the IP checksum within the packet. 1998 * 1999 * This helper works in combination with **bpf_csum_diff**\ (), 2000 * which does not update the checksum in-place, but offers more 2001 * flexibility and can handle sizes larger than 2 or 4 for the 2002 * checksum to update. 2003 * 2004 * A call to this helper is susceptible to change the underlying 2005 * packet buffer. Therefore, at load time, all checks on pointers 2006 * previously done by the verifier are invalidated and must be 2007 * performed again, if the helper is used in combination with 2008 * direct packet access. 2009 * Return 2010 * 0 on success, or a negative error in case of failure. 2011 * 2012 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 2013 * Description 2014 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 2015 * packet associated to *skb*. Computation is incremental, so the 2016 * helper must know the former value of the header field that was 2017 * modified (*from*), the new value of this field (*to*), and the 2018 * number of bytes (2 or 4) for this field, stored on the lowest 2019 * four bits of *flags*. Alternatively, it is possible to store 2020 * the difference between the previous and the new values of the 2021 * header field in *to*, by setting *from* and the four lowest 2022 * bits of *flags* to 0. For both methods, *offset* indicates the 2023 * location of the IP checksum within the packet. In addition to 2024 * the size of the field, *flags* can be added (bitwise OR) actual 2025 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 2026 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 2027 * for updates resulting in a null checksum the value is set to 2028 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 2029 * the checksum is to be computed against a pseudo-header. 2030 * 2031 * This helper works in combination with **bpf_csum_diff**\ (), 2032 * which does not update the checksum in-place, but offers more 2033 * flexibility and can handle sizes larger than 2 or 4 for the 2034 * checksum to update. 2035 * 2036 * A call to this helper is susceptible to change the underlying 2037 * packet buffer. Therefore, at load time, all checks on pointers 2038 * previously done by the verifier are invalidated and must be 2039 * performed again, if the helper is used in combination with 2040 * direct packet access. 2041 * Return 2042 * 0 on success, or a negative error in case of failure. 2043 * 2044 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2045 * Description 2046 * This special helper is used to trigger a "tail call", or in 2047 * other words, to jump into another eBPF program. The same stack 2048 * frame is used (but values on stack and in registers for the 2049 * caller are not accessible to the callee). This mechanism allows 2050 * for program chaining, either for raising the maximum number of 2051 * available eBPF instructions, or to execute given programs in 2052 * conditional blocks. For security reasons, there is an upper 2053 * limit to the number of successive tail calls that can be 2054 * performed. 2055 * 2056 * Upon call of this helper, the program attempts to jump into a 2057 * program referenced at index *index* in *prog_array_map*, a 2058 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2059 * *ctx*, a pointer to the context. 2060 * 2061 * If the call succeeds, the kernel immediately runs the first 2062 * instruction of the new program. This is not a function call, 2063 * and it never returns to the previous program. If the call 2064 * fails, then the helper has no effect, and the caller continues 2065 * to run its subsequent instructions. A call can fail if the 2066 * destination program for the jump does not exist (i.e. *index* 2067 * is superior to the number of entries in *prog_array_map*), or 2068 * if the maximum number of tail calls has been reached for this 2069 * chain of programs. This limit is defined in the kernel by the 2070 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2071 * which is currently set to 33. 2072 * Return 2073 * 0 on success, or a negative error in case of failure. 2074 * 2075 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2076 * Description 2077 * Clone and redirect the packet associated to *skb* to another 2078 * net device of index *ifindex*. Both ingress and egress 2079 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2080 * value in *flags* is used to make the distinction (ingress path 2081 * is selected if the flag is present, egress path otherwise). 2082 * This is the only flag supported for now. 2083 * 2084 * In comparison with **bpf_redirect**\ () helper, 2085 * **bpf_clone_redirect**\ () has the associated cost of 2086 * duplicating the packet buffer, but this can be executed out of 2087 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2088 * efficient, but it is handled through an action code where the 2089 * redirection happens only after the eBPF program has returned. 2090 * 2091 * A call to this helper is susceptible to change the underlying 2092 * packet buffer. Therefore, at load time, all checks on pointers 2093 * previously done by the verifier are invalidated and must be 2094 * performed again, if the helper is used in combination with 2095 * direct packet access. 2096 * Return 2097 * 0 on success, or a negative error in case of failure. Positive 2098 * error indicates a potential drop or congestion in the target 2099 * device. The particular positive error codes are not defined. 2100 * 2101 * u64 bpf_get_current_pid_tgid(void) 2102 * Description 2103 * Get the current pid and tgid. 2104 * Return 2105 * A 64-bit integer containing the current tgid and pid, and 2106 * created as such: 2107 * *current_task*\ **->tgid << 32 \|** 2108 * *current_task*\ **->pid**. 2109 * 2110 * u64 bpf_get_current_uid_gid(void) 2111 * Description 2112 * Get the current uid and gid. 2113 * Return 2114 * A 64-bit integer containing the current GID and UID, and 2115 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2116 * 2117 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2118 * Description 2119 * Copy the **comm** attribute of the current task into *buf* of 2120 * *size_of_buf*. The **comm** attribute contains the name of 2121 * the executable (excluding the path) for the current task. The 2122 * *size_of_buf* must be strictly positive. On success, the 2123 * helper makes sure that the *buf* is NUL-terminated. On failure, 2124 * it is filled with zeroes. 2125 * Return 2126 * 0 on success, or a negative error in case of failure. 2127 * 2128 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2129 * Description 2130 * Retrieve the classid for the current task, i.e. for the net_cls 2131 * cgroup to which *skb* belongs. 2132 * 2133 * This helper can be used on TC egress path, but not on ingress. 2134 * 2135 * The net_cls cgroup provides an interface to tag network packets 2136 * based on a user-provided identifier for all traffic coming from 2137 * the tasks belonging to the related cgroup. See also the related 2138 * kernel documentation, available from the Linux sources in file 2139 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2140 * 2141 * The Linux kernel has two versions for cgroups: there are 2142 * cgroups v1 and cgroups v2. Both are available to users, who can 2143 * use a mixture of them, but note that the net_cls cgroup is for 2144 * cgroup v1 only. This makes it incompatible with BPF programs 2145 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2146 * only hold data for one version of cgroups at a time). 2147 * 2148 * This helper is only available is the kernel was compiled with 2149 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2150 * "**y**" or to "**m**". 2151 * Return 2152 * The classid, or 0 for the default unconfigured classid. 2153 * 2154 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2155 * Description 2156 * Push a *vlan_tci* (VLAN tag control information) of protocol 2157 * *vlan_proto* to the packet associated to *skb*, then update 2158 * the checksum. Note that if *vlan_proto* is different from 2159 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2160 * be **ETH_P_8021Q**. 2161 * 2162 * A call to this helper is susceptible to change the underlying 2163 * packet buffer. Therefore, at load time, all checks on pointers 2164 * previously done by the verifier are invalidated and must be 2165 * performed again, if the helper is used in combination with 2166 * direct packet access. 2167 * Return 2168 * 0 on success, or a negative error in case of failure. 2169 * 2170 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2171 * Description 2172 * Pop a VLAN header from the packet associated to *skb*. 2173 * 2174 * A call to this helper is susceptible to change the underlying 2175 * packet buffer. Therefore, at load time, all checks on pointers 2176 * previously done by the verifier are invalidated and must be 2177 * performed again, if the helper is used in combination with 2178 * direct packet access. 2179 * Return 2180 * 0 on success, or a negative error in case of failure. 2181 * 2182 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2183 * Description 2184 * Get tunnel metadata. This helper takes a pointer *key* to an 2185 * empty **struct bpf_tunnel_key** of **size**, that will be 2186 * filled with tunnel metadata for the packet associated to *skb*. 2187 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2188 * indicates that the tunnel is based on IPv6 protocol instead of 2189 * IPv4. 2190 * 2191 * The **struct bpf_tunnel_key** is an object that generalizes the 2192 * principal parameters used by various tunneling protocols into a 2193 * single struct. This way, it can be used to easily make a 2194 * decision based on the contents of the encapsulation header, 2195 * "summarized" in this struct. In particular, it holds the IP 2196 * address of the remote end (IPv4 or IPv6, depending on the case) 2197 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2198 * this struct exposes the *key*\ **->tunnel_id**, which is 2199 * generally mapped to a VNI (Virtual Network Identifier), making 2200 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2201 * () helper. 2202 * 2203 * Let's imagine that the following code is part of a program 2204 * attached to the TC ingress interface, on one end of a GRE 2205 * tunnel, and is supposed to filter out all messages coming from 2206 * remote ends with IPv4 address other than 10.0.0.1: 2207 * 2208 * :: 2209 * 2210 * int ret; 2211 * struct bpf_tunnel_key key = {}; 2212 * 2213 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2214 * if (ret < 0) 2215 * return TC_ACT_SHOT; // drop packet 2216 * 2217 * if (key.remote_ipv4 != 0x0a000001) 2218 * return TC_ACT_SHOT; // drop packet 2219 * 2220 * return TC_ACT_OK; // accept packet 2221 * 2222 * This interface can also be used with all encapsulation devices 2223 * that can operate in "collect metadata" mode: instead of having 2224 * one network device per specific configuration, the "collect 2225 * metadata" mode only requires a single device where the 2226 * configuration can be extracted from this helper. 2227 * 2228 * This can be used together with various tunnels such as VXLan, 2229 * Geneve, GRE or IP in IP (IPIP). 2230 * Return 2231 * 0 on success, or a negative error in case of failure. 2232 * 2233 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2234 * Description 2235 * Populate tunnel metadata for packet associated to *skb.* The 2236 * tunnel metadata is set to the contents of *key*, of *size*. The 2237 * *flags* can be set to a combination of the following values: 2238 * 2239 * **BPF_F_TUNINFO_IPV6** 2240 * Indicate that the tunnel is based on IPv6 protocol 2241 * instead of IPv4. 2242 * **BPF_F_ZERO_CSUM_TX** 2243 * For IPv4 packets, add a flag to tunnel metadata 2244 * indicating that checksum computation should be skipped 2245 * and checksum set to zeroes. 2246 * **BPF_F_DONT_FRAGMENT** 2247 * Add a flag to tunnel metadata indicating that the 2248 * packet should not be fragmented. 2249 * **BPF_F_SEQ_NUMBER** 2250 * Add a flag to tunnel metadata indicating that a 2251 * sequence number should be added to tunnel header before 2252 * sending the packet. This flag was added for GRE 2253 * encapsulation, but might be used with other protocols 2254 * as well in the future. 2255 * **BPF_F_NO_TUNNEL_KEY** 2256 * Add a flag to tunnel metadata indicating that no tunnel 2257 * key should be set in the resulting tunnel header. 2258 * 2259 * Here is a typical usage on the transmit path: 2260 * 2261 * :: 2262 * 2263 * struct bpf_tunnel_key key; 2264 * populate key ... 2265 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2266 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2267 * 2268 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2269 * helper for additional information. 2270 * Return 2271 * 0 on success, or a negative error in case of failure. 2272 * 2273 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2274 * Description 2275 * Read the value of a perf event counter. This helper relies on a 2276 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2277 * the perf event counter is selected when *map* is updated with 2278 * perf event file descriptors. The *map* is an array whose size 2279 * is the number of available CPUs, and each cell contains a value 2280 * relative to one CPU. The value to retrieve is indicated by 2281 * *flags*, that contains the index of the CPU to look up, masked 2282 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2283 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2284 * current CPU should be retrieved. 2285 * 2286 * Note that before Linux 4.13, only hardware perf event can be 2287 * retrieved. 2288 * 2289 * Also, be aware that the newer helper 2290 * **bpf_perf_event_read_value**\ () is recommended over 2291 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2292 * quirks where error and counter value are used as a return code 2293 * (which is wrong to do since ranges may overlap). This issue is 2294 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2295 * time provides more features over the **bpf_perf_event_read**\ 2296 * () interface. Please refer to the description of 2297 * **bpf_perf_event_read_value**\ () for details. 2298 * Return 2299 * The value of the perf event counter read from the map, or a 2300 * negative error code in case of failure. 2301 * 2302 * long bpf_redirect(u32 ifindex, u64 flags) 2303 * Description 2304 * Redirect the packet to another net device of index *ifindex*. 2305 * This helper is somewhat similar to **bpf_clone_redirect**\ 2306 * (), except that the packet is not cloned, which provides 2307 * increased performance. 2308 * 2309 * Except for XDP, both ingress and egress interfaces can be used 2310 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2311 * to make the distinction (ingress path is selected if the flag 2312 * is present, egress path otherwise). Currently, XDP only 2313 * supports redirection to the egress interface, and accepts no 2314 * flag at all. 2315 * 2316 * The same effect can also be attained with the more generic 2317 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2318 * redirect target instead of providing it directly to the helper. 2319 * Return 2320 * For XDP, the helper returns **XDP_REDIRECT** on success or 2321 * **XDP_ABORTED** on error. For other program types, the values 2322 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2323 * error. 2324 * 2325 * u32 bpf_get_route_realm(struct sk_buff *skb) 2326 * Description 2327 * Retrieve the realm or the route, that is to say the 2328 * **tclassid** field of the destination for the *skb*. The 2329 * identifier retrieved is a user-provided tag, similar to the 2330 * one used with the net_cls cgroup (see description for 2331 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2332 * held by a route (a destination entry), not by a task. 2333 * 2334 * Retrieving this identifier works with the clsact TC egress hook 2335 * (see also **tc-bpf(8)**), or alternatively on conventional 2336 * classful egress qdiscs, but not on TC ingress path. In case of 2337 * clsact TC egress hook, this has the advantage that, internally, 2338 * the destination entry has not been dropped yet in the transmit 2339 * path. Therefore, the destination entry does not need to be 2340 * artificially held via **netif_keep_dst**\ () for a classful 2341 * qdisc until the *skb* is freed. 2342 * 2343 * This helper is available only if the kernel was compiled with 2344 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2345 * Return 2346 * The realm of the route for the packet associated to *skb*, or 0 2347 * if none was found. 2348 * 2349 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2350 * Description 2351 * Write raw *data* blob into a special BPF perf event held by 2352 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2353 * event must have the following attributes: **PERF_SAMPLE_RAW** 2354 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2355 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2356 * 2357 * The *flags* are used to indicate the index in *map* for which 2358 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2359 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2360 * to indicate that the index of the current CPU core should be 2361 * used. 2362 * 2363 * The value to write, of *size*, is passed through eBPF stack and 2364 * pointed by *data*. 2365 * 2366 * The context of the program *ctx* needs also be passed to the 2367 * helper. 2368 * 2369 * On user space, a program willing to read the values needs to 2370 * call **perf_event_open**\ () on the perf event (either for 2371 * one or for all CPUs) and to store the file descriptor into the 2372 * *map*. This must be done before the eBPF program can send data 2373 * into it. An example is available in file 2374 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2375 * tree (the eBPF program counterpart is in 2376 * *samples/bpf/trace_output_kern.c*). 2377 * 2378 * **bpf_perf_event_output**\ () achieves better performance 2379 * than **bpf_trace_printk**\ () for sharing data with user 2380 * space, and is much better suitable for streaming data from eBPF 2381 * programs. 2382 * 2383 * Note that this helper is not restricted to tracing use cases 2384 * and can be used with programs attached to TC or XDP as well, 2385 * where it allows for passing data to user space listeners. Data 2386 * can be: 2387 * 2388 * * Only custom structs, 2389 * * Only the packet payload, or 2390 * * A combination of both. 2391 * Return 2392 * 0 on success, or a negative error in case of failure. 2393 * 2394 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2395 * Description 2396 * This helper was provided as an easy way to load data from a 2397 * packet. It can be used to load *len* bytes from *offset* from 2398 * the packet associated to *skb*, into the buffer pointed by 2399 * *to*. 2400 * 2401 * Since Linux 4.7, usage of this helper has mostly been replaced 2402 * by "direct packet access", enabling packet data to be 2403 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2404 * pointing respectively to the first byte of packet data and to 2405 * the byte after the last byte of packet data. However, it 2406 * remains useful if one wishes to read large quantities of data 2407 * at once from a packet into the eBPF stack. 2408 * Return 2409 * 0 on success, or a negative error in case of failure. 2410 * 2411 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2412 * Description 2413 * Walk a user or a kernel stack and return its id. To achieve 2414 * this, the helper needs *ctx*, which is a pointer to the context 2415 * on which the tracing program is executed, and a pointer to a 2416 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2417 * 2418 * The last argument, *flags*, holds the number of stack frames to 2419 * skip (from 0 to 255), masked with 2420 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2421 * a combination of the following flags: 2422 * 2423 * **BPF_F_USER_STACK** 2424 * Collect a user space stack instead of a kernel stack. 2425 * **BPF_F_FAST_STACK_CMP** 2426 * Compare stacks by hash only. 2427 * **BPF_F_REUSE_STACKID** 2428 * If two different stacks hash into the same *stackid*, 2429 * discard the old one. 2430 * 2431 * The stack id retrieved is a 32 bit long integer handle which 2432 * can be further combined with other data (including other stack 2433 * ids) and used as a key into maps. This can be useful for 2434 * generating a variety of graphs (such as flame graphs or off-cpu 2435 * graphs). 2436 * 2437 * For walking a stack, this helper is an improvement over 2438 * **bpf_probe_read**\ (), which can be used with unrolled loops 2439 * but is not efficient and consumes a lot of eBPF instructions. 2440 * Instead, **bpf_get_stackid**\ () can collect up to 2441 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2442 * this limit can be controlled with the **sysctl** program, and 2443 * that it should be manually increased in order to profile long 2444 * user stacks (such as stacks for Java programs). To do so, use: 2445 * 2446 * :: 2447 * 2448 * # sysctl kernel.perf_event_max_stack=<new value> 2449 * Return 2450 * The positive or null stack id on success, or a negative error 2451 * in case of failure. 2452 * 2453 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2454 * Description 2455 * Compute a checksum difference, from the raw buffer pointed by 2456 * *from*, of length *from_size* (that must be a multiple of 4), 2457 * towards the raw buffer pointed by *to*, of size *to_size* 2458 * (same remark). An optional *seed* can be added to the value 2459 * (this can be cascaded, the seed may come from a previous call 2460 * to the helper). 2461 * 2462 * This is flexible enough to be used in several ways: 2463 * 2464 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2465 * checksum, it can be used when pushing new data. 2466 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2467 * checksum, it can be used when removing data from a packet. 2468 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2469 * can be used to compute a diff. Note that *from_size* and 2470 * *to_size* do not need to be equal. 2471 * 2472 * This helper can be used in combination with 2473 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2474 * which one can feed in the difference computed with 2475 * **bpf_csum_diff**\ (). 2476 * Return 2477 * The checksum result, or a negative error code in case of 2478 * failure. 2479 * 2480 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2481 * Description 2482 * Retrieve tunnel options metadata for the packet associated to 2483 * *skb*, and store the raw tunnel option data to the buffer *opt* 2484 * of *size*. 2485 * 2486 * This helper can be used with encapsulation devices that can 2487 * operate in "collect metadata" mode (please refer to the related 2488 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2489 * more details). A particular example where this can be used is 2490 * in combination with the Geneve encapsulation protocol, where it 2491 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2492 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2493 * the eBPF program. This allows for full customization of these 2494 * headers. 2495 * Return 2496 * The size of the option data retrieved. 2497 * 2498 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2499 * Description 2500 * Set tunnel options metadata for the packet associated to *skb* 2501 * to the option data contained in the raw buffer *opt* of *size*. 2502 * 2503 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2504 * helper for additional information. 2505 * Return 2506 * 0 on success, or a negative error in case of failure. 2507 * 2508 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2509 * Description 2510 * Change the protocol of the *skb* to *proto*. Currently 2511 * supported are transition from IPv4 to IPv6, and from IPv6 to 2512 * IPv4. The helper takes care of the groundwork for the 2513 * transition, including resizing the socket buffer. The eBPF 2514 * program is expected to fill the new headers, if any, via 2515 * **skb_store_bytes**\ () and to recompute the checksums with 2516 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2517 * (). The main case for this helper is to perform NAT64 2518 * operations out of an eBPF program. 2519 * 2520 * Internally, the GSO type is marked as dodgy so that headers are 2521 * checked and segments are recalculated by the GSO/GRO engine. 2522 * The size for GSO target is adapted as well. 2523 * 2524 * All values for *flags* are reserved for future usage, and must 2525 * be left at zero. 2526 * 2527 * A call to this helper is susceptible to change the underlying 2528 * packet buffer. Therefore, at load time, all checks on pointers 2529 * previously done by the verifier are invalidated and must be 2530 * performed again, if the helper is used in combination with 2531 * direct packet access. 2532 * Return 2533 * 0 on success, or a negative error in case of failure. 2534 * 2535 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2536 * Description 2537 * Change the packet type for the packet associated to *skb*. This 2538 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2539 * the eBPF program does not have a write access to *skb*\ 2540 * **->pkt_type** beside this helper. Using a helper here allows 2541 * for graceful handling of errors. 2542 * 2543 * The major use case is to change incoming *skb*s to 2544 * **PACKET_HOST** in a programmatic way instead of having to 2545 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2546 * example. 2547 * 2548 * Note that *type* only allows certain values. At this time, they 2549 * are: 2550 * 2551 * **PACKET_HOST** 2552 * Packet is for us. 2553 * **PACKET_BROADCAST** 2554 * Send packet to all. 2555 * **PACKET_MULTICAST** 2556 * Send packet to group. 2557 * **PACKET_OTHERHOST** 2558 * Send packet to someone else. 2559 * Return 2560 * 0 on success, or a negative error in case of failure. 2561 * 2562 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2563 * Description 2564 * Check whether *skb* is a descendant of the cgroup2 held by 2565 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2566 * Return 2567 * The return value depends on the result of the test, and can be: 2568 * 2569 * * 0, if the *skb* failed the cgroup2 descendant test. 2570 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2571 * * A negative error code, if an error occurred. 2572 * 2573 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2574 * Description 2575 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2576 * not set, in particular if the hash was cleared due to mangling, 2577 * recompute this hash. Later accesses to the hash can be done 2578 * directly with *skb*\ **->hash**. 2579 * 2580 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2581 * prototype with **bpf_skb_change_proto**\ (), or calling 2582 * **bpf_skb_store_bytes**\ () with the 2583 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2584 * the hash and to trigger a new computation for the next call to 2585 * **bpf_get_hash_recalc**\ (). 2586 * Return 2587 * The 32-bit hash. 2588 * 2589 * u64 bpf_get_current_task(void) 2590 * Description 2591 * Get the current task. 2592 * Return 2593 * A pointer to the current task struct. 2594 * 2595 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2596 * Description 2597 * Attempt in a safe way to write *len* bytes from the buffer 2598 * *src* to *dst* in memory. It only works for threads that are in 2599 * user context, and *dst* must be a valid user space address. 2600 * 2601 * This helper should not be used to implement any kind of 2602 * security mechanism because of TOC-TOU attacks, but rather to 2603 * debug, divert, and manipulate execution of semi-cooperative 2604 * processes. 2605 * 2606 * Keep in mind that this feature is meant for experiments, and it 2607 * has a risk of crashing the system and running programs. 2608 * Therefore, when an eBPF program using this helper is attached, 2609 * a warning including PID and process name is printed to kernel 2610 * logs. 2611 * Return 2612 * 0 on success, or a negative error in case of failure. 2613 * 2614 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2615 * Description 2616 * Check whether the probe is being run is the context of a given 2617 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2618 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2619 * Return 2620 * The return value depends on the result of the test, and can be: 2621 * 2622 * * 1, if current task belongs to the cgroup2. 2623 * * 0, if current task does not belong to the cgroup2. 2624 * * A negative error code, if an error occurred. 2625 * 2626 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2627 * Description 2628 * Resize (trim or grow) the packet associated to *skb* to the 2629 * new *len*. The *flags* are reserved for future usage, and must 2630 * be left at zero. 2631 * 2632 * The basic idea is that the helper performs the needed work to 2633 * change the size of the packet, then the eBPF program rewrites 2634 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2635 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2636 * and others. This helper is a slow path utility intended for 2637 * replies with control messages. And because it is targeted for 2638 * slow path, the helper itself can afford to be slow: it 2639 * implicitly linearizes, unclones and drops offloads from the 2640 * *skb*. 2641 * 2642 * A call to this helper is susceptible to change the underlying 2643 * packet buffer. Therefore, at load time, all checks on pointers 2644 * previously done by the verifier are invalidated and must be 2645 * performed again, if the helper is used in combination with 2646 * direct packet access. 2647 * Return 2648 * 0 on success, or a negative error in case of failure. 2649 * 2650 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2651 * Description 2652 * Pull in non-linear data in case the *skb* is non-linear and not 2653 * all of *len* are part of the linear section. Make *len* bytes 2654 * from *skb* readable and writable. If a zero value is passed for 2655 * *len*, then all bytes in the linear part of *skb* will be made 2656 * readable and writable. 2657 * 2658 * This helper is only needed for reading and writing with direct 2659 * packet access. 2660 * 2661 * For direct packet access, testing that offsets to access 2662 * are within packet boundaries (test on *skb*\ **->data_end**) is 2663 * susceptible to fail if offsets are invalid, or if the requested 2664 * data is in non-linear parts of the *skb*. On failure the 2665 * program can just bail out, or in the case of a non-linear 2666 * buffer, use a helper to make the data available. The 2667 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2668 * the data. Another one consists in using **bpf_skb_pull_data** 2669 * to pull in once the non-linear parts, then retesting and 2670 * eventually access the data. 2671 * 2672 * At the same time, this also makes sure the *skb* is uncloned, 2673 * which is a necessary condition for direct write. As this needs 2674 * to be an invariant for the write part only, the verifier 2675 * detects writes and adds a prologue that is calling 2676 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2677 * the very beginning in case it is indeed cloned. 2678 * 2679 * A call to this helper is susceptible to change the underlying 2680 * packet buffer. Therefore, at load time, all checks on pointers 2681 * previously done by the verifier are invalidated and must be 2682 * performed again, if the helper is used in combination with 2683 * direct packet access. 2684 * Return 2685 * 0 on success, or a negative error in case of failure. 2686 * 2687 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2688 * Description 2689 * Add the checksum *csum* into *skb*\ **->csum** in case the 2690 * driver has supplied a checksum for the entire packet into that 2691 * field. Return an error otherwise. This helper is intended to be 2692 * used in combination with **bpf_csum_diff**\ (), in particular 2693 * when the checksum needs to be updated after data has been 2694 * written into the packet through direct packet access. 2695 * Return 2696 * The checksum on success, or a negative error code in case of 2697 * failure. 2698 * 2699 * void bpf_set_hash_invalid(struct sk_buff *skb) 2700 * Description 2701 * Invalidate the current *skb*\ **->hash**. It can be used after 2702 * mangling on headers through direct packet access, in order to 2703 * indicate that the hash is outdated and to trigger a 2704 * recalculation the next time the kernel tries to access this 2705 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2706 * Return 2707 * void. 2708 * 2709 * long bpf_get_numa_node_id(void) 2710 * Description 2711 * Return the id of the current NUMA node. The primary use case 2712 * for this helper is the selection of sockets for the local NUMA 2713 * node, when the program is attached to sockets using the 2714 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2715 * but the helper is also available to other eBPF program types, 2716 * similarly to **bpf_get_smp_processor_id**\ (). 2717 * Return 2718 * The id of current NUMA node. 2719 * 2720 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2721 * Description 2722 * Grows headroom of packet associated to *skb* and adjusts the 2723 * offset of the MAC header accordingly, adding *len* bytes of 2724 * space. It automatically extends and reallocates memory as 2725 * required. 2726 * 2727 * This helper can be used on a layer 3 *skb* to push a MAC header 2728 * for redirection into a layer 2 device. 2729 * 2730 * All values for *flags* are reserved for future usage, and must 2731 * be left at zero. 2732 * 2733 * A call to this helper is susceptible to change the underlying 2734 * packet buffer. Therefore, at load time, all checks on pointers 2735 * previously done by the verifier are invalidated and must be 2736 * performed again, if the helper is used in combination with 2737 * direct packet access. 2738 * Return 2739 * 0 on success, or a negative error in case of failure. 2740 * 2741 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2742 * Description 2743 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2744 * it is possible to use a negative value for *delta*. This helper 2745 * can be used to prepare the packet for pushing or popping 2746 * headers. 2747 * 2748 * A call to this helper is susceptible to change the underlying 2749 * packet buffer. Therefore, at load time, all checks on pointers 2750 * previously done by the verifier are invalidated and must be 2751 * performed again, if the helper is used in combination with 2752 * direct packet access. 2753 * Return 2754 * 0 on success, or a negative error in case of failure. 2755 * 2756 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2757 * Description 2758 * Copy a NUL terminated string from an unsafe kernel address 2759 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2760 * more details. 2761 * 2762 * Generally, use **bpf_probe_read_user_str**\ () or 2763 * **bpf_probe_read_kernel_str**\ () instead. 2764 * Return 2765 * On success, the strictly positive length of the string, 2766 * including the trailing NUL character. On error, a negative 2767 * value. 2768 * 2769 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2770 * Description 2771 * If the **struct sk_buff** pointed by *skb* has a known socket, 2772 * retrieve the cookie (generated by the kernel) of this socket. 2773 * If no cookie has been set yet, generate a new cookie. Once 2774 * generated, the socket cookie remains stable for the life of the 2775 * socket. This helper can be useful for monitoring per socket 2776 * networking traffic statistics as it provides a global socket 2777 * identifier that can be assumed unique. 2778 * Return 2779 * A 8-byte long unique number on success, or 0 if the socket 2780 * field is missing inside *skb*. 2781 * 2782 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2783 * Description 2784 * Equivalent to bpf_get_socket_cookie() helper that accepts 2785 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2786 * Return 2787 * A 8-byte long unique number. 2788 * 2789 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2790 * Description 2791 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2792 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2793 * Return 2794 * A 8-byte long unique number. 2795 * 2796 * u64 bpf_get_socket_cookie(struct sock *sk) 2797 * Description 2798 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2799 * *sk*, but gets socket from a BTF **struct sock**. This helper 2800 * also works for sleepable programs. 2801 * Return 2802 * A 8-byte long unique number or 0 if *sk* is NULL. 2803 * 2804 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2805 * Description 2806 * Get the owner UID of the socked associated to *skb*. 2807 * Return 2808 * The owner UID of the socket associated to *skb*. If the socket 2809 * is **NULL**, or if it is not a full socket (i.e. if it is a 2810 * time-wait or a request socket instead), **overflowuid** value 2811 * is returned (note that **overflowuid** might also be the actual 2812 * UID value for the socket). 2813 * 2814 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2815 * Description 2816 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2817 * to value *hash*. 2818 * Return 2819 * 0 2820 * 2821 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2822 * Description 2823 * Emulate a call to **setsockopt()** on the socket associated to 2824 * *bpf_socket*, which must be a full socket. The *level* at 2825 * which the option resides and the name *optname* of the option 2826 * must be specified, see **setsockopt(2)** for more information. 2827 * The option value of length *optlen* is pointed by *optval*. 2828 * 2829 * *bpf_socket* should be one of the following: 2830 * 2831 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2832 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2833 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2834 * 2835 * This helper actually implements a subset of **setsockopt()**. 2836 * It supports the following *level*\ s: 2837 * 2838 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2839 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2840 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2841 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2842 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2843 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2844 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2845 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2846 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2847 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2848 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2849 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2850 * **TCP_BPF_RTO_MIN**. 2851 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2852 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2853 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2854 * Return 2855 * 0 on success, or a negative error in case of failure. 2856 * 2857 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2858 * Description 2859 * Grow or shrink the room for data in the packet associated to 2860 * *skb* by *len_diff*, and according to the selected *mode*. 2861 * 2862 * By default, the helper will reset any offloaded checksum 2863 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2864 * by the following flag: 2865 * 2866 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2867 * checksum data of the skb to CHECKSUM_NONE. 2868 * 2869 * There are two supported modes at this time: 2870 * 2871 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2872 * (room space is added or removed between the layer 2 and 2873 * layer 3 headers). 2874 * 2875 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2876 * (room space is added or removed between the layer 3 and 2877 * layer 4 headers). 2878 * 2879 * The following flags are supported at this time: 2880 * 2881 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2882 * Adjusting mss in this way is not allowed for datagrams. 2883 * 2884 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2885 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2886 * Any new space is reserved to hold a tunnel header. 2887 * Configure skb offsets and other fields accordingly. 2888 * 2889 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2890 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2891 * Use with ENCAP_L3 flags to further specify the tunnel type. 2892 * 2893 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2894 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2895 * type; *len* is the length of the inner MAC header. 2896 * 2897 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2898 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2899 * L2 type as Ethernet. 2900 * 2901 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2902 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2903 * Indicate the new IP header version after decapsulating the outer 2904 * IP header. Used when the inner and outer IP versions are different. 2905 * 2906 * A call to this helper is susceptible to change the underlying 2907 * packet buffer. Therefore, at load time, all checks on pointers 2908 * previously done by the verifier are invalidated and must be 2909 * performed again, if the helper is used in combination with 2910 * direct packet access. 2911 * Return 2912 * 0 on success, or a negative error in case of failure. 2913 * 2914 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2915 * Description 2916 * Redirect the packet to the endpoint referenced by *map* at 2917 * index *key*. Depending on its type, this *map* can contain 2918 * references to net devices (for forwarding packets through other 2919 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2920 * but this is only implemented for native XDP (with driver 2921 * support) as of this writing). 2922 * 2923 * The lower two bits of *flags* are used as the return code if 2924 * the map lookup fails. This is so that the return value can be 2925 * one of the XDP program return codes up to **XDP_TX**, as chosen 2926 * by the caller. The higher bits of *flags* can be set to 2927 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2928 * 2929 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2930 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2931 * interface will be excluded when do broadcasting. 2932 * 2933 * See also **bpf_redirect**\ (), which only supports redirecting 2934 * to an ifindex, but doesn't require a map to do so. 2935 * Return 2936 * **XDP_REDIRECT** on success, or the value of the two lower bits 2937 * of the *flags* argument on error. 2938 * 2939 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2940 * Description 2941 * Redirect the packet to the socket referenced by *map* (of type 2942 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2943 * egress interfaces can be used for redirection. The 2944 * **BPF_F_INGRESS** value in *flags* is used to make the 2945 * distinction (ingress path is selected if the flag is present, 2946 * egress path otherwise). This is the only flag supported for now. 2947 * Return 2948 * **SK_PASS** on success, or **SK_DROP** on error. 2949 * 2950 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2951 * Description 2952 * Add an entry to, or update a *map* referencing sockets. The 2953 * *skops* is used as a new value for the entry associated to 2954 * *key*. *flags* is one of: 2955 * 2956 * **BPF_NOEXIST** 2957 * The entry for *key* must not exist in the map. 2958 * **BPF_EXIST** 2959 * The entry for *key* must already exist in the map. 2960 * **BPF_ANY** 2961 * No condition on the existence of the entry for *key*. 2962 * 2963 * If the *map* has eBPF programs (parser and verdict), those will 2964 * be inherited by the socket being added. If the socket is 2965 * already attached to eBPF programs, this results in an error. 2966 * Return 2967 * 0 on success, or a negative error in case of failure. 2968 * 2969 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2970 * Description 2971 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2972 * *delta* (which can be positive or negative). Note that this 2973 * operation modifies the address stored in *xdp_md*\ **->data**, 2974 * so the latter must be loaded only after the helper has been 2975 * called. 2976 * 2977 * The use of *xdp_md*\ **->data_meta** is optional and programs 2978 * are not required to use it. The rationale is that when the 2979 * packet is processed with XDP (e.g. as DoS filter), it is 2980 * possible to push further meta data along with it before passing 2981 * to the stack, and to give the guarantee that an ingress eBPF 2982 * program attached as a TC classifier on the same device can pick 2983 * this up for further post-processing. Since TC works with socket 2984 * buffers, it remains possible to set from XDP the **mark** or 2985 * **priority** pointers, or other pointers for the socket buffer. 2986 * Having this scratch space generic and programmable allows for 2987 * more flexibility as the user is free to store whatever meta 2988 * data they need. 2989 * 2990 * A call to this helper is susceptible to change the underlying 2991 * packet buffer. Therefore, at load time, all checks on pointers 2992 * previously done by the verifier are invalidated and must be 2993 * performed again, if the helper is used in combination with 2994 * direct packet access. 2995 * Return 2996 * 0 on success, or a negative error in case of failure. 2997 * 2998 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2999 * Description 3000 * Read the value of a perf event counter, and store it into *buf* 3001 * of size *buf_size*. This helper relies on a *map* of type 3002 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 3003 * counter is selected when *map* is updated with perf event file 3004 * descriptors. The *map* is an array whose size is the number of 3005 * available CPUs, and each cell contains a value relative to one 3006 * CPU. The value to retrieve is indicated by *flags*, that 3007 * contains the index of the CPU to look up, masked with 3008 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 3009 * **BPF_F_CURRENT_CPU** to indicate that the value for the 3010 * current CPU should be retrieved. 3011 * 3012 * This helper behaves in a way close to 3013 * **bpf_perf_event_read**\ () helper, save that instead of 3014 * just returning the value observed, it fills the *buf* 3015 * structure. This allows for additional data to be retrieved: in 3016 * particular, the enabled and running times (in *buf*\ 3017 * **->enabled** and *buf*\ **->running**, respectively) are 3018 * copied. In general, **bpf_perf_event_read_value**\ () is 3019 * recommended over **bpf_perf_event_read**\ (), which has some 3020 * ABI issues and provides fewer functionalities. 3021 * 3022 * These values are interesting, because hardware PMU (Performance 3023 * Monitoring Unit) counters are limited resources. When there are 3024 * more PMU based perf events opened than available counters, 3025 * kernel will multiplex these events so each event gets certain 3026 * percentage (but not all) of the PMU time. In case that 3027 * multiplexing happens, the number of samples or counter value 3028 * will not reflect the case compared to when no multiplexing 3029 * occurs. This makes comparison between different runs difficult. 3030 * Typically, the counter value should be normalized before 3031 * comparing to other experiments. The usual normalization is done 3032 * as follows. 3033 * 3034 * :: 3035 * 3036 * normalized_counter = counter * t_enabled / t_running 3037 * 3038 * Where t_enabled is the time enabled for event and t_running is 3039 * the time running for event since last normalization. The 3040 * enabled and running times are accumulated since the perf event 3041 * open. To achieve scaling factor between two invocations of an 3042 * eBPF program, users can use CPU id as the key (which is 3043 * typical for perf array usage model) to remember the previous 3044 * value and do the calculation inside the eBPF program. 3045 * Return 3046 * 0 on success, or a negative error in case of failure. 3047 * 3048 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3049 * Description 3050 * For an eBPF program attached to a perf event, retrieve the 3051 * value of the event counter associated to *ctx* and store it in 3052 * the structure pointed by *buf* and of size *buf_size*. Enabled 3053 * and running times are also stored in the structure (see 3054 * description of helper **bpf_perf_event_read_value**\ () for 3055 * more details). 3056 * Return 3057 * 0 on success, or a negative error in case of failure. 3058 * 3059 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3060 * Description 3061 * Emulate a call to **getsockopt()** on the socket associated to 3062 * *bpf_socket*, which must be a full socket. The *level* at 3063 * which the option resides and the name *optname* of the option 3064 * must be specified, see **getsockopt(2)** for more information. 3065 * The retrieved value is stored in the structure pointed by 3066 * *opval* and of length *optlen*. 3067 * 3068 * *bpf_socket* should be one of the following: 3069 * 3070 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3071 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3072 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3073 * 3074 * This helper actually implements a subset of **getsockopt()**. 3075 * It supports the same set of *optname*\ s that is supported by 3076 * the **bpf_setsockopt**\ () helper. The exceptions are 3077 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3078 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3079 * Return 3080 * 0 on success, or a negative error in case of failure. 3081 * 3082 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3083 * Description 3084 * Used for error injection, this helper uses kprobes to override 3085 * the return value of the probed function, and to set it to *rc*. 3086 * The first argument is the context *regs* on which the kprobe 3087 * works. 3088 * 3089 * This helper works by setting the PC (program counter) 3090 * to an override function which is run in place of the original 3091 * probed function. This means the probed function is not run at 3092 * all. The replacement function just returns with the required 3093 * value. 3094 * 3095 * This helper has security implications, and thus is subject to 3096 * restrictions. It is only available if the kernel was compiled 3097 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3098 * option, and in this case it only works on functions tagged with 3099 * **ALLOW_ERROR_INJECTION** in the kernel code. 3100 * 3101 * Also, the helper is only available for the architectures having 3102 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 3103 * x86 architecture is the only one to support this feature. 3104 * Return 3105 * 0 3106 * 3107 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3108 * Description 3109 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3110 * for the full TCP socket associated to *bpf_sock_ops* to 3111 * *argval*. 3112 * 3113 * The primary use of this field is to determine if there should 3114 * be calls to eBPF programs of type 3115 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3116 * code. A program of the same type can change its value, per 3117 * connection and as necessary, when the connection is 3118 * established. This field is directly accessible for reading, but 3119 * this helper must be used for updates in order to return an 3120 * error if an eBPF program tries to set a callback that is not 3121 * supported in the current kernel. 3122 * 3123 * *argval* is a flag array which can combine these flags: 3124 * 3125 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3126 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3127 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3128 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3129 * 3130 * Therefore, this function can be used to clear a callback flag by 3131 * setting the appropriate bit to zero. e.g. to disable the RTO 3132 * callback: 3133 * 3134 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3135 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3136 * 3137 * Here are some examples of where one could call such eBPF 3138 * program: 3139 * 3140 * * When RTO fires. 3141 * * When a packet is retransmitted. 3142 * * When the connection terminates. 3143 * * When a packet is sent. 3144 * * When a packet is received. 3145 * Return 3146 * Code **-EINVAL** if the socket is not a full TCP socket; 3147 * otherwise, a positive number containing the bits that could not 3148 * be set is returned (which comes down to 0 if all bits were set 3149 * as required). 3150 * 3151 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3152 * Description 3153 * This helper is used in programs implementing policies at the 3154 * socket level. If the message *msg* is allowed to pass (i.e. if 3155 * the verdict eBPF program returns **SK_PASS**), redirect it to 3156 * the socket referenced by *map* (of type 3157 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3158 * egress interfaces can be used for redirection. The 3159 * **BPF_F_INGRESS** value in *flags* is used to make the 3160 * distinction (ingress path is selected if the flag is present, 3161 * egress path otherwise). This is the only flag supported for now. 3162 * Return 3163 * **SK_PASS** on success, or **SK_DROP** on error. 3164 * 3165 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3166 * Description 3167 * For socket policies, apply the verdict of the eBPF program to 3168 * the next *bytes* (number of bytes) of message *msg*. 3169 * 3170 * For example, this helper can be used in the following cases: 3171 * 3172 * * A single **sendmsg**\ () or **sendfile**\ () system call 3173 * contains multiple logical messages that the eBPF program is 3174 * supposed to read and for which it should apply a verdict. 3175 * * An eBPF program only cares to read the first *bytes* of a 3176 * *msg*. If the message has a large payload, then setting up 3177 * and calling the eBPF program repeatedly for all bytes, even 3178 * though the verdict is already known, would create unnecessary 3179 * overhead. 3180 * 3181 * When called from within an eBPF program, the helper sets a 3182 * counter internal to the BPF infrastructure, that is used to 3183 * apply the last verdict to the next *bytes*. If *bytes* is 3184 * smaller than the current data being processed from a 3185 * **sendmsg**\ () or **sendfile**\ () system call, the first 3186 * *bytes* will be sent and the eBPF program will be re-run with 3187 * the pointer for start of data pointing to byte number *bytes* 3188 * **+ 1**. If *bytes* is larger than the current data being 3189 * processed, then the eBPF verdict will be applied to multiple 3190 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3191 * consumed. 3192 * 3193 * Note that if a socket closes with the internal counter holding 3194 * a non-zero value, this is not a problem because data is not 3195 * being buffered for *bytes* and is sent as it is received. 3196 * Return 3197 * 0 3198 * 3199 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3200 * Description 3201 * For socket policies, prevent the execution of the verdict eBPF 3202 * program for message *msg* until *bytes* (byte number) have been 3203 * accumulated. 3204 * 3205 * This can be used when one needs a specific number of bytes 3206 * before a verdict can be assigned, even if the data spans 3207 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3208 * case would be a user calling **sendmsg**\ () repeatedly with 3209 * 1-byte long message segments. Obviously, this is bad for 3210 * performance, but it is still valid. If the eBPF program needs 3211 * *bytes* bytes to validate a header, this helper can be used to 3212 * prevent the eBPF program to be called again until *bytes* have 3213 * been accumulated. 3214 * Return 3215 * 0 3216 * 3217 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3218 * Description 3219 * For socket policies, pull in non-linear data from user space 3220 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3221 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3222 * respectively. 3223 * 3224 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3225 * *msg* it can only parse data that the (**data**, **data_end**) 3226 * pointers have already consumed. For **sendmsg**\ () hooks this 3227 * is likely the first scatterlist element. But for calls relying 3228 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3229 * be the range (**0**, **0**) because the data is shared with 3230 * user space and by default the objective is to avoid allowing 3231 * user space to modify data while (or after) eBPF verdict is 3232 * being decided. This helper can be used to pull in data and to 3233 * set the start and end pointer to given values. Data will be 3234 * copied if necessary (i.e. if data was not linear and if start 3235 * and end pointers do not point to the same chunk). 3236 * 3237 * A call to this helper is susceptible to change the underlying 3238 * packet buffer. Therefore, at load time, all checks on pointers 3239 * previously done by the verifier are invalidated and must be 3240 * performed again, if the helper is used in combination with 3241 * direct packet access. 3242 * 3243 * All values for *flags* are reserved for future usage, and must 3244 * be left at zero. 3245 * Return 3246 * 0 on success, or a negative error in case of failure. 3247 * 3248 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3249 * Description 3250 * Bind the socket associated to *ctx* to the address pointed by 3251 * *addr*, of length *addr_len*. This allows for making outgoing 3252 * connection from the desired IP address, which can be useful for 3253 * example when all processes inside a cgroup should use one 3254 * single IP address on a host that has multiple IP configured. 3255 * 3256 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3257 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3258 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3259 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3260 * behavior and lets the kernel efficiently pick up an unused 3261 * port as long as 4-tuple is unique. Passing non-zero port might 3262 * lead to degraded performance. 3263 * Return 3264 * 0 on success, or a negative error in case of failure. 3265 * 3266 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3267 * Description 3268 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3269 * possible to both shrink and grow the packet tail. 3270 * Shrink done via *delta* being a negative integer. 3271 * 3272 * A call to this helper is susceptible to change the underlying 3273 * packet buffer. Therefore, at load time, all checks on pointers 3274 * previously done by the verifier are invalidated and must be 3275 * performed again, if the helper is used in combination with 3276 * direct packet access. 3277 * Return 3278 * 0 on success, or a negative error in case of failure. 3279 * 3280 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3281 * Description 3282 * Retrieve the XFRM state (IP transform framework, see also 3283 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3284 * 3285 * The retrieved value is stored in the **struct bpf_xfrm_state** 3286 * pointed by *xfrm_state* and of length *size*. 3287 * 3288 * All values for *flags* are reserved for future usage, and must 3289 * be left at zero. 3290 * 3291 * This helper is available only if the kernel was compiled with 3292 * **CONFIG_XFRM** configuration option. 3293 * Return 3294 * 0 on success, or a negative error in case of failure. 3295 * 3296 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3297 * Description 3298 * Return a user or a kernel stack in bpf program provided buffer. 3299 * To achieve this, the helper needs *ctx*, which is a pointer 3300 * to the context on which the tracing program is executed. 3301 * To store the stacktrace, the bpf program provides *buf* with 3302 * a nonnegative *size*. 3303 * 3304 * The last argument, *flags*, holds the number of stack frames to 3305 * skip (from 0 to 255), masked with 3306 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3307 * the following flags: 3308 * 3309 * **BPF_F_USER_STACK** 3310 * Collect a user space stack instead of a kernel stack. 3311 * **BPF_F_USER_BUILD_ID** 3312 * Collect (build_id, file_offset) instead of ips for user 3313 * stack, only valid if **BPF_F_USER_STACK** is also 3314 * specified. 3315 * 3316 * *file_offset* is an offset relative to the beginning 3317 * of the executable or shared object file backing the vma 3318 * which the *ip* falls in. It is *not* an offset relative 3319 * to that object's base address. Accordingly, it must be 3320 * adjusted by adding (sh_addr - sh_offset), where 3321 * sh_{addr,offset} correspond to the executable section 3322 * containing *file_offset* in the object, for comparisons 3323 * to symbols' st_value to be valid. 3324 * 3325 * **bpf_get_stack**\ () can collect up to 3326 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3327 * to sufficient large buffer size. Note that 3328 * this limit can be controlled with the **sysctl** program, and 3329 * that it should be manually increased in order to profile long 3330 * user stacks (such as stacks for Java programs). To do so, use: 3331 * 3332 * :: 3333 * 3334 * # sysctl kernel.perf_event_max_stack=<new value> 3335 * Return 3336 * The non-negative copied *buf* length equal to or less than 3337 * *size* on success, or a negative error in case of failure. 3338 * 3339 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3340 * Description 3341 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3342 * it provides an easy way to load *len* bytes from *offset* 3343 * from the packet associated to *skb*, into the buffer pointed 3344 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3345 * a fifth argument *start_header* exists in order to select a 3346 * base offset to start from. *start_header* can be one of: 3347 * 3348 * **BPF_HDR_START_MAC** 3349 * Base offset to load data from is *skb*'s mac header. 3350 * **BPF_HDR_START_NET** 3351 * Base offset to load data from is *skb*'s network header. 3352 * 3353 * In general, "direct packet access" is the preferred method to 3354 * access packet data, however, this helper is in particular useful 3355 * in socket filters where *skb*\ **->data** does not always point 3356 * to the start of the mac header and where "direct packet access" 3357 * is not available. 3358 * Return 3359 * 0 on success, or a negative error in case of failure. 3360 * 3361 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3362 * Description 3363 * Do FIB lookup in kernel tables using parameters in *params*. 3364 * If lookup is successful and result shows packet is to be 3365 * forwarded, the neighbor tables are searched for the nexthop. 3366 * If successful (ie., FIB lookup shows forwarding and nexthop 3367 * is resolved), the nexthop address is returned in ipv4_dst 3368 * or ipv6_dst based on family, smac is set to mac address of 3369 * egress device, dmac is set to nexthop mac address, rt_metric 3370 * is set to metric from route (IPv4/IPv6 only), and ifindex 3371 * is set to the device index of the nexthop from the FIB lookup. 3372 * 3373 * *plen* argument is the size of the passed in struct. 3374 * *flags* argument can be a combination of one or more of the 3375 * following values: 3376 * 3377 * **BPF_FIB_LOOKUP_DIRECT** 3378 * Do a direct table lookup vs full lookup using FIB 3379 * rules. 3380 * **BPF_FIB_LOOKUP_TBID** 3381 * Used with BPF_FIB_LOOKUP_DIRECT. 3382 * Use the routing table ID present in *params*->tbid 3383 * for the fib lookup. 3384 * **BPF_FIB_LOOKUP_OUTPUT** 3385 * Perform lookup from an egress perspective (default is 3386 * ingress). 3387 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3388 * Skip the neighbour table lookup. *params*->dmac 3389 * and *params*->smac will not be set as output. A common 3390 * use case is to call **bpf_redirect_neigh**\ () after 3391 * doing **bpf_fib_lookup**\ (). 3392 * **BPF_FIB_LOOKUP_SRC** 3393 * Derive and set source IP addr in *params*->ipv{4,6}_src 3394 * for the nexthop. If the src addr cannot be derived, 3395 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3396 * case, *params*->dmac and *params*->smac are not set either. 3397 * 3398 * *ctx* is either **struct xdp_md** for XDP programs or 3399 * **struct sk_buff** tc cls_act programs. 3400 * Return 3401 * * < 0 if any input argument is invalid 3402 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3403 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3404 * packet is not forwarded or needs assist from full stack 3405 * 3406 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3407 * was exceeded and output params->mtu_result contains the MTU. 3408 * 3409 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3410 * Description 3411 * Add an entry to, or update a sockhash *map* referencing sockets. 3412 * The *skops* is used as a new value for the entry associated to 3413 * *key*. *flags* is one of: 3414 * 3415 * **BPF_NOEXIST** 3416 * The entry for *key* must not exist in the map. 3417 * **BPF_EXIST** 3418 * The entry for *key* must already exist in the map. 3419 * **BPF_ANY** 3420 * No condition on the existence of the entry for *key*. 3421 * 3422 * If the *map* has eBPF programs (parser and verdict), those will 3423 * be inherited by the socket being added. If the socket is 3424 * already attached to eBPF programs, this results in an error. 3425 * Return 3426 * 0 on success, or a negative error in case of failure. 3427 * 3428 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3429 * Description 3430 * This helper is used in programs implementing policies at the 3431 * socket level. If the message *msg* is allowed to pass (i.e. if 3432 * the verdict eBPF program returns **SK_PASS**), redirect it to 3433 * the socket referenced by *map* (of type 3434 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3435 * egress interfaces can be used for redirection. The 3436 * **BPF_F_INGRESS** value in *flags* is used to make the 3437 * distinction (ingress path is selected if the flag is present, 3438 * egress path otherwise). This is the only flag supported for now. 3439 * Return 3440 * **SK_PASS** on success, or **SK_DROP** on error. 3441 * 3442 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3443 * Description 3444 * This helper is used in programs implementing policies at the 3445 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3446 * if the verdict eBPF program returns **SK_PASS**), redirect it 3447 * to the socket referenced by *map* (of type 3448 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3449 * egress interfaces can be used for redirection. The 3450 * **BPF_F_INGRESS** value in *flags* is used to make the 3451 * distinction (ingress path is selected if the flag is present, 3452 * egress otherwise). This is the only flag supported for now. 3453 * Return 3454 * **SK_PASS** on success, or **SK_DROP** on error. 3455 * 3456 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3457 * Description 3458 * Encapsulate the packet associated to *skb* within a Layer 3 3459 * protocol header. This header is provided in the buffer at 3460 * address *hdr*, with *len* its size in bytes. *type* indicates 3461 * the protocol of the header and can be one of: 3462 * 3463 * **BPF_LWT_ENCAP_SEG6** 3464 * IPv6 encapsulation with Segment Routing Header 3465 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3466 * the IPv6 header is computed by the kernel. 3467 * **BPF_LWT_ENCAP_SEG6_INLINE** 3468 * Only works if *skb* contains an IPv6 packet. Insert a 3469 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3470 * the IPv6 header. 3471 * **BPF_LWT_ENCAP_IP** 3472 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3473 * must be IPv4 or IPv6, followed by zero or more 3474 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3475 * total bytes in all prepended headers. Please note that 3476 * if **skb_is_gso**\ (*skb*) is true, no more than two 3477 * headers can be prepended, and the inner header, if 3478 * present, should be either GRE or UDP/GUE. 3479 * 3480 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3481 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3482 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3483 * **BPF_PROG_TYPE_LWT_XMIT**. 3484 * 3485 * A call to this helper is susceptible to change the underlying 3486 * packet buffer. Therefore, at load time, all checks on pointers 3487 * previously done by the verifier are invalidated and must be 3488 * performed again, if the helper is used in combination with 3489 * direct packet access. 3490 * Return 3491 * 0 on success, or a negative error in case of failure. 3492 * 3493 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3494 * Description 3495 * Store *len* bytes from address *from* into the packet 3496 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3497 * inside the outermost IPv6 Segment Routing Header can be 3498 * modified through this helper. 3499 * 3500 * A call to this helper is susceptible to change the underlying 3501 * packet buffer. Therefore, at load time, all checks on pointers 3502 * previously done by the verifier are invalidated and must be 3503 * performed again, if the helper is used in combination with 3504 * direct packet access. 3505 * Return 3506 * 0 on success, or a negative error in case of failure. 3507 * 3508 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3509 * Description 3510 * Adjust the size allocated to TLVs in the outermost IPv6 3511 * Segment Routing Header contained in the packet associated to 3512 * *skb*, at position *offset* by *delta* bytes. Only offsets 3513 * after the segments are accepted. *delta* can be as well 3514 * positive (growing) as negative (shrinking). 3515 * 3516 * A call to this helper is susceptible to change the underlying 3517 * packet buffer. Therefore, at load time, all checks on pointers 3518 * previously done by the verifier are invalidated and must be 3519 * performed again, if the helper is used in combination with 3520 * direct packet access. 3521 * Return 3522 * 0 on success, or a negative error in case of failure. 3523 * 3524 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3525 * Description 3526 * Apply an IPv6 Segment Routing action of type *action* to the 3527 * packet associated to *skb*. Each action takes a parameter 3528 * contained at address *param*, and of length *param_len* bytes. 3529 * *action* can be one of: 3530 * 3531 * **SEG6_LOCAL_ACTION_END_X** 3532 * End.X action: Endpoint with Layer-3 cross-connect. 3533 * Type of *param*: **struct in6_addr**. 3534 * **SEG6_LOCAL_ACTION_END_T** 3535 * End.T action: Endpoint with specific IPv6 table lookup. 3536 * Type of *param*: **int**. 3537 * **SEG6_LOCAL_ACTION_END_B6** 3538 * End.B6 action: Endpoint bound to an SRv6 policy. 3539 * Type of *param*: **struct ipv6_sr_hdr**. 3540 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3541 * End.B6.Encap action: Endpoint bound to an SRv6 3542 * encapsulation policy. 3543 * Type of *param*: **struct ipv6_sr_hdr**. 3544 * 3545 * A call to this helper is susceptible to change the underlying 3546 * packet buffer. Therefore, at load time, all checks on pointers 3547 * previously done by the verifier are invalidated and must be 3548 * performed again, if the helper is used in combination with 3549 * direct packet access. 3550 * Return 3551 * 0 on success, or a negative error in case of failure. 3552 * 3553 * long bpf_rc_repeat(void *ctx) 3554 * Description 3555 * This helper is used in programs implementing IR decoding, to 3556 * report a successfully decoded repeat key message. This delays 3557 * the generation of a key up event for previously generated 3558 * key down event. 3559 * 3560 * Some IR protocols like NEC have a special IR message for 3561 * repeating last button, for when a button is held down. 3562 * 3563 * The *ctx* should point to the lirc sample as passed into 3564 * the program. 3565 * 3566 * This helper is only available is the kernel was compiled with 3567 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3568 * "**y**". 3569 * Return 3570 * 0 3571 * 3572 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3573 * Description 3574 * This helper is used in programs implementing IR decoding, to 3575 * report a successfully decoded key press with *scancode*, 3576 * *toggle* value in the given *protocol*. The scancode will be 3577 * translated to a keycode using the rc keymap, and reported as 3578 * an input key down event. After a period a key up event is 3579 * generated. This period can be extended by calling either 3580 * **bpf_rc_keydown**\ () again with the same values, or calling 3581 * **bpf_rc_repeat**\ (). 3582 * 3583 * Some protocols include a toggle bit, in case the button was 3584 * released and pressed again between consecutive scancodes. 3585 * 3586 * The *ctx* should point to the lirc sample as passed into 3587 * the program. 3588 * 3589 * The *protocol* is the decoded protocol number (see 3590 * **enum rc_proto** for some predefined values). 3591 * 3592 * This helper is only available is the kernel was compiled with 3593 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3594 * "**y**". 3595 * Return 3596 * 0 3597 * 3598 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3599 * Description 3600 * Return the cgroup v2 id of the socket associated with the *skb*. 3601 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3602 * helper for cgroup v1 by providing a tag resp. identifier that 3603 * can be matched on or used for map lookups e.g. to implement 3604 * policy. The cgroup v2 id of a given path in the hierarchy is 3605 * exposed in user space through the f_handle API in order to get 3606 * to the same 64-bit id. 3607 * 3608 * This helper can be used on TC egress path, but not on ingress, 3609 * and is available only if the kernel was compiled with the 3610 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3611 * Return 3612 * The id is returned or 0 in case the id could not be retrieved. 3613 * 3614 * u64 bpf_get_current_cgroup_id(void) 3615 * Description 3616 * Get the current cgroup id based on the cgroup within which 3617 * the current task is running. 3618 * Return 3619 * A 64-bit integer containing the current cgroup id based 3620 * on the cgroup within which the current task is running. 3621 * 3622 * void *bpf_get_local_storage(void *map, u64 flags) 3623 * Description 3624 * Get the pointer to the local storage area. 3625 * The type and the size of the local storage is defined 3626 * by the *map* argument. 3627 * The *flags* meaning is specific for each map type, 3628 * and has to be 0 for cgroup local storage. 3629 * 3630 * Depending on the BPF program type, a local storage area 3631 * can be shared between multiple instances of the BPF program, 3632 * running simultaneously. 3633 * 3634 * A user should care about the synchronization by himself. 3635 * For example, by using the **BPF_ATOMIC** instructions to alter 3636 * the shared data. 3637 * Return 3638 * A pointer to the local storage area. 3639 * 3640 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3641 * Description 3642 * Select a **SO_REUSEPORT** socket from a 3643 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3644 * It checks the selected socket is matching the incoming 3645 * request in the socket buffer. 3646 * Return 3647 * 0 on success, or a negative error in case of failure. 3648 * 3649 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3650 * Description 3651 * Return id of cgroup v2 that is ancestor of cgroup associated 3652 * with the *skb* at the *ancestor_level*. The root cgroup is at 3653 * *ancestor_level* zero and each step down the hierarchy 3654 * increments the level. If *ancestor_level* == level of cgroup 3655 * associated with *skb*, then return value will be same as that 3656 * of **bpf_skb_cgroup_id**\ (). 3657 * 3658 * The helper is useful to implement policies based on cgroups 3659 * that are upper in hierarchy than immediate cgroup associated 3660 * with *skb*. 3661 * 3662 * The format of returned id and helper limitations are same as in 3663 * **bpf_skb_cgroup_id**\ (). 3664 * Return 3665 * The id is returned or 0 in case the id could not be retrieved. 3666 * 3667 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3668 * Description 3669 * Look for TCP socket matching *tuple*, optionally in a child 3670 * network namespace *netns*. The return value must be checked, 3671 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3672 * 3673 * The *ctx* should point to the context of the program, such as 3674 * the skb or socket (depending on the hook in use). This is used 3675 * to determine the base network namespace for the lookup. 3676 * 3677 * *tuple_size* must be one of: 3678 * 3679 * **sizeof**\ (*tuple*\ **->ipv4**) 3680 * Look for an IPv4 socket. 3681 * **sizeof**\ (*tuple*\ **->ipv6**) 3682 * Look for an IPv6 socket. 3683 * 3684 * If the *netns* is a negative signed 32-bit integer, then the 3685 * socket lookup table in the netns associated with the *ctx* 3686 * will be used. For the TC hooks, this is the netns of the device 3687 * in the skb. For socket hooks, this is the netns of the socket. 3688 * If *netns* is any other signed 32-bit value greater than or 3689 * equal to zero then it specifies the ID of the netns relative to 3690 * the netns associated with the *ctx*. *netns* values beyond the 3691 * range of 32-bit integers are reserved for future use. 3692 * 3693 * All values for *flags* are reserved for future usage, and must 3694 * be left at zero. 3695 * 3696 * This helper is available only if the kernel was compiled with 3697 * **CONFIG_NET** configuration option. 3698 * Return 3699 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3700 * For sockets with reuseport option, the **struct bpf_sock** 3701 * result is from *reuse*\ **->socks**\ [] using the hash of the 3702 * tuple. 3703 * 3704 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3705 * Description 3706 * Look for UDP socket matching *tuple*, optionally in a child 3707 * network namespace *netns*. The return value must be checked, 3708 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3709 * 3710 * The *ctx* should point to the context of the program, such as 3711 * the skb or socket (depending on the hook in use). This is used 3712 * to determine the base network namespace for the lookup. 3713 * 3714 * *tuple_size* must be one of: 3715 * 3716 * **sizeof**\ (*tuple*\ **->ipv4**) 3717 * Look for an IPv4 socket. 3718 * **sizeof**\ (*tuple*\ **->ipv6**) 3719 * Look for an IPv6 socket. 3720 * 3721 * If the *netns* is a negative signed 32-bit integer, then the 3722 * socket lookup table in the netns associated with the *ctx* 3723 * will be used. For the TC hooks, this is the netns of the device 3724 * in the skb. For socket hooks, this is the netns of the socket. 3725 * If *netns* is any other signed 32-bit value greater than or 3726 * equal to zero then it specifies the ID of the netns relative to 3727 * the netns associated with the *ctx*. *netns* values beyond the 3728 * range of 32-bit integers are reserved for future use. 3729 * 3730 * All values for *flags* are reserved for future usage, and must 3731 * be left at zero. 3732 * 3733 * This helper is available only if the kernel was compiled with 3734 * **CONFIG_NET** configuration option. 3735 * Return 3736 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3737 * For sockets with reuseport option, the **struct bpf_sock** 3738 * result is from *reuse*\ **->socks**\ [] using the hash of the 3739 * tuple. 3740 * 3741 * long bpf_sk_release(void *sock) 3742 * Description 3743 * Release the reference held by *sock*. *sock* must be a 3744 * non-**NULL** pointer that was returned from 3745 * **bpf_sk_lookup_xxx**\ (). 3746 * Return 3747 * 0 on success, or a negative error in case of failure. 3748 * 3749 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3750 * Description 3751 * Push an element *value* in *map*. *flags* is one of: 3752 * 3753 * **BPF_EXIST** 3754 * If the queue/stack is full, the oldest element is 3755 * removed to make room for this. 3756 * Return 3757 * 0 on success, or a negative error in case of failure. 3758 * 3759 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3760 * Description 3761 * Pop an element from *map*. 3762 * Return 3763 * 0 on success, or a negative error in case of failure. 3764 * 3765 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3766 * Description 3767 * Get an element from *map* without removing it. 3768 * Return 3769 * 0 on success, or a negative error in case of failure. 3770 * 3771 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3772 * Description 3773 * For socket policies, insert *len* bytes into *msg* at offset 3774 * *start*. 3775 * 3776 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3777 * *msg* it may want to insert metadata or options into the *msg*. 3778 * This can later be read and used by any of the lower layer BPF 3779 * hooks. 3780 * 3781 * This helper may fail if under memory pressure (a malloc 3782 * fails) in these cases BPF programs will get an appropriate 3783 * error and BPF programs will need to handle them. 3784 * Return 3785 * 0 on success, or a negative error in case of failure. 3786 * 3787 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3788 * Description 3789 * Will remove *len* bytes from a *msg* starting at byte *start*. 3790 * This may result in **ENOMEM** errors under certain situations if 3791 * an allocation and copy are required due to a full ring buffer. 3792 * However, the helper will try to avoid doing the allocation 3793 * if possible. Other errors can occur if input parameters are 3794 * invalid either due to *start* byte not being valid part of *msg* 3795 * payload and/or *pop* value being to large. 3796 * Return 3797 * 0 on success, or a negative error in case of failure. 3798 * 3799 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3800 * Description 3801 * This helper is used in programs implementing IR decoding, to 3802 * report a successfully decoded pointer movement. 3803 * 3804 * The *ctx* should point to the lirc sample as passed into 3805 * the program. 3806 * 3807 * This helper is only available is the kernel was compiled with 3808 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3809 * "**y**". 3810 * Return 3811 * 0 3812 * 3813 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3814 * Description 3815 * Acquire a spinlock represented by the pointer *lock*, which is 3816 * stored as part of a value of a map. Taking the lock allows to 3817 * safely update the rest of the fields in that value. The 3818 * spinlock can (and must) later be released with a call to 3819 * **bpf_spin_unlock**\ (\ *lock*\ ). 3820 * 3821 * Spinlocks in BPF programs come with a number of restrictions 3822 * and constraints: 3823 * 3824 * * **bpf_spin_lock** objects are only allowed inside maps of 3825 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3826 * list could be extended in the future). 3827 * * BTF description of the map is mandatory. 3828 * * The BPF program can take ONE lock at a time, since taking two 3829 * or more could cause dead locks. 3830 * * Only one **struct bpf_spin_lock** is allowed per map element. 3831 * * When the lock is taken, calls (either BPF to BPF or helpers) 3832 * are not allowed. 3833 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3834 * allowed inside a spinlock-ed region. 3835 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3836 * the lock, on all execution paths, before it returns. 3837 * * The BPF program can access **struct bpf_spin_lock** only via 3838 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3839 * helpers. Loading or storing data into the **struct 3840 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3841 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3842 * of the map value must be a struct and have **struct 3843 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3844 * Nested lock inside another struct is not allowed. 3845 * * The **struct bpf_spin_lock** *lock* field in a map value must 3846 * be aligned on a multiple of 4 bytes in that value. 3847 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3848 * the **bpf_spin_lock** field to user space. 3849 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3850 * a BPF program, do not update the **bpf_spin_lock** field. 3851 * * **bpf_spin_lock** cannot be on the stack or inside a 3852 * networking packet (it can only be inside of a map values). 3853 * * **bpf_spin_lock** is available to root only. 3854 * * Tracing programs and socket filter programs cannot use 3855 * **bpf_spin_lock**\ () due to insufficient preemption checks 3856 * (but this may change in the future). 3857 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3858 * Return 3859 * 0 3860 * 3861 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3862 * Description 3863 * Release the *lock* previously locked by a call to 3864 * **bpf_spin_lock**\ (\ *lock*\ ). 3865 * Return 3866 * 0 3867 * 3868 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3869 * Description 3870 * This helper gets a **struct bpf_sock** pointer such 3871 * that all the fields in this **bpf_sock** can be accessed. 3872 * Return 3873 * A **struct bpf_sock** pointer on success, or **NULL** in 3874 * case of failure. 3875 * 3876 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3877 * Description 3878 * This helper gets a **struct bpf_tcp_sock** pointer from a 3879 * **struct bpf_sock** pointer. 3880 * Return 3881 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3882 * case of failure. 3883 * 3884 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3885 * Description 3886 * Set ECN (Explicit Congestion Notification) field of IP header 3887 * to **CE** (Congestion Encountered) if current value is **ECT** 3888 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3889 * and IPv4. 3890 * Return 3891 * 1 if the **CE** flag is set (either by the current helper call 3892 * or because it was already present), 0 if it is not set. 3893 * 3894 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3895 * Description 3896 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3897 * **bpf_sk_release**\ () is unnecessary and not allowed. 3898 * Return 3899 * A **struct bpf_sock** pointer on success, or **NULL** in 3900 * case of failure. 3901 * 3902 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3903 * Description 3904 * Look for TCP socket matching *tuple*, optionally in a child 3905 * network namespace *netns*. The return value must be checked, 3906 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3907 * 3908 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3909 * that it also returns timewait or request sockets. Use 3910 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3911 * full structure. 3912 * 3913 * This helper is available only if the kernel was compiled with 3914 * **CONFIG_NET** configuration option. 3915 * Return 3916 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3917 * For sockets with reuseport option, the **struct bpf_sock** 3918 * result is from *reuse*\ **->socks**\ [] using the hash of the 3919 * tuple. 3920 * 3921 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3922 * Description 3923 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3924 * the listening socket in *sk*. 3925 * 3926 * *iph* points to the start of the IPv4 or IPv6 header, while 3927 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3928 * **sizeof**\ (**struct ipv6hdr**). 3929 * 3930 * *th* points to the start of the TCP header, while *th_len* 3931 * contains the length of the TCP header (at least 3932 * **sizeof**\ (**struct tcphdr**)). 3933 * Return 3934 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3935 * error otherwise. 3936 * 3937 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3938 * Description 3939 * Get name of sysctl in /proc/sys/ and copy it into provided by 3940 * program buffer *buf* of size *buf_len*. 3941 * 3942 * The buffer is always NUL terminated, unless it's zero-sized. 3943 * 3944 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3945 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3946 * only (e.g. "tcp_mem"). 3947 * Return 3948 * Number of character copied (not including the trailing NUL). 3949 * 3950 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3951 * truncated name in this case). 3952 * 3953 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3954 * Description 3955 * Get current value of sysctl as it is presented in /proc/sys 3956 * (incl. newline, etc), and copy it as a string into provided 3957 * by program buffer *buf* of size *buf_len*. 3958 * 3959 * The whole value is copied, no matter what file position user 3960 * space issued e.g. sys_read at. 3961 * 3962 * The buffer is always NUL terminated, unless it's zero-sized. 3963 * Return 3964 * Number of character copied (not including the trailing NUL). 3965 * 3966 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3967 * truncated name in this case). 3968 * 3969 * **-EINVAL** if current value was unavailable, e.g. because 3970 * sysctl is uninitialized and read returns -EIO for it. 3971 * 3972 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3973 * Description 3974 * Get new value being written by user space to sysctl (before 3975 * the actual write happens) and copy it as a string into 3976 * provided by program buffer *buf* of size *buf_len*. 3977 * 3978 * User space may write new value at file position > 0. 3979 * 3980 * The buffer is always NUL terminated, unless it's zero-sized. 3981 * Return 3982 * Number of character copied (not including the trailing NUL). 3983 * 3984 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3985 * truncated name in this case). 3986 * 3987 * **-EINVAL** if sysctl is being read. 3988 * 3989 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3990 * Description 3991 * Override new value being written by user space to sysctl with 3992 * value provided by program in buffer *buf* of size *buf_len*. 3993 * 3994 * *buf* should contain a string in same form as provided by user 3995 * space on sysctl write. 3996 * 3997 * User space may write new value at file position > 0. To override 3998 * the whole sysctl value file position should be set to zero. 3999 * Return 4000 * 0 on success. 4001 * 4002 * **-E2BIG** if the *buf_len* is too big. 4003 * 4004 * **-EINVAL** if sysctl is being read. 4005 * 4006 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 4007 * Description 4008 * Convert the initial part of the string from buffer *buf* of 4009 * size *buf_len* to a long integer according to the given base 4010 * and save the result in *res*. 4011 * 4012 * The string may begin with an arbitrary amount of white space 4013 * (as determined by **isspace**\ (3)) followed by a single 4014 * optional '**-**' sign. 4015 * 4016 * Five least significant bits of *flags* encode base, other bits 4017 * are currently unused. 4018 * 4019 * Base must be either 8, 10, 16 or 0 to detect it automatically 4020 * similar to user space **strtol**\ (3). 4021 * Return 4022 * Number of characters consumed on success. Must be positive but 4023 * no more than *buf_len*. 4024 * 4025 * **-EINVAL** if no valid digits were found or unsupported base 4026 * was provided. 4027 * 4028 * **-ERANGE** if resulting value was out of range. 4029 * 4030 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 4031 * Description 4032 * Convert the initial part of the string from buffer *buf* of 4033 * size *buf_len* to an unsigned long integer according to the 4034 * given base and save the result in *res*. 4035 * 4036 * The string may begin with an arbitrary amount of white space 4037 * (as determined by **isspace**\ (3)). 4038 * 4039 * Five least significant bits of *flags* encode base, other bits 4040 * are currently unused. 4041 * 4042 * Base must be either 8, 10, 16 or 0 to detect it automatically 4043 * similar to user space **strtoul**\ (3). 4044 * Return 4045 * Number of characters consumed on success. Must be positive but 4046 * no more than *buf_len*. 4047 * 4048 * **-EINVAL** if no valid digits were found or unsupported base 4049 * was provided. 4050 * 4051 * **-ERANGE** if resulting value was out of range. 4052 * 4053 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4054 * Description 4055 * Get a bpf-local-storage from a *sk*. 4056 * 4057 * Logically, it could be thought of getting the value from 4058 * a *map* with *sk* as the **key**. From this 4059 * perspective, the usage is not much different from 4060 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4061 * helper enforces the key must be a full socket and the map must 4062 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4063 * 4064 * Underneath, the value is stored locally at *sk* instead of 4065 * the *map*. The *map* is used as the bpf-local-storage 4066 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4067 * searched against all bpf-local-storages residing at *sk*. 4068 * 4069 * *sk* is a kernel **struct sock** pointer for LSM program. 4070 * *sk* is a **struct bpf_sock** pointer for other program types. 4071 * 4072 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4073 * used such that a new bpf-local-storage will be 4074 * created if one does not exist. *value* can be used 4075 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4076 * the initial value of a bpf-local-storage. If *value* is 4077 * **NULL**, the new bpf-local-storage will be zero initialized. 4078 * Return 4079 * A bpf-local-storage pointer is returned on success. 4080 * 4081 * **NULL** if not found or there was an error in adding 4082 * a new bpf-local-storage. 4083 * 4084 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4085 * Description 4086 * Delete a bpf-local-storage from a *sk*. 4087 * Return 4088 * 0 on success. 4089 * 4090 * **-ENOENT** if the bpf-local-storage cannot be found. 4091 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4092 * 4093 * long bpf_send_signal(u32 sig) 4094 * Description 4095 * Send signal *sig* to the process of the current task. 4096 * The signal may be delivered to any of this process's threads. 4097 * Return 4098 * 0 on success or successfully queued. 4099 * 4100 * **-EBUSY** if work queue under nmi is full. 4101 * 4102 * **-EINVAL** if *sig* is invalid. 4103 * 4104 * **-EPERM** if no permission to send the *sig*. 4105 * 4106 * **-EAGAIN** if bpf program can try again. 4107 * 4108 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4109 * Description 4110 * Try to issue a SYN cookie for the packet with corresponding 4111 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4112 * 4113 * *iph* points to the start of the IPv4 or IPv6 header, while 4114 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4115 * **sizeof**\ (**struct ipv6hdr**). 4116 * 4117 * *th* points to the start of the TCP header, while *th_len* 4118 * contains the length of the TCP header with options (at least 4119 * **sizeof**\ (**struct tcphdr**)). 4120 * Return 4121 * On success, lower 32 bits hold the generated SYN cookie in 4122 * followed by 16 bits which hold the MSS value for that cookie, 4123 * and the top 16 bits are unused. 4124 * 4125 * On failure, the returned value is one of the following: 4126 * 4127 * **-EINVAL** SYN cookie cannot be issued due to error 4128 * 4129 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4130 * 4131 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4132 * 4133 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4134 * 4135 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4136 * Description 4137 * Write raw *data* blob into a special BPF perf event held by 4138 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4139 * event must have the following attributes: **PERF_SAMPLE_RAW** 4140 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4141 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4142 * 4143 * The *flags* are used to indicate the index in *map* for which 4144 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4145 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4146 * to indicate that the index of the current CPU core should be 4147 * used. 4148 * 4149 * The value to write, of *size*, is passed through eBPF stack and 4150 * pointed by *data*. 4151 * 4152 * *ctx* is a pointer to in-kernel struct sk_buff. 4153 * 4154 * This helper is similar to **bpf_perf_event_output**\ () but 4155 * restricted to raw_tracepoint bpf programs. 4156 * Return 4157 * 0 on success, or a negative error in case of failure. 4158 * 4159 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4160 * Description 4161 * Safely attempt to read *size* bytes from user space address 4162 * *unsafe_ptr* and store the data in *dst*. 4163 * Return 4164 * 0 on success, or a negative error in case of failure. 4165 * 4166 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4167 * Description 4168 * Safely attempt to read *size* bytes from kernel space address 4169 * *unsafe_ptr* and store the data in *dst*. 4170 * Return 4171 * 0 on success, or a negative error in case of failure. 4172 * 4173 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4174 * Description 4175 * Copy a NUL terminated string from an unsafe user address 4176 * *unsafe_ptr* to *dst*. The *size* should include the 4177 * terminating NUL byte. In case the string length is smaller than 4178 * *size*, the target is not padded with further NUL bytes. If the 4179 * string length is larger than *size*, just *size*-1 bytes are 4180 * copied and the last byte is set to NUL. 4181 * 4182 * On success, returns the number of bytes that were written, 4183 * including the terminal NUL. This makes this helper useful in 4184 * tracing programs for reading strings, and more importantly to 4185 * get its length at runtime. See the following snippet: 4186 * 4187 * :: 4188 * 4189 * SEC("kprobe/sys_open") 4190 * void bpf_sys_open(struct pt_regs *ctx) 4191 * { 4192 * char buf[PATHLEN]; // PATHLEN is defined to 256 4193 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4194 * ctx->di); 4195 * 4196 * // Consume buf, for example push it to 4197 * // userspace via bpf_perf_event_output(); we 4198 * // can use res (the string length) as event 4199 * // size, after checking its boundaries. 4200 * } 4201 * 4202 * In comparison, using **bpf_probe_read_user**\ () helper here 4203 * instead to read the string would require to estimate the length 4204 * at compile time, and would often result in copying more memory 4205 * than necessary. 4206 * 4207 * Another useful use case is when parsing individual process 4208 * arguments or individual environment variables navigating 4209 * *current*\ **->mm->arg_start** and *current*\ 4210 * **->mm->env_start**: using this helper and the return value, 4211 * one can quickly iterate at the right offset of the memory area. 4212 * Return 4213 * On success, the strictly positive length of the output string, 4214 * including the trailing NUL character. On error, a negative 4215 * value. 4216 * 4217 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4218 * Description 4219 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4220 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4221 * Return 4222 * On success, the strictly positive length of the string, including 4223 * the trailing NUL character. On error, a negative value. 4224 * 4225 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4226 * Description 4227 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4228 * *rcv_nxt* is the ack_seq to be sent out. 4229 * Return 4230 * 0 on success, or a negative error in case of failure. 4231 * 4232 * long bpf_send_signal_thread(u32 sig) 4233 * Description 4234 * Send signal *sig* to the thread corresponding to the current task. 4235 * Return 4236 * 0 on success or successfully queued. 4237 * 4238 * **-EBUSY** if work queue under nmi is full. 4239 * 4240 * **-EINVAL** if *sig* is invalid. 4241 * 4242 * **-EPERM** if no permission to send the *sig*. 4243 * 4244 * **-EAGAIN** if bpf program can try again. 4245 * 4246 * u64 bpf_jiffies64(void) 4247 * Description 4248 * Obtain the 64bit jiffies 4249 * Return 4250 * The 64 bit jiffies 4251 * 4252 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4253 * Description 4254 * For an eBPF program attached to a perf event, retrieve the 4255 * branch records (**struct perf_branch_entry**) associated to *ctx* 4256 * and store it in the buffer pointed by *buf* up to size 4257 * *size* bytes. 4258 * Return 4259 * On success, number of bytes written to *buf*. On error, a 4260 * negative value. 4261 * 4262 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4263 * instead return the number of bytes required to store all the 4264 * branch entries. If this flag is set, *buf* may be NULL. 4265 * 4266 * **-EINVAL** if arguments invalid or **size** not a multiple 4267 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4268 * 4269 * **-ENOENT** if architecture does not support branch records. 4270 * 4271 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4272 * Description 4273 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4274 * *namespace* will be returned in *nsdata*. 4275 * Return 4276 * 0 on success, or one of the following in case of failure: 4277 * 4278 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4279 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4280 * 4281 * **-ENOENT** if pidns does not exists for the current task. 4282 * 4283 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4284 * Description 4285 * Write raw *data* blob into a special BPF perf event held by 4286 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4287 * event must have the following attributes: **PERF_SAMPLE_RAW** 4288 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4289 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4290 * 4291 * The *flags* are used to indicate the index in *map* for which 4292 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4293 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4294 * to indicate that the index of the current CPU core should be 4295 * used. 4296 * 4297 * The value to write, of *size*, is passed through eBPF stack and 4298 * pointed by *data*. 4299 * 4300 * *ctx* is a pointer to in-kernel struct xdp_buff. 4301 * 4302 * This helper is similar to **bpf_perf_eventoutput**\ () but 4303 * restricted to raw_tracepoint bpf programs. 4304 * Return 4305 * 0 on success, or a negative error in case of failure. 4306 * 4307 * u64 bpf_get_netns_cookie(void *ctx) 4308 * Description 4309 * Retrieve the cookie (generated by the kernel) of the network 4310 * namespace the input *ctx* is associated with. The network 4311 * namespace cookie remains stable for its lifetime and provides 4312 * a global identifier that can be assumed unique. If *ctx* is 4313 * NULL, then the helper returns the cookie for the initial 4314 * network namespace. The cookie itself is very similar to that 4315 * of **bpf_get_socket_cookie**\ () helper, but for network 4316 * namespaces instead of sockets. 4317 * Return 4318 * A 8-byte long opaque number. 4319 * 4320 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4321 * Description 4322 * Return id of cgroup v2 that is ancestor of the cgroup associated 4323 * with the current task at the *ancestor_level*. The root cgroup 4324 * is at *ancestor_level* zero and each step down the hierarchy 4325 * increments the level. If *ancestor_level* == level of cgroup 4326 * associated with the current task, then return value will be the 4327 * same as that of **bpf_get_current_cgroup_id**\ (). 4328 * 4329 * The helper is useful to implement policies based on cgroups 4330 * that are upper in hierarchy than immediate cgroup associated 4331 * with the current task. 4332 * 4333 * The format of returned id and helper limitations are same as in 4334 * **bpf_get_current_cgroup_id**\ (). 4335 * Return 4336 * The id is returned or 0 in case the id could not be retrieved. 4337 * 4338 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4339 * Description 4340 * Helper is overloaded depending on BPF program type. This 4341 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4342 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4343 * 4344 * Assign the *sk* to the *skb*. When combined with appropriate 4345 * routing configuration to receive the packet towards the socket, 4346 * will cause *skb* to be delivered to the specified socket. 4347 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4348 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4349 * interfere with successful delivery to the socket. 4350 * 4351 * This operation is only valid from TC ingress path. 4352 * 4353 * The *flags* argument must be zero. 4354 * Return 4355 * 0 on success, or a negative error in case of failure: 4356 * 4357 * **-EINVAL** if specified *flags* are not supported. 4358 * 4359 * **-ENOENT** if the socket is unavailable for assignment. 4360 * 4361 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4362 * 4363 * **-EOPNOTSUPP** if the operation is not supported, for example 4364 * a call from outside of TC ingress. 4365 * 4366 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4367 * Description 4368 * Helper is overloaded depending on BPF program type. This 4369 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4370 * 4371 * Select the *sk* as a result of a socket lookup. 4372 * 4373 * For the operation to succeed passed socket must be compatible 4374 * with the packet description provided by the *ctx* object. 4375 * 4376 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4377 * be an exact match. While IP family (**AF_INET** or 4378 * **AF_INET6**) must be compatible, that is IPv6 sockets 4379 * that are not v6-only can be selected for IPv4 packets. 4380 * 4381 * Only TCP listeners and UDP unconnected sockets can be 4382 * selected. *sk* can also be NULL to reset any previous 4383 * selection. 4384 * 4385 * *flags* argument can combination of following values: 4386 * 4387 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4388 * socket selection, potentially done by a BPF program 4389 * that ran before us. 4390 * 4391 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4392 * load-balancing within reuseport group for the socket 4393 * being selected. 4394 * 4395 * On success *ctx->sk* will point to the selected socket. 4396 * 4397 * Return 4398 * 0 on success, or a negative errno in case of failure. 4399 * 4400 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4401 * not compatible with packet family (*ctx->family*). 4402 * 4403 * * **-EEXIST** if socket has been already selected, 4404 * potentially by another program, and 4405 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4406 * 4407 * * **-EINVAL** if unsupported flags were specified. 4408 * 4409 * * **-EPROTOTYPE** if socket L4 protocol 4410 * (*sk->protocol*) doesn't match packet protocol 4411 * (*ctx->protocol*). 4412 * 4413 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4414 * state (TCP listening or UDP unconnected). 4415 * 4416 * u64 bpf_ktime_get_boot_ns(void) 4417 * Description 4418 * Return the time elapsed since system boot, in nanoseconds. 4419 * Does include the time the system was suspended. 4420 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4421 * Return 4422 * Current *ktime*. 4423 * 4424 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4425 * Description 4426 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4427 * out the format string. 4428 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4429 * the format string itself. The *data* and *data_len* are format string 4430 * arguments. The *data* are a **u64** array and corresponding format string 4431 * values are stored in the array. For strings and pointers where pointees 4432 * are accessed, only the pointer values are stored in the *data* array. 4433 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4434 * 4435 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4436 * Reading kernel memory may fail due to either invalid address or 4437 * valid address but requiring a major memory fault. If reading kernel memory 4438 * fails, the string for **%s** will be an empty string, and the ip 4439 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4440 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4441 * Return 4442 * 0 on success, or a negative error in case of failure: 4443 * 4444 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4445 * by returning 1 from bpf program. 4446 * 4447 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4448 * 4449 * **-E2BIG** if *fmt* contains too many format specifiers. 4450 * 4451 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4452 * 4453 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4454 * Description 4455 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4456 * The *m* represents the seq_file. The *data* and *len* represent the 4457 * data to write in bytes. 4458 * Return 4459 * 0 on success, or a negative error in case of failure: 4460 * 4461 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4462 * 4463 * u64 bpf_sk_cgroup_id(void *sk) 4464 * Description 4465 * Return the cgroup v2 id of the socket *sk*. 4466 * 4467 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4468 * returned from **bpf_sk_lookup_xxx**\ (), 4469 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4470 * same as in **bpf_skb_cgroup_id**\ (). 4471 * 4472 * This helper is available only if the kernel was compiled with 4473 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4474 * Return 4475 * The id is returned or 0 in case the id could not be retrieved. 4476 * 4477 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4478 * Description 4479 * Return id of cgroup v2 that is ancestor of cgroup associated 4480 * with the *sk* at the *ancestor_level*. The root cgroup is at 4481 * *ancestor_level* zero and each step down the hierarchy 4482 * increments the level. If *ancestor_level* == level of cgroup 4483 * associated with *sk*, then return value will be same as that 4484 * of **bpf_sk_cgroup_id**\ (). 4485 * 4486 * The helper is useful to implement policies based on cgroups 4487 * that are upper in hierarchy than immediate cgroup associated 4488 * with *sk*. 4489 * 4490 * The format of returned id and helper limitations are same as in 4491 * **bpf_sk_cgroup_id**\ (). 4492 * Return 4493 * The id is returned or 0 in case the id could not be retrieved. 4494 * 4495 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4496 * Description 4497 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4498 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4499 * of new data availability is sent. 4500 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4501 * of new data availability is sent unconditionally. 4502 * If **0** is specified in *flags*, an adaptive notification 4503 * of new data availability is sent. 4504 * 4505 * An adaptive notification is a notification sent whenever the user-space 4506 * process has caught up and consumed all available payloads. In case the user-space 4507 * process is still processing a previous payload, then no notification is needed 4508 * as it will process the newly added payload automatically. 4509 * Return 4510 * 0 on success, or a negative error in case of failure. 4511 * 4512 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4513 * Description 4514 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4515 * *flags* must be 0. 4516 * Return 4517 * Valid pointer with *size* bytes of memory available; NULL, 4518 * otherwise. 4519 * 4520 * void bpf_ringbuf_submit(void *data, u64 flags) 4521 * Description 4522 * Submit reserved ring buffer sample, pointed to by *data*. 4523 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4524 * of new data availability is sent. 4525 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4526 * of new data availability is sent unconditionally. 4527 * If **0** is specified in *flags*, an adaptive notification 4528 * of new data availability is sent. 4529 * 4530 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4531 * Return 4532 * Nothing. Always succeeds. 4533 * 4534 * void bpf_ringbuf_discard(void *data, u64 flags) 4535 * Description 4536 * Discard reserved ring buffer sample, pointed to by *data*. 4537 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4538 * of new data availability is sent. 4539 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4540 * of new data availability is sent unconditionally. 4541 * If **0** is specified in *flags*, an adaptive notification 4542 * of new data availability is sent. 4543 * 4544 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4545 * Return 4546 * Nothing. Always succeeds. 4547 * 4548 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4549 * Description 4550 * Query various characteristics of provided ring buffer. What 4551 * exactly is queries is determined by *flags*: 4552 * 4553 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4554 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4555 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4556 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4557 * 4558 * Data returned is just a momentary snapshot of actual values 4559 * and could be inaccurate, so this facility should be used to 4560 * power heuristics and for reporting, not to make 100% correct 4561 * calculation. 4562 * Return 4563 * Requested value, or 0, if *flags* are not recognized. 4564 * 4565 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4566 * Description 4567 * Change the skbs checksum level by one layer up or down, or 4568 * reset it entirely to none in order to have the stack perform 4569 * checksum validation. The level is applicable to the following 4570 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4571 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4572 * through **bpf_skb_adjust_room**\ () helper with passing in 4573 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4574 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4575 * the UDP header is removed. Similarly, an encap of the latter 4576 * into the former could be accompanied by a helper call to 4577 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4578 * skb is still intended to be processed in higher layers of the 4579 * stack instead of just egressing at tc. 4580 * 4581 * There are three supported level settings at this time: 4582 * 4583 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4584 * with CHECKSUM_UNNECESSARY. 4585 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4586 * with CHECKSUM_UNNECESSARY. 4587 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4588 * sets CHECKSUM_NONE to force checksum validation by the stack. 4589 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4590 * skb->csum_level. 4591 * Return 4592 * 0 on success, or a negative error in case of failure. In the 4593 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4594 * is returned or the error code -EACCES in case the skb is not 4595 * subject to CHECKSUM_UNNECESSARY. 4596 * 4597 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4598 * Description 4599 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4600 * Return 4601 * *sk* if casting is valid, or **NULL** otherwise. 4602 * 4603 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4604 * Description 4605 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4606 * Return 4607 * *sk* if casting is valid, or **NULL** otherwise. 4608 * 4609 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4610 * Description 4611 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4612 * Return 4613 * *sk* if casting is valid, or **NULL** otherwise. 4614 * 4615 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4616 * Description 4617 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4618 * Return 4619 * *sk* if casting is valid, or **NULL** otherwise. 4620 * 4621 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4622 * Description 4623 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4624 * Return 4625 * *sk* if casting is valid, or **NULL** otherwise. 4626 * 4627 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4628 * Description 4629 * Return a user or a kernel stack in bpf program provided buffer. 4630 * Note: the user stack will only be populated if the *task* is 4631 * the current task; all other tasks will return -EOPNOTSUPP. 4632 * To achieve this, the helper needs *task*, which is a valid 4633 * pointer to **struct task_struct**. To store the stacktrace, the 4634 * bpf program provides *buf* with a nonnegative *size*. 4635 * 4636 * The last argument, *flags*, holds the number of stack frames to 4637 * skip (from 0 to 255), masked with 4638 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4639 * the following flags: 4640 * 4641 * **BPF_F_USER_STACK** 4642 * Collect a user space stack instead of a kernel stack. 4643 * The *task* must be the current task. 4644 * **BPF_F_USER_BUILD_ID** 4645 * Collect buildid+offset instead of ips for user stack, 4646 * only valid if **BPF_F_USER_STACK** is also specified. 4647 * 4648 * **bpf_get_task_stack**\ () can collect up to 4649 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4650 * to sufficient large buffer size. Note that 4651 * this limit can be controlled with the **sysctl** program, and 4652 * that it should be manually increased in order to profile long 4653 * user stacks (such as stacks for Java programs). To do so, use: 4654 * 4655 * :: 4656 * 4657 * # sysctl kernel.perf_event_max_stack=<new value> 4658 * Return 4659 * The non-negative copied *buf* length equal to or less than 4660 * *size* on success, or a negative error in case of failure. 4661 * 4662 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4663 * Description 4664 * Load header option. Support reading a particular TCP header 4665 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4666 * 4667 * If *flags* is 0, it will search the option from the 4668 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4669 * has details on what skb_data contains under different 4670 * *skops*\ **->op**. 4671 * 4672 * The first byte of the *searchby_res* specifies the 4673 * kind that it wants to search. 4674 * 4675 * If the searching kind is an experimental kind 4676 * (i.e. 253 or 254 according to RFC6994). It also 4677 * needs to specify the "magic" which is either 4678 * 2 bytes or 4 bytes. It then also needs to 4679 * specify the size of the magic by using 4680 * the 2nd byte which is "kind-length" of a TCP 4681 * header option and the "kind-length" also 4682 * includes the first 2 bytes "kind" and "kind-length" 4683 * itself as a normal TCP header option also does. 4684 * 4685 * For example, to search experimental kind 254 with 4686 * 2 byte magic 0xeB9F, the searchby_res should be 4687 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4688 * 4689 * To search for the standard window scale option (3), 4690 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4691 * Note, kind-length must be 0 for regular option. 4692 * 4693 * Searching for No-Op (0) and End-of-Option-List (1) are 4694 * not supported. 4695 * 4696 * *len* must be at least 2 bytes which is the minimal size 4697 * of a header option. 4698 * 4699 * Supported flags: 4700 * 4701 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4702 * saved_syn packet or the just-received syn packet. 4703 * 4704 * Return 4705 * > 0 when found, the header option is copied to *searchby_res*. 4706 * The return value is the total length copied. On failure, a 4707 * negative error code is returned: 4708 * 4709 * **-EINVAL** if a parameter is invalid. 4710 * 4711 * **-ENOMSG** if the option is not found. 4712 * 4713 * **-ENOENT** if no syn packet is available when 4714 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4715 * 4716 * **-ENOSPC** if there is not enough space. Only *len* number of 4717 * bytes are copied. 4718 * 4719 * **-EFAULT** on failure to parse the header options in the 4720 * packet. 4721 * 4722 * **-EPERM** if the helper cannot be used under the current 4723 * *skops*\ **->op**. 4724 * 4725 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4726 * Description 4727 * Store header option. The data will be copied 4728 * from buffer *from* with length *len* to the TCP header. 4729 * 4730 * The buffer *from* should have the whole option that 4731 * includes the kind, kind-length, and the actual 4732 * option data. The *len* must be at least kind-length 4733 * long. The kind-length does not have to be 4 byte 4734 * aligned. The kernel will take care of the padding 4735 * and setting the 4 bytes aligned value to th->doff. 4736 * 4737 * This helper will check for duplicated option 4738 * by searching the same option in the outgoing skb. 4739 * 4740 * This helper can only be called during 4741 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4742 * 4743 * Return 4744 * 0 on success, or negative error in case of failure: 4745 * 4746 * **-EINVAL** If param is invalid. 4747 * 4748 * **-ENOSPC** if there is not enough space in the header. 4749 * Nothing has been written 4750 * 4751 * **-EEXIST** if the option already exists. 4752 * 4753 * **-EFAULT** on failure to parse the existing header options. 4754 * 4755 * **-EPERM** if the helper cannot be used under the current 4756 * *skops*\ **->op**. 4757 * 4758 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4759 * Description 4760 * Reserve *len* bytes for the bpf header option. The 4761 * space will be used by **bpf_store_hdr_opt**\ () later in 4762 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4763 * 4764 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4765 * the total number of bytes will be reserved. 4766 * 4767 * This helper can only be called during 4768 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4769 * 4770 * Return 4771 * 0 on success, or negative error in case of failure: 4772 * 4773 * **-EINVAL** if a parameter is invalid. 4774 * 4775 * **-ENOSPC** if there is not enough space in the header. 4776 * 4777 * **-EPERM** if the helper cannot be used under the current 4778 * *skops*\ **->op**. 4779 * 4780 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4781 * Description 4782 * Get a bpf_local_storage from an *inode*. 4783 * 4784 * Logically, it could be thought of as getting the value from 4785 * a *map* with *inode* as the **key**. From this 4786 * perspective, the usage is not much different from 4787 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4788 * helper enforces the key must be an inode and the map must also 4789 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4790 * 4791 * Underneath, the value is stored locally at *inode* instead of 4792 * the *map*. The *map* is used as the bpf-local-storage 4793 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4794 * searched against all bpf_local_storage residing at *inode*. 4795 * 4796 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4797 * used such that a new bpf_local_storage will be 4798 * created if one does not exist. *value* can be used 4799 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4800 * the initial value of a bpf_local_storage. If *value* is 4801 * **NULL**, the new bpf_local_storage will be zero initialized. 4802 * Return 4803 * A bpf_local_storage pointer is returned on success. 4804 * 4805 * **NULL** if not found or there was an error in adding 4806 * a new bpf_local_storage. 4807 * 4808 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4809 * Description 4810 * Delete a bpf_local_storage from an *inode*. 4811 * Return 4812 * 0 on success. 4813 * 4814 * **-ENOENT** if the bpf_local_storage cannot be found. 4815 * 4816 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4817 * Description 4818 * Return full path for given **struct path** object, which 4819 * needs to be the kernel BTF *path* object. The path is 4820 * returned in the provided buffer *buf* of size *sz* and 4821 * is zero terminated. 4822 * 4823 * Return 4824 * On success, the strictly positive length of the string, 4825 * including the trailing NUL character. On error, a negative 4826 * value. 4827 * 4828 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4829 * Description 4830 * Read *size* bytes from user space address *user_ptr* and store 4831 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4832 * Return 4833 * 0 on success, or a negative error in case of failure. 4834 * 4835 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4836 * Description 4837 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4838 * using *ptr*->type_id. This value should specify the type 4839 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4840 * can be used to look up vmlinux BTF type ids. Traversing the 4841 * data structure using BTF, the type information and values are 4842 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4843 * the pointer data is carried out to avoid kernel crashes during 4844 * operation. Smaller types can use string space on the stack; 4845 * larger programs can use map data to store the string 4846 * representation. 4847 * 4848 * The string can be subsequently shared with userspace via 4849 * bpf_perf_event_output() or ring buffer interfaces. 4850 * bpf_trace_printk() is to be avoided as it places too small 4851 * a limit on string size to be useful. 4852 * 4853 * *flags* is a combination of 4854 * 4855 * **BTF_F_COMPACT** 4856 * no formatting around type information 4857 * **BTF_F_NONAME** 4858 * no struct/union member names/types 4859 * **BTF_F_PTR_RAW** 4860 * show raw (unobfuscated) pointer values; 4861 * equivalent to printk specifier %px. 4862 * **BTF_F_ZERO** 4863 * show zero-valued struct/union members; they 4864 * are not displayed by default 4865 * 4866 * Return 4867 * The number of bytes that were written (or would have been 4868 * written if output had to be truncated due to string size), 4869 * or a negative error in cases of failure. 4870 * 4871 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4872 * Description 4873 * Use BTF to write to seq_write a string representation of 4874 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4875 * *flags* are identical to those used for bpf_snprintf_btf. 4876 * Return 4877 * 0 on success or a negative error in case of failure. 4878 * 4879 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4880 * Description 4881 * See **bpf_get_cgroup_classid**\ () for the main description. 4882 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4883 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4884 * associated socket instead of the current process. 4885 * Return 4886 * The id is returned or 0 in case the id could not be retrieved. 4887 * 4888 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4889 * Description 4890 * Redirect the packet to another net device of index *ifindex* 4891 * and fill in L2 addresses from neighboring subsystem. This helper 4892 * is somewhat similar to **bpf_redirect**\ (), except that it 4893 * populates L2 addresses as well, meaning, internally, the helper 4894 * relies on the neighbor lookup for the L2 address of the nexthop. 4895 * 4896 * The helper will perform a FIB lookup based on the skb's 4897 * networking header to get the address of the next hop, unless 4898 * this is supplied by the caller in the *params* argument. The 4899 * *plen* argument indicates the len of *params* and should be set 4900 * to 0 if *params* is NULL. 4901 * 4902 * The *flags* argument is reserved and must be 0. The helper is 4903 * currently only supported for tc BPF program types, and enabled 4904 * for IPv4 and IPv6 protocols. 4905 * Return 4906 * The helper returns **TC_ACT_REDIRECT** on success or 4907 * **TC_ACT_SHOT** on error. 4908 * 4909 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4910 * Description 4911 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4912 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4913 * extern variable decorated with '__ksym'. For ksym, there is a 4914 * global var (either static or global) defined of the same name 4915 * in the kernel. The ksym is percpu if the global var is percpu. 4916 * The returned pointer points to the global percpu var on *cpu*. 4917 * 4918 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4919 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4920 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4921 * bpf_per_cpu_ptr() must check the returned value. 4922 * Return 4923 * A pointer pointing to the kernel percpu variable on *cpu*, or 4924 * NULL, if *cpu* is invalid. 4925 * 4926 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4927 * Description 4928 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4929 * pointer to the percpu kernel variable on this cpu. See the 4930 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4931 * 4932 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4933 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4934 * never return NULL. 4935 * Return 4936 * A pointer pointing to the kernel percpu variable on this cpu. 4937 * 4938 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4939 * Description 4940 * Redirect the packet to another net device of index *ifindex*. 4941 * This helper is somewhat similar to **bpf_redirect**\ (), except 4942 * that the redirection happens to the *ifindex*' peer device and 4943 * the netns switch takes place from ingress to ingress without 4944 * going through the CPU's backlog queue. 4945 * 4946 * The *flags* argument is reserved and must be 0. The helper is 4947 * currently only supported for tc BPF program types at the 4948 * ingress hook and for veth and netkit target device types. The 4949 * peer device must reside in a different network namespace. 4950 * Return 4951 * The helper returns **TC_ACT_REDIRECT** on success or 4952 * **TC_ACT_SHOT** on error. 4953 * 4954 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4955 * Description 4956 * Get a bpf_local_storage from the *task*. 4957 * 4958 * Logically, it could be thought of as getting the value from 4959 * a *map* with *task* as the **key**. From this 4960 * perspective, the usage is not much different from 4961 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4962 * helper enforces the key must be a task_struct and the map must also 4963 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4964 * 4965 * Underneath, the value is stored locally at *task* instead of 4966 * the *map*. The *map* is used as the bpf-local-storage 4967 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4968 * searched against all bpf_local_storage residing at *task*. 4969 * 4970 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4971 * used such that a new bpf_local_storage will be 4972 * created if one does not exist. *value* can be used 4973 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4974 * the initial value of a bpf_local_storage. If *value* is 4975 * **NULL**, the new bpf_local_storage will be zero initialized. 4976 * Return 4977 * A bpf_local_storage pointer is returned on success. 4978 * 4979 * **NULL** if not found or there was an error in adding 4980 * a new bpf_local_storage. 4981 * 4982 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4983 * Description 4984 * Delete a bpf_local_storage from a *task*. 4985 * Return 4986 * 0 on success. 4987 * 4988 * **-ENOENT** if the bpf_local_storage cannot be found. 4989 * 4990 * struct task_struct *bpf_get_current_task_btf(void) 4991 * Description 4992 * Return a BTF pointer to the "current" task. 4993 * This pointer can also be used in helpers that accept an 4994 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4995 * Return 4996 * Pointer to the current task. 4997 * 4998 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4999 * Description 5000 * Set or clear certain options on *bprm*: 5001 * 5002 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 5003 * which sets the **AT_SECURE** auxv for glibc. The bit 5004 * is cleared if the flag is not specified. 5005 * Return 5006 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 5007 * 5008 * u64 bpf_ktime_get_coarse_ns(void) 5009 * Description 5010 * Return a coarse-grained version of the time elapsed since 5011 * system boot, in nanoseconds. Does not include time the system 5012 * was suspended. 5013 * 5014 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 5015 * Return 5016 * Current *ktime*. 5017 * 5018 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 5019 * Description 5020 * Returns the stored IMA hash of the *inode* (if it's available). 5021 * If the hash is larger than *size*, then only *size* 5022 * bytes will be copied to *dst* 5023 * Return 5024 * The **hash_algo** is returned on success, 5025 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 5026 * invalid arguments are passed. 5027 * 5028 * struct socket *bpf_sock_from_file(struct file *file) 5029 * Description 5030 * If the given file represents a socket, returns the associated 5031 * socket. 5032 * Return 5033 * A pointer to a struct socket on success or NULL if the file is 5034 * not a socket. 5035 * 5036 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 5037 * Description 5038 * Check packet size against exceeding MTU of net device (based 5039 * on *ifindex*). This helper will likely be used in combination 5040 * with helpers that adjust/change the packet size. 5041 * 5042 * The argument *len_diff* can be used for querying with a planned 5043 * size change. This allows to check MTU prior to changing packet 5044 * ctx. Providing a *len_diff* adjustment that is larger than the 5045 * actual packet size (resulting in negative packet size) will in 5046 * principle not exceed the MTU, which is why it is not considered 5047 * a failure. Other BPF helpers are needed for performing the 5048 * planned size change; therefore the responsibility for catching 5049 * a negative packet size belongs in those helpers. 5050 * 5051 * Specifying *ifindex* zero means the MTU check is performed 5052 * against the current net device. This is practical if this isn't 5053 * used prior to redirect. 5054 * 5055 * On input *mtu_len* must be a valid pointer, else verifier will 5056 * reject BPF program. If the value *mtu_len* is initialized to 5057 * zero then the ctx packet size is use. When value *mtu_len* is 5058 * provided as input this specify the L3 length that the MTU check 5059 * is done against. Remember XDP and TC length operate at L2, but 5060 * this value is L3 as this correlate to MTU and IP-header tot_len 5061 * values which are L3 (similar behavior as bpf_fib_lookup). 5062 * 5063 * The Linux kernel route table can configure MTUs on a more 5064 * specific per route level, which is not provided by this helper. 5065 * For route level MTU checks use the **bpf_fib_lookup**\ () 5066 * helper. 5067 * 5068 * *ctx* is either **struct xdp_md** for XDP programs or 5069 * **struct sk_buff** for tc cls_act programs. 5070 * 5071 * The *flags* argument can be a combination of one or more of the 5072 * following values: 5073 * 5074 * **BPF_MTU_CHK_SEGS** 5075 * This flag will only works for *ctx* **struct sk_buff**. 5076 * If packet context contains extra packet segment buffers 5077 * (often knows as GSO skb), then MTU check is harder to 5078 * check at this point, because in transmit path it is 5079 * possible for the skb packet to get re-segmented 5080 * (depending on net device features). This could still be 5081 * a MTU violation, so this flag enables performing MTU 5082 * check against segments, with a different violation 5083 * return code to tell it apart. Check cannot use len_diff. 5084 * 5085 * On return *mtu_len* pointer contains the MTU value of the net 5086 * device. Remember the net device configured MTU is the L3 size, 5087 * which is returned here and XDP and TC length operate at L2. 5088 * Helper take this into account for you, but remember when using 5089 * MTU value in your BPF-code. 5090 * 5091 * Return 5092 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5093 * 5094 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5095 * 5096 * MTU violations return positive values, but also populate MTU 5097 * value in *mtu_len* pointer, as this can be needed for 5098 * implementing PMTU handing: 5099 * 5100 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5101 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5102 * 5103 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5104 * Description 5105 * For each element in **map**, call **callback_fn** function with 5106 * **map**, **callback_ctx** and other map-specific parameters. 5107 * The **callback_fn** should be a static function and 5108 * the **callback_ctx** should be a pointer to the stack. 5109 * The **flags** is used to control certain aspects of the helper. 5110 * Currently, the **flags** must be 0. 5111 * 5112 * The following are a list of supported map types and their 5113 * respective expected callback signatures: 5114 * 5115 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5116 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5117 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5118 * 5119 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5120 * 5121 * For per_cpu maps, the map_value is the value on the cpu where the 5122 * bpf_prog is running. 5123 * 5124 * If **callback_fn** return 0, the helper will continue to the next 5125 * element. If return value is 1, the helper will skip the rest of 5126 * elements and return. Other return values are not used now. 5127 * 5128 * Return 5129 * The number of traversed map elements for success, **-EINVAL** for 5130 * invalid **flags**. 5131 * 5132 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5133 * Description 5134 * Outputs a string into the **str** buffer of size **str_size** 5135 * based on a format string stored in a read-only map pointed by 5136 * **fmt**. 5137 * 5138 * Each format specifier in **fmt** corresponds to one u64 element 5139 * in the **data** array. For strings and pointers where pointees 5140 * are accessed, only the pointer values are stored in the *data* 5141 * array. The *data_len* is the size of *data* in bytes - must be 5142 * a multiple of 8. 5143 * 5144 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5145 * memory. Reading kernel memory may fail due to either invalid 5146 * address or valid address but requiring a major memory fault. If 5147 * reading kernel memory fails, the string for **%s** will be an 5148 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5149 * Not returning error to bpf program is consistent with what 5150 * **bpf_trace_printk**\ () does for now. 5151 * 5152 * Return 5153 * The strictly positive length of the formatted string, including 5154 * the trailing zero character. If the return value is greater than 5155 * **str_size**, **str** contains a truncated string, guaranteed to 5156 * be zero-terminated except when **str_size** is 0. 5157 * 5158 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5159 * 5160 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5161 * Description 5162 * Execute bpf syscall with given arguments. 5163 * Return 5164 * A syscall result. 5165 * 5166 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5167 * Description 5168 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5169 * Return 5170 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5171 * 5172 * long bpf_sys_close(u32 fd) 5173 * Description 5174 * Execute close syscall for given FD. 5175 * Return 5176 * A syscall result. 5177 * 5178 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5179 * Description 5180 * Initialize the timer. 5181 * First 4 bits of *flags* specify clockid. 5182 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5183 * All other bits of *flags* are reserved. 5184 * The verifier will reject the program if *timer* is not from 5185 * the same *map*. 5186 * Return 5187 * 0 on success. 5188 * **-EBUSY** if *timer* is already initialized. 5189 * **-EINVAL** if invalid *flags* are passed. 5190 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5191 * The user space should either hold a file descriptor to a map with timers 5192 * or pin such map in bpffs. When map is unpinned or file descriptor is 5193 * closed all timers in the map will be cancelled and freed. 5194 * 5195 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5196 * Description 5197 * Configure the timer to call *callback_fn* static function. 5198 * Return 5199 * 0 on success. 5200 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5201 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5202 * The user space should either hold a file descriptor to a map with timers 5203 * or pin such map in bpffs. When map is unpinned or file descriptor is 5204 * closed all timers in the map will be cancelled and freed. 5205 * 5206 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5207 * Description 5208 * Set timer expiration N nanoseconds from the current time. The 5209 * configured callback will be invoked in soft irq context on some cpu 5210 * and will not repeat unless another bpf_timer_start() is made. 5211 * In such case the next invocation can migrate to a different cpu. 5212 * Since struct bpf_timer is a field inside map element the map 5213 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5214 * of BPF program to make sure that callback_fn code stays valid. 5215 * When user space reference to a map reaches zero all timers 5216 * in a map are cancelled and corresponding program's refcnts are 5217 * decremented. This is done to make sure that Ctrl-C of a user 5218 * process doesn't leave any timers running. If map is pinned in 5219 * bpffs the callback_fn can re-arm itself indefinitely. 5220 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5221 * cancel and free the timer in the given map element. 5222 * The map can contain timers that invoke callback_fn-s from different 5223 * programs. The same callback_fn can serve different timers from 5224 * different maps if key/value layout matches across maps. 5225 * Every bpf_timer_set_callback() can have different callback_fn. 5226 * 5227 * *flags* can be one of: 5228 * 5229 * **BPF_F_TIMER_ABS** 5230 * Start the timer in absolute expire value instead of the 5231 * default relative one. 5232 * **BPF_F_TIMER_CPU_PIN** 5233 * Timer will be pinned to the CPU of the caller. 5234 * 5235 * Return 5236 * 0 on success. 5237 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5238 * or invalid *flags* are passed. 5239 * 5240 * long bpf_timer_cancel(struct bpf_timer *timer) 5241 * Description 5242 * Cancel the timer and wait for callback_fn to finish if it was running. 5243 * Return 5244 * 0 if the timer was not active. 5245 * 1 if the timer was active. 5246 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5247 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5248 * own timer which would have led to a deadlock otherwise. 5249 * 5250 * u64 bpf_get_func_ip(void *ctx) 5251 * Description 5252 * Get address of the traced function (for tracing and kprobe programs). 5253 * 5254 * When called for kprobe program attached as uprobe it returns 5255 * probe address for both entry and return uprobe. 5256 * 5257 * Return 5258 * Address of the traced function for kprobe. 5259 * 0 for kprobes placed within the function (not at the entry). 5260 * Address of the probe for uprobe and return uprobe. 5261 * 5262 * u64 bpf_get_attach_cookie(void *ctx) 5263 * Description 5264 * Get bpf_cookie value provided (optionally) during the program 5265 * attachment. It might be different for each individual 5266 * attachment, even if BPF program itself is the same. 5267 * Expects BPF program context *ctx* as a first argument. 5268 * 5269 * Supported for the following program types: 5270 * - kprobe/uprobe; 5271 * - tracepoint; 5272 * - perf_event. 5273 * Return 5274 * Value specified by user at BPF link creation/attachment time 5275 * or 0, if it was not specified. 5276 * 5277 * long bpf_task_pt_regs(struct task_struct *task) 5278 * Description 5279 * Get the struct pt_regs associated with **task**. 5280 * Return 5281 * A pointer to struct pt_regs. 5282 * 5283 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5284 * Description 5285 * Get branch trace from hardware engines like Intel LBR. The 5286 * hardware engine is stopped shortly after the helper is 5287 * called. Therefore, the user need to filter branch entries 5288 * based on the actual use case. To capture branch trace 5289 * before the trigger point of the BPF program, the helper 5290 * should be called at the beginning of the BPF program. 5291 * 5292 * The data is stored as struct perf_branch_entry into output 5293 * buffer *entries*. *size* is the size of *entries* in bytes. 5294 * *flags* is reserved for now and must be zero. 5295 * 5296 * Return 5297 * On success, number of bytes written to *buf*. On error, a 5298 * negative value. 5299 * 5300 * **-EINVAL** if *flags* is not zero. 5301 * 5302 * **-ENOENT** if architecture does not support branch records. 5303 * 5304 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5305 * Description 5306 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5307 * to format and can handle more format args as a result. 5308 * 5309 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5310 * Return 5311 * The number of bytes written to the buffer, or a negative error 5312 * in case of failure. 5313 * 5314 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5315 * Description 5316 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5317 * Return 5318 * *sk* if casting is valid, or **NULL** otherwise. 5319 * 5320 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5321 * Description 5322 * Get the address of a kernel symbol, returned in *res*. *res* is 5323 * set to 0 if the symbol is not found. 5324 * Return 5325 * On success, zero. On error, a negative value. 5326 * 5327 * **-EINVAL** if *flags* is not zero. 5328 * 5329 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5330 * 5331 * **-ENOENT** if symbol is not found. 5332 * 5333 * **-EPERM** if caller does not have permission to obtain kernel address. 5334 * 5335 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5336 * Description 5337 * Find vma of *task* that contains *addr*, call *callback_fn* 5338 * function with *task*, *vma*, and *callback_ctx*. 5339 * The *callback_fn* should be a static function and 5340 * the *callback_ctx* should be a pointer to the stack. 5341 * The *flags* is used to control certain aspects of the helper. 5342 * Currently, the *flags* must be 0. 5343 * 5344 * The expected callback signature is 5345 * 5346 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5347 * 5348 * Return 5349 * 0 on success. 5350 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5351 * **-EBUSY** if failed to try lock mmap_lock. 5352 * **-EINVAL** for invalid **flags**. 5353 * 5354 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5355 * Description 5356 * For **nr_loops**, call **callback_fn** function 5357 * with **callback_ctx** as the context parameter. 5358 * The **callback_fn** should be a static function and 5359 * the **callback_ctx** should be a pointer to the stack. 5360 * The **flags** is used to control certain aspects of the helper. 5361 * Currently, the **flags** must be 0. Currently, nr_loops is 5362 * limited to 1 << 23 (~8 million) loops. 5363 * 5364 * long (\*callback_fn)(u32 index, void \*ctx); 5365 * 5366 * where **index** is the current index in the loop. The index 5367 * is zero-indexed. 5368 * 5369 * If **callback_fn** returns 0, the helper will continue to the next 5370 * loop. If return value is 1, the helper will skip the rest of 5371 * the loops and return. Other return values are not used now, 5372 * and will be rejected by the verifier. 5373 * 5374 * Return 5375 * The number of loops performed, **-EINVAL** for invalid **flags**, 5376 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5377 * 5378 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5379 * Description 5380 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5381 * to be null-terminated and **s1_sz** is the maximum storage 5382 * size of **s1**. **s2** must be a read-only string. 5383 * Return 5384 * An integer less than, equal to, or greater than zero 5385 * if the first **s1_sz** bytes of **s1** is found to be 5386 * less than, to match, or be greater than **s2**. 5387 * 5388 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5389 * Description 5390 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5391 * returned in **value**. 5392 * 5393 * Return 5394 * 0 on success. 5395 * **-EINVAL** if n >= argument register count of traced function. 5396 * 5397 * long bpf_get_func_ret(void *ctx, u64 *value) 5398 * Description 5399 * Get return value of the traced function (for tracing programs) 5400 * in **value**. 5401 * 5402 * Return 5403 * 0 on success. 5404 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5405 * 5406 * long bpf_get_func_arg_cnt(void *ctx) 5407 * Description 5408 * Get number of registers of the traced function (for tracing programs) where 5409 * function arguments are stored in these registers. 5410 * 5411 * Return 5412 * The number of argument registers of the traced function. 5413 * 5414 * int bpf_get_retval(void) 5415 * Description 5416 * Get the BPF program's return value that will be returned to the upper layers. 5417 * 5418 * This helper is currently supported by cgroup programs and only by the hooks 5419 * where BPF program's return value is returned to the userspace via errno. 5420 * Return 5421 * The BPF program's return value. 5422 * 5423 * int bpf_set_retval(int retval) 5424 * Description 5425 * Set the BPF program's return value that will be returned to the upper layers. 5426 * 5427 * This helper is currently supported by cgroup programs and only by the hooks 5428 * where BPF program's return value is returned to the userspace via errno. 5429 * 5430 * Note that there is the following corner case where the program exports an error 5431 * via bpf_set_retval but signals success via 'return 1': 5432 * 5433 * bpf_set_retval(-EPERM); 5434 * return 1; 5435 * 5436 * In this case, the BPF program's return value will use helper's -EPERM. This 5437 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5438 * 5439 * Return 5440 * 0 on success, or a negative error in case of failure. 5441 * 5442 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5443 * Description 5444 * Get the total size of a given xdp buff (linear and paged area) 5445 * Return 5446 * The total size of a given xdp buffer. 5447 * 5448 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5449 * Description 5450 * This helper is provided as an easy way to load data from a 5451 * xdp buffer. It can be used to load *len* bytes from *offset* from 5452 * the frame associated to *xdp_md*, into the buffer pointed by 5453 * *buf*. 5454 * Return 5455 * 0 on success, or a negative error in case of failure. 5456 * 5457 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5458 * Description 5459 * Store *len* bytes from buffer *buf* into the frame 5460 * associated to *xdp_md*, at *offset*. 5461 * Return 5462 * 0 on success, or a negative error in case of failure. 5463 * 5464 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5465 * Description 5466 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5467 * address space, and stores the data in *dst*. *flags* is not 5468 * used yet and is provided for future extensibility. This helper 5469 * can only be used by sleepable programs. 5470 * Return 5471 * 0 on success, or a negative error in case of failure. On error 5472 * *dst* buffer is zeroed out. 5473 * 5474 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5475 * Description 5476 * Change the __sk_buff->tstamp_type to *tstamp_type* 5477 * and set *tstamp* to the __sk_buff->tstamp together. 5478 * 5479 * If there is no need to change the __sk_buff->tstamp_type, 5480 * the tstamp value can be directly written to __sk_buff->tstamp 5481 * instead. 5482 * 5483 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5484 * will be kept during bpf_redirect_*(). A non zero 5485 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5486 * *tstamp_type*. 5487 * 5488 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5489 * with a zero *tstamp*. 5490 * 5491 * Only IPv4 and IPv6 skb->protocol are supported. 5492 * 5493 * This function is most useful when it needs to set a 5494 * mono delivery time to __sk_buff->tstamp and then 5495 * bpf_redirect_*() to the egress of an iface. For example, 5496 * changing the (rcv) timestamp in __sk_buff->tstamp at 5497 * ingress to a mono delivery time and then bpf_redirect_*() 5498 * to sch_fq@phy-dev. 5499 * Return 5500 * 0 on success. 5501 * **-EINVAL** for invalid input 5502 * **-EOPNOTSUPP** for unsupported protocol 5503 * 5504 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5505 * Description 5506 * Returns a calculated IMA hash of the *file*. 5507 * If the hash is larger than *size*, then only *size* 5508 * bytes will be copied to *dst* 5509 * Return 5510 * The **hash_algo** is returned on success, 5511 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5512 * invalid arguments are passed. 5513 * 5514 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5515 * Description 5516 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5517 * old value. *ptr* can be NULL, otherwise it must be a referenced 5518 * pointer which will be released when this helper is called. 5519 * Return 5520 * The old value of kptr (which can be NULL). The returned pointer 5521 * if not NULL, is a reference which must be released using its 5522 * corresponding release function, or moved into a BPF map before 5523 * program exit. 5524 * 5525 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5526 * Description 5527 * Perform a lookup in *percpu map* for an entry associated to 5528 * *key* on *cpu*. 5529 * Return 5530 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5531 * was found or *cpu* is invalid. 5532 * 5533 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5534 * Description 5535 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5536 * Return 5537 * *sk* if casting is valid, or **NULL** otherwise. 5538 * 5539 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5540 * Description 5541 * Get a dynptr to local memory *data*. 5542 * 5543 * *data* must be a ptr to a map value. 5544 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5545 * *flags* is currently unused. 5546 * Return 5547 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5548 * -EINVAL if flags is not 0. 5549 * 5550 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5551 * Description 5552 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5553 * through the dynptr interface. *flags* must be 0. 5554 * 5555 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5556 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5557 * reservation fails. This is enforced by the verifier. 5558 * Return 5559 * 0 on success, or a negative error in case of failure. 5560 * 5561 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5562 * Description 5563 * Submit reserved ring buffer sample, pointed to by *data*, 5564 * through the dynptr interface. This is a no-op if the dynptr is 5565 * invalid/null. 5566 * 5567 * For more information on *flags*, please see 5568 * 'bpf_ringbuf_submit'. 5569 * Return 5570 * Nothing. Always succeeds. 5571 * 5572 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5573 * Description 5574 * Discard reserved ring buffer sample through the dynptr 5575 * interface. This is a no-op if the dynptr is invalid/null. 5576 * 5577 * For more information on *flags*, please see 5578 * 'bpf_ringbuf_discard'. 5579 * Return 5580 * Nothing. Always succeeds. 5581 * 5582 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5583 * Description 5584 * Read *len* bytes from *src* into *dst*, starting from *offset* 5585 * into *src*. 5586 * *flags* is currently unused. 5587 * Return 5588 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5589 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5590 * *flags* is not 0. 5591 * 5592 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5593 * Description 5594 * Write *len* bytes from *src* into *dst*, starting from *offset* 5595 * into *dst*. 5596 * 5597 * *flags* must be 0 except for skb-type dynptrs. 5598 * 5599 * For skb-type dynptrs: 5600 * * All data slices of the dynptr are automatically 5601 * invalidated after **bpf_dynptr_write**\ (). This is 5602 * because writing may pull the skb and change the 5603 * underlying packet buffer. 5604 * 5605 * * For *flags*, please see the flags accepted by 5606 * **bpf_skb_store_bytes**\ (). 5607 * Return 5608 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5609 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5610 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5611 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5612 * 5613 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5614 * Description 5615 * Get a pointer to the underlying dynptr data. 5616 * 5617 * *len* must be a statically known value. The returned data slice 5618 * is invalidated whenever the dynptr is invalidated. 5619 * 5620 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5621 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5622 * Return 5623 * Pointer to the underlying dynptr data, NULL if the dynptr is 5624 * read-only, if the dynptr is invalid, or if the offset and length 5625 * is out of bounds. 5626 * 5627 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5628 * Description 5629 * Try to issue a SYN cookie for the packet with corresponding 5630 * IPv4/TCP headers, *iph* and *th*, without depending on a 5631 * listening socket. 5632 * 5633 * *iph* points to the IPv4 header. 5634 * 5635 * *th* points to the start of the TCP header, while *th_len* 5636 * contains the length of the TCP header (at least 5637 * **sizeof**\ (**struct tcphdr**)). 5638 * Return 5639 * On success, lower 32 bits hold the generated SYN cookie in 5640 * followed by 16 bits which hold the MSS value for that cookie, 5641 * and the top 16 bits are unused. 5642 * 5643 * On failure, the returned value is one of the following: 5644 * 5645 * **-EINVAL** if *th_len* is invalid. 5646 * 5647 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5648 * Description 5649 * Try to issue a SYN cookie for the packet with corresponding 5650 * IPv6/TCP headers, *iph* and *th*, without depending on a 5651 * listening socket. 5652 * 5653 * *iph* points to the IPv6 header. 5654 * 5655 * *th* points to the start of the TCP header, while *th_len* 5656 * contains the length of the TCP header (at least 5657 * **sizeof**\ (**struct tcphdr**)). 5658 * Return 5659 * On success, lower 32 bits hold the generated SYN cookie in 5660 * followed by 16 bits which hold the MSS value for that cookie, 5661 * and the top 16 bits are unused. 5662 * 5663 * On failure, the returned value is one of the following: 5664 * 5665 * **-EINVAL** if *th_len* is invalid. 5666 * 5667 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5668 * 5669 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5670 * Description 5671 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5672 * without depending on a listening socket. 5673 * 5674 * *iph* points to the IPv4 header. 5675 * 5676 * *th* points to the TCP header. 5677 * Return 5678 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5679 * 5680 * On failure, the returned value is one of the following: 5681 * 5682 * **-EACCES** if the SYN cookie is not valid. 5683 * 5684 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5685 * Description 5686 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5687 * without depending on a listening socket. 5688 * 5689 * *iph* points to the IPv6 header. 5690 * 5691 * *th* points to the TCP header. 5692 * Return 5693 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5694 * 5695 * On failure, the returned value is one of the following: 5696 * 5697 * **-EACCES** if the SYN cookie is not valid. 5698 * 5699 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5700 * 5701 * u64 bpf_ktime_get_tai_ns(void) 5702 * Description 5703 * A nonsettable system-wide clock derived from wall-clock time but 5704 * ignoring leap seconds. This clock does not experience 5705 * discontinuities and backwards jumps caused by NTP inserting leap 5706 * seconds as CLOCK_REALTIME does. 5707 * 5708 * See: **clock_gettime**\ (**CLOCK_TAI**) 5709 * Return 5710 * Current *ktime*. 5711 * 5712 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5713 * Description 5714 * Drain samples from the specified user ring buffer, and invoke 5715 * the provided callback for each such sample: 5716 * 5717 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5718 * 5719 * If **callback_fn** returns 0, the helper will continue to try 5720 * and drain the next sample, up to a maximum of 5721 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5722 * the helper will skip the rest of the samples and return. Other 5723 * return values are not used now, and will be rejected by the 5724 * verifier. 5725 * Return 5726 * The number of drained samples if no error was encountered while 5727 * draining samples, or 0 if no samples were present in the ring 5728 * buffer. If a user-space producer was epoll-waiting on this map, 5729 * and at least one sample was drained, they will receive an event 5730 * notification notifying them of available space in the ring 5731 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5732 * function, no wakeup notification will be sent. If the 5733 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5734 * be sent even if no sample was drained. 5735 * 5736 * On failure, the returned value is one of the following: 5737 * 5738 * **-EBUSY** if the ring buffer is contended, and another calling 5739 * context was concurrently draining the ring buffer. 5740 * 5741 * **-EINVAL** if user-space is not properly tracking the ring 5742 * buffer due to the producer position not being aligned to 8 5743 * bytes, a sample not being aligned to 8 bytes, or the producer 5744 * position not matching the advertised length of a sample. 5745 * 5746 * **-E2BIG** if user-space has tried to publish a sample which is 5747 * larger than the size of the ring buffer, or which cannot fit 5748 * within a struct bpf_dynptr. 5749 * 5750 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5751 * Description 5752 * Get a bpf_local_storage from the *cgroup*. 5753 * 5754 * Logically, it could be thought of as getting the value from 5755 * a *map* with *cgroup* as the **key**. From this 5756 * perspective, the usage is not much different from 5757 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5758 * helper enforces the key must be a cgroup struct and the map must also 5759 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5760 * 5761 * In reality, the local-storage value is embedded directly inside of the 5762 * *cgroup* object itself, rather than being located in the 5763 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5764 * queried for some *map* on a *cgroup* object, the kernel will perform an 5765 * O(n) iteration over all of the live local-storage values for that 5766 * *cgroup* object until the local-storage value for the *map* is found. 5767 * 5768 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5769 * used such that a new bpf_local_storage will be 5770 * created if one does not exist. *value* can be used 5771 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5772 * the initial value of a bpf_local_storage. If *value* is 5773 * **NULL**, the new bpf_local_storage will be zero initialized. 5774 * Return 5775 * A bpf_local_storage pointer is returned on success. 5776 * 5777 * **NULL** if not found or there was an error in adding 5778 * a new bpf_local_storage. 5779 * 5780 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5781 * Description 5782 * Delete a bpf_local_storage from a *cgroup*. 5783 * Return 5784 * 0 on success. 5785 * 5786 * **-ENOENT** if the bpf_local_storage cannot be found. 5787 */ 5788 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5789 FN(unspec, 0, ##ctx) \ 5790 FN(map_lookup_elem, 1, ##ctx) \ 5791 FN(map_update_elem, 2, ##ctx) \ 5792 FN(map_delete_elem, 3, ##ctx) \ 5793 FN(probe_read, 4, ##ctx) \ 5794 FN(ktime_get_ns, 5, ##ctx) \ 5795 FN(trace_printk, 6, ##ctx) \ 5796 FN(get_prandom_u32, 7, ##ctx) \ 5797 FN(get_smp_processor_id, 8, ##ctx) \ 5798 FN(skb_store_bytes, 9, ##ctx) \ 5799 FN(l3_csum_replace, 10, ##ctx) \ 5800 FN(l4_csum_replace, 11, ##ctx) \ 5801 FN(tail_call, 12, ##ctx) \ 5802 FN(clone_redirect, 13, ##ctx) \ 5803 FN(get_current_pid_tgid, 14, ##ctx) \ 5804 FN(get_current_uid_gid, 15, ##ctx) \ 5805 FN(get_current_comm, 16, ##ctx) \ 5806 FN(get_cgroup_classid, 17, ##ctx) \ 5807 FN(skb_vlan_push, 18, ##ctx) \ 5808 FN(skb_vlan_pop, 19, ##ctx) \ 5809 FN(skb_get_tunnel_key, 20, ##ctx) \ 5810 FN(skb_set_tunnel_key, 21, ##ctx) \ 5811 FN(perf_event_read, 22, ##ctx) \ 5812 FN(redirect, 23, ##ctx) \ 5813 FN(get_route_realm, 24, ##ctx) \ 5814 FN(perf_event_output, 25, ##ctx) \ 5815 FN(skb_load_bytes, 26, ##ctx) \ 5816 FN(get_stackid, 27, ##ctx) \ 5817 FN(csum_diff, 28, ##ctx) \ 5818 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5819 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5820 FN(skb_change_proto, 31, ##ctx) \ 5821 FN(skb_change_type, 32, ##ctx) \ 5822 FN(skb_under_cgroup, 33, ##ctx) \ 5823 FN(get_hash_recalc, 34, ##ctx) \ 5824 FN(get_current_task, 35, ##ctx) \ 5825 FN(probe_write_user, 36, ##ctx) \ 5826 FN(current_task_under_cgroup, 37, ##ctx) \ 5827 FN(skb_change_tail, 38, ##ctx) \ 5828 FN(skb_pull_data, 39, ##ctx) \ 5829 FN(csum_update, 40, ##ctx) \ 5830 FN(set_hash_invalid, 41, ##ctx) \ 5831 FN(get_numa_node_id, 42, ##ctx) \ 5832 FN(skb_change_head, 43, ##ctx) \ 5833 FN(xdp_adjust_head, 44, ##ctx) \ 5834 FN(probe_read_str, 45, ##ctx) \ 5835 FN(get_socket_cookie, 46, ##ctx) \ 5836 FN(get_socket_uid, 47, ##ctx) \ 5837 FN(set_hash, 48, ##ctx) \ 5838 FN(setsockopt, 49, ##ctx) \ 5839 FN(skb_adjust_room, 50, ##ctx) \ 5840 FN(redirect_map, 51, ##ctx) \ 5841 FN(sk_redirect_map, 52, ##ctx) \ 5842 FN(sock_map_update, 53, ##ctx) \ 5843 FN(xdp_adjust_meta, 54, ##ctx) \ 5844 FN(perf_event_read_value, 55, ##ctx) \ 5845 FN(perf_prog_read_value, 56, ##ctx) \ 5846 FN(getsockopt, 57, ##ctx) \ 5847 FN(override_return, 58, ##ctx) \ 5848 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5849 FN(msg_redirect_map, 60, ##ctx) \ 5850 FN(msg_apply_bytes, 61, ##ctx) \ 5851 FN(msg_cork_bytes, 62, ##ctx) \ 5852 FN(msg_pull_data, 63, ##ctx) \ 5853 FN(bind, 64, ##ctx) \ 5854 FN(xdp_adjust_tail, 65, ##ctx) \ 5855 FN(skb_get_xfrm_state, 66, ##ctx) \ 5856 FN(get_stack, 67, ##ctx) \ 5857 FN(skb_load_bytes_relative, 68, ##ctx) \ 5858 FN(fib_lookup, 69, ##ctx) \ 5859 FN(sock_hash_update, 70, ##ctx) \ 5860 FN(msg_redirect_hash, 71, ##ctx) \ 5861 FN(sk_redirect_hash, 72, ##ctx) \ 5862 FN(lwt_push_encap, 73, ##ctx) \ 5863 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5864 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5865 FN(lwt_seg6_action, 76, ##ctx) \ 5866 FN(rc_repeat, 77, ##ctx) \ 5867 FN(rc_keydown, 78, ##ctx) \ 5868 FN(skb_cgroup_id, 79, ##ctx) \ 5869 FN(get_current_cgroup_id, 80, ##ctx) \ 5870 FN(get_local_storage, 81, ##ctx) \ 5871 FN(sk_select_reuseport, 82, ##ctx) \ 5872 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5873 FN(sk_lookup_tcp, 84, ##ctx) \ 5874 FN(sk_lookup_udp, 85, ##ctx) \ 5875 FN(sk_release, 86, ##ctx) \ 5876 FN(map_push_elem, 87, ##ctx) \ 5877 FN(map_pop_elem, 88, ##ctx) \ 5878 FN(map_peek_elem, 89, ##ctx) \ 5879 FN(msg_push_data, 90, ##ctx) \ 5880 FN(msg_pop_data, 91, ##ctx) \ 5881 FN(rc_pointer_rel, 92, ##ctx) \ 5882 FN(spin_lock, 93, ##ctx) \ 5883 FN(spin_unlock, 94, ##ctx) \ 5884 FN(sk_fullsock, 95, ##ctx) \ 5885 FN(tcp_sock, 96, ##ctx) \ 5886 FN(skb_ecn_set_ce, 97, ##ctx) \ 5887 FN(get_listener_sock, 98, ##ctx) \ 5888 FN(skc_lookup_tcp, 99, ##ctx) \ 5889 FN(tcp_check_syncookie, 100, ##ctx) \ 5890 FN(sysctl_get_name, 101, ##ctx) \ 5891 FN(sysctl_get_current_value, 102, ##ctx) \ 5892 FN(sysctl_get_new_value, 103, ##ctx) \ 5893 FN(sysctl_set_new_value, 104, ##ctx) \ 5894 FN(strtol, 105, ##ctx) \ 5895 FN(strtoul, 106, ##ctx) \ 5896 FN(sk_storage_get, 107, ##ctx) \ 5897 FN(sk_storage_delete, 108, ##ctx) \ 5898 FN(send_signal, 109, ##ctx) \ 5899 FN(tcp_gen_syncookie, 110, ##ctx) \ 5900 FN(skb_output, 111, ##ctx) \ 5901 FN(probe_read_user, 112, ##ctx) \ 5902 FN(probe_read_kernel, 113, ##ctx) \ 5903 FN(probe_read_user_str, 114, ##ctx) \ 5904 FN(probe_read_kernel_str, 115, ##ctx) \ 5905 FN(tcp_send_ack, 116, ##ctx) \ 5906 FN(send_signal_thread, 117, ##ctx) \ 5907 FN(jiffies64, 118, ##ctx) \ 5908 FN(read_branch_records, 119, ##ctx) \ 5909 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5910 FN(xdp_output, 121, ##ctx) \ 5911 FN(get_netns_cookie, 122, ##ctx) \ 5912 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5913 FN(sk_assign, 124, ##ctx) \ 5914 FN(ktime_get_boot_ns, 125, ##ctx) \ 5915 FN(seq_printf, 126, ##ctx) \ 5916 FN(seq_write, 127, ##ctx) \ 5917 FN(sk_cgroup_id, 128, ##ctx) \ 5918 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5919 FN(ringbuf_output, 130, ##ctx) \ 5920 FN(ringbuf_reserve, 131, ##ctx) \ 5921 FN(ringbuf_submit, 132, ##ctx) \ 5922 FN(ringbuf_discard, 133, ##ctx) \ 5923 FN(ringbuf_query, 134, ##ctx) \ 5924 FN(csum_level, 135, ##ctx) \ 5925 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5926 FN(skc_to_tcp_sock, 137, ##ctx) \ 5927 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5928 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5929 FN(skc_to_udp6_sock, 140, ##ctx) \ 5930 FN(get_task_stack, 141, ##ctx) \ 5931 FN(load_hdr_opt, 142, ##ctx) \ 5932 FN(store_hdr_opt, 143, ##ctx) \ 5933 FN(reserve_hdr_opt, 144, ##ctx) \ 5934 FN(inode_storage_get, 145, ##ctx) \ 5935 FN(inode_storage_delete, 146, ##ctx) \ 5936 FN(d_path, 147, ##ctx) \ 5937 FN(copy_from_user, 148, ##ctx) \ 5938 FN(snprintf_btf, 149, ##ctx) \ 5939 FN(seq_printf_btf, 150, ##ctx) \ 5940 FN(skb_cgroup_classid, 151, ##ctx) \ 5941 FN(redirect_neigh, 152, ##ctx) \ 5942 FN(per_cpu_ptr, 153, ##ctx) \ 5943 FN(this_cpu_ptr, 154, ##ctx) \ 5944 FN(redirect_peer, 155, ##ctx) \ 5945 FN(task_storage_get, 156, ##ctx) \ 5946 FN(task_storage_delete, 157, ##ctx) \ 5947 FN(get_current_task_btf, 158, ##ctx) \ 5948 FN(bprm_opts_set, 159, ##ctx) \ 5949 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5950 FN(ima_inode_hash, 161, ##ctx) \ 5951 FN(sock_from_file, 162, ##ctx) \ 5952 FN(check_mtu, 163, ##ctx) \ 5953 FN(for_each_map_elem, 164, ##ctx) \ 5954 FN(snprintf, 165, ##ctx) \ 5955 FN(sys_bpf, 166, ##ctx) \ 5956 FN(btf_find_by_name_kind, 167, ##ctx) \ 5957 FN(sys_close, 168, ##ctx) \ 5958 FN(timer_init, 169, ##ctx) \ 5959 FN(timer_set_callback, 170, ##ctx) \ 5960 FN(timer_start, 171, ##ctx) \ 5961 FN(timer_cancel, 172, ##ctx) \ 5962 FN(get_func_ip, 173, ##ctx) \ 5963 FN(get_attach_cookie, 174, ##ctx) \ 5964 FN(task_pt_regs, 175, ##ctx) \ 5965 FN(get_branch_snapshot, 176, ##ctx) \ 5966 FN(trace_vprintk, 177, ##ctx) \ 5967 FN(skc_to_unix_sock, 178, ##ctx) \ 5968 FN(kallsyms_lookup_name, 179, ##ctx) \ 5969 FN(find_vma, 180, ##ctx) \ 5970 FN(loop, 181, ##ctx) \ 5971 FN(strncmp, 182, ##ctx) \ 5972 FN(get_func_arg, 183, ##ctx) \ 5973 FN(get_func_ret, 184, ##ctx) \ 5974 FN(get_func_arg_cnt, 185, ##ctx) \ 5975 FN(get_retval, 186, ##ctx) \ 5976 FN(set_retval, 187, ##ctx) \ 5977 FN(xdp_get_buff_len, 188, ##ctx) \ 5978 FN(xdp_load_bytes, 189, ##ctx) \ 5979 FN(xdp_store_bytes, 190, ##ctx) \ 5980 FN(copy_from_user_task, 191, ##ctx) \ 5981 FN(skb_set_tstamp, 192, ##ctx) \ 5982 FN(ima_file_hash, 193, ##ctx) \ 5983 FN(kptr_xchg, 194, ##ctx) \ 5984 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5985 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5986 FN(dynptr_from_mem, 197, ##ctx) \ 5987 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5988 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5989 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5990 FN(dynptr_read, 201, ##ctx) \ 5991 FN(dynptr_write, 202, ##ctx) \ 5992 FN(dynptr_data, 203, ##ctx) \ 5993 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5994 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5995 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5996 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5997 FN(ktime_get_tai_ns, 208, ##ctx) \ 5998 FN(user_ringbuf_drain, 209, ##ctx) \ 5999 FN(cgrp_storage_get, 210, ##ctx) \ 6000 FN(cgrp_storage_delete, 211, ##ctx) \ 6001 /* */ 6002 6003 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 6004 * know or care about integer value that is now passed as second argument 6005 */ 6006 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 6007 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 6008 6009 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 6010 * function eBPF program intends to call 6011 */ 6012 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 6013 enum bpf_func_id { 6014 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 6015 __BPF_FUNC_MAX_ID, 6016 }; 6017 #undef __BPF_ENUM_FN 6018 6019 /* All flags used by eBPF helper functions, placed here. */ 6020 6021 /* BPF_FUNC_skb_store_bytes flags. */ 6022 enum { 6023 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 6024 BPF_F_INVALIDATE_HASH = (1ULL << 1), 6025 }; 6026 6027 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 6028 * First 4 bits are for passing the header field size. 6029 */ 6030 enum { 6031 BPF_F_HDR_FIELD_MASK = 0xfULL, 6032 }; 6033 6034 /* BPF_FUNC_l4_csum_replace flags. */ 6035 enum { 6036 BPF_F_PSEUDO_HDR = (1ULL << 4), 6037 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 6038 BPF_F_MARK_ENFORCE = (1ULL << 6), 6039 }; 6040 6041 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 6042 enum { 6043 BPF_F_INGRESS = (1ULL << 0), 6044 }; 6045 6046 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6047 enum { 6048 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6049 }; 6050 6051 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6052 enum { 6053 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6054 BPF_F_USER_STACK = (1ULL << 8), 6055 /* flags used by BPF_FUNC_get_stackid only. */ 6056 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6057 BPF_F_REUSE_STACKID = (1ULL << 10), 6058 /* flags used by BPF_FUNC_get_stack only. */ 6059 BPF_F_USER_BUILD_ID = (1ULL << 11), 6060 }; 6061 6062 /* BPF_FUNC_skb_set_tunnel_key flags. */ 6063 enum { 6064 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6065 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6066 BPF_F_SEQ_NUMBER = (1ULL << 3), 6067 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6068 }; 6069 6070 /* BPF_FUNC_skb_get_tunnel_key flags. */ 6071 enum { 6072 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6073 }; 6074 6075 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6076 * BPF_FUNC_perf_event_read_value flags. 6077 */ 6078 enum { 6079 BPF_F_INDEX_MASK = 0xffffffffULL, 6080 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6081 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 6082 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6083 }; 6084 6085 /* Current network namespace */ 6086 enum { 6087 BPF_F_CURRENT_NETNS = (-1L), 6088 }; 6089 6090 /* BPF_FUNC_csum_level level values. */ 6091 enum { 6092 BPF_CSUM_LEVEL_QUERY, 6093 BPF_CSUM_LEVEL_INC, 6094 BPF_CSUM_LEVEL_DEC, 6095 BPF_CSUM_LEVEL_RESET, 6096 }; 6097 6098 /* BPF_FUNC_skb_adjust_room flags. */ 6099 enum { 6100 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6101 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6102 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6103 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6104 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6105 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6106 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6107 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6108 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6109 }; 6110 6111 enum { 6112 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6113 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6114 }; 6115 6116 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6117 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6118 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6119 6120 /* BPF_FUNC_sysctl_get_name flags. */ 6121 enum { 6122 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6123 }; 6124 6125 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 6126 enum { 6127 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6128 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6129 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6130 */ 6131 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6132 }; 6133 6134 /* BPF_FUNC_read_branch_records flags. */ 6135 enum { 6136 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6137 }; 6138 6139 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6140 * BPF_FUNC_bpf_ringbuf_output flags. 6141 */ 6142 enum { 6143 BPF_RB_NO_WAKEUP = (1ULL << 0), 6144 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6145 }; 6146 6147 /* BPF_FUNC_bpf_ringbuf_query flags */ 6148 enum { 6149 BPF_RB_AVAIL_DATA = 0, 6150 BPF_RB_RING_SIZE = 1, 6151 BPF_RB_CONS_POS = 2, 6152 BPF_RB_PROD_POS = 3, 6153 }; 6154 6155 /* BPF ring buffer constants */ 6156 enum { 6157 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6158 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6159 BPF_RINGBUF_HDR_SZ = 8, 6160 }; 6161 6162 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6163 enum { 6164 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6165 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6166 }; 6167 6168 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6169 enum bpf_adj_room_mode { 6170 BPF_ADJ_ROOM_NET, 6171 BPF_ADJ_ROOM_MAC, 6172 }; 6173 6174 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6175 enum bpf_hdr_start_off { 6176 BPF_HDR_START_MAC, 6177 BPF_HDR_START_NET, 6178 }; 6179 6180 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6181 enum bpf_lwt_encap_mode { 6182 BPF_LWT_ENCAP_SEG6, 6183 BPF_LWT_ENCAP_SEG6_INLINE, 6184 BPF_LWT_ENCAP_IP, 6185 }; 6186 6187 /* Flags for bpf_bprm_opts_set helper */ 6188 enum { 6189 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6190 }; 6191 6192 /* Flags for bpf_redirect_map helper */ 6193 enum { 6194 BPF_F_BROADCAST = (1ULL << 3), 6195 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 6196 }; 6197 6198 #define __bpf_md_ptr(type, name) \ 6199 union { \ 6200 type name; \ 6201 __u64 :64; \ 6202 } __attribute__((aligned(8))) 6203 6204 enum { 6205 BPF_SKB_TSTAMP_UNSPEC, 6206 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 6207 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 6208 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 6209 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 6210 */ 6211 }; 6212 6213 /* user accessible mirror of in-kernel sk_buff. 6214 * new fields can only be added to the end of this structure 6215 */ 6216 struct __sk_buff { 6217 __u32 len; 6218 __u32 pkt_type; 6219 __u32 mark; 6220 __u32 queue_mapping; 6221 __u32 protocol; 6222 __u32 vlan_present; 6223 __u32 vlan_tci; 6224 __u32 vlan_proto; 6225 __u32 priority; 6226 __u32 ingress_ifindex; 6227 __u32 ifindex; 6228 __u32 tc_index; 6229 __u32 cb[5]; 6230 __u32 hash; 6231 __u32 tc_classid; 6232 __u32 data; 6233 __u32 data_end; 6234 __u32 napi_id; 6235 6236 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6237 __u32 family; 6238 __u32 remote_ip4; /* Stored in network byte order */ 6239 __u32 local_ip4; /* Stored in network byte order */ 6240 __u32 remote_ip6[4]; /* Stored in network byte order */ 6241 __u32 local_ip6[4]; /* Stored in network byte order */ 6242 __u32 remote_port; /* Stored in network byte order */ 6243 __u32 local_port; /* stored in host byte order */ 6244 /* ... here. */ 6245 6246 __u32 data_meta; 6247 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6248 __u64 tstamp; 6249 __u32 wire_len; 6250 __u32 gso_segs; 6251 __bpf_md_ptr(struct bpf_sock *, sk); 6252 __u32 gso_size; 6253 __u8 tstamp_type; 6254 __u32 :24; /* Padding, future use. */ 6255 __u64 hwtstamp; 6256 }; 6257 6258 struct bpf_tunnel_key { 6259 __u32 tunnel_id; 6260 union { 6261 __u32 remote_ipv4; 6262 __u32 remote_ipv6[4]; 6263 }; 6264 __u8 tunnel_tos; 6265 __u8 tunnel_ttl; 6266 union { 6267 __u16 tunnel_ext; /* compat */ 6268 __be16 tunnel_flags; 6269 }; 6270 __u32 tunnel_label; 6271 union { 6272 __u32 local_ipv4; 6273 __u32 local_ipv6[4]; 6274 }; 6275 }; 6276 6277 /* user accessible mirror of in-kernel xfrm_state. 6278 * new fields can only be added to the end of this structure 6279 */ 6280 struct bpf_xfrm_state { 6281 __u32 reqid; 6282 __u32 spi; /* Stored in network byte order */ 6283 __u16 family; 6284 __u16 ext; /* Padding, future use. */ 6285 union { 6286 __u32 remote_ipv4; /* Stored in network byte order */ 6287 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6288 }; 6289 }; 6290 6291 /* Generic BPF return codes which all BPF program types may support. 6292 * The values are binary compatible with their TC_ACT_* counter-part to 6293 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6294 * programs. 6295 * 6296 * XDP is handled seprately, see XDP_*. 6297 */ 6298 enum bpf_ret_code { 6299 BPF_OK = 0, 6300 /* 1 reserved */ 6301 BPF_DROP = 2, 6302 /* 3-6 reserved */ 6303 BPF_REDIRECT = 7, 6304 /* >127 are reserved for prog type specific return codes. 6305 * 6306 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6307 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6308 * changed and should be routed based on its new L3 header. 6309 * (This is an L3 redirect, as opposed to L2 redirect 6310 * represented by BPF_REDIRECT above). 6311 */ 6312 BPF_LWT_REROUTE = 128, 6313 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6314 * to indicate that no custom dissection was performed, and 6315 * fallback to standard dissector is requested. 6316 */ 6317 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6318 }; 6319 6320 struct bpf_sock { 6321 __u32 bound_dev_if; 6322 __u32 family; 6323 __u32 type; 6324 __u32 protocol; 6325 __u32 mark; 6326 __u32 priority; 6327 /* IP address also allows 1 and 2 bytes access */ 6328 __u32 src_ip4; 6329 __u32 src_ip6[4]; 6330 __u32 src_port; /* host byte order */ 6331 __be16 dst_port; /* network byte order */ 6332 __u16 :16; /* zero padding */ 6333 __u32 dst_ip4; 6334 __u32 dst_ip6[4]; 6335 __u32 state; 6336 __s32 rx_queue_mapping; 6337 }; 6338 6339 struct bpf_tcp_sock { 6340 __u32 snd_cwnd; /* Sending congestion window */ 6341 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6342 __u32 rtt_min; 6343 __u32 snd_ssthresh; /* Slow start size threshold */ 6344 __u32 rcv_nxt; /* What we want to receive next */ 6345 __u32 snd_nxt; /* Next sequence we send */ 6346 __u32 snd_una; /* First byte we want an ack for */ 6347 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6348 __u32 ecn_flags; /* ECN status bits. */ 6349 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6350 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6351 __u32 packets_out; /* Packets which are "in flight" */ 6352 __u32 retrans_out; /* Retransmitted packets out */ 6353 __u32 total_retrans; /* Total retransmits for entire connection */ 6354 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6355 * total number of segments in. 6356 */ 6357 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6358 * total number of data segments in. 6359 */ 6360 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6361 * The total number of segments sent. 6362 */ 6363 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6364 * total number of data segments sent. 6365 */ 6366 __u32 lost_out; /* Lost packets */ 6367 __u32 sacked_out; /* SACK'd packets */ 6368 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6369 * sum(delta(rcv_nxt)), or how many bytes 6370 * were acked. 6371 */ 6372 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6373 * sum(delta(snd_una)), or how many bytes 6374 * were acked. 6375 */ 6376 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6377 * total number of DSACK blocks received 6378 */ 6379 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6380 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6381 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6382 }; 6383 6384 struct bpf_sock_tuple { 6385 union { 6386 struct { 6387 __be32 saddr; 6388 __be32 daddr; 6389 __be16 sport; 6390 __be16 dport; 6391 } ipv4; 6392 struct { 6393 __be32 saddr[4]; 6394 __be32 daddr[4]; 6395 __be16 sport; 6396 __be16 dport; 6397 } ipv6; 6398 }; 6399 }; 6400 6401 /* (Simplified) user return codes for tcx prog type. 6402 * A valid tcx program must return one of these defined values. All other 6403 * return codes are reserved for future use. Must remain compatible with 6404 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6405 * return codes are mapped to TCX_NEXT. 6406 */ 6407 enum tcx_action_base { 6408 TCX_NEXT = -1, 6409 TCX_PASS = 0, 6410 TCX_DROP = 2, 6411 TCX_REDIRECT = 7, 6412 }; 6413 6414 struct bpf_xdp_sock { 6415 __u32 queue_id; 6416 }; 6417 6418 #define XDP_PACKET_HEADROOM 256 6419 6420 /* User return codes for XDP prog type. 6421 * A valid XDP program must return one of these defined values. All other 6422 * return codes are reserved for future use. Unknown return codes will 6423 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6424 */ 6425 enum xdp_action { 6426 XDP_ABORTED = 0, 6427 XDP_DROP, 6428 XDP_PASS, 6429 XDP_TX, 6430 XDP_REDIRECT, 6431 }; 6432 6433 /* user accessible metadata for XDP packet hook 6434 * new fields must be added to the end of this structure 6435 */ 6436 struct xdp_md { 6437 __u32 data; 6438 __u32 data_end; 6439 __u32 data_meta; 6440 /* Below access go through struct xdp_rxq_info */ 6441 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6442 __u32 rx_queue_index; /* rxq->queue_index */ 6443 6444 __u32 egress_ifindex; /* txq->dev->ifindex */ 6445 }; 6446 6447 /* DEVMAP map-value layout 6448 * 6449 * The struct data-layout of map-value is a configuration interface. 6450 * New members can only be added to the end of this structure. 6451 */ 6452 struct bpf_devmap_val { 6453 __u32 ifindex; /* device index */ 6454 union { 6455 int fd; /* prog fd on map write */ 6456 __u32 id; /* prog id on map read */ 6457 } bpf_prog; 6458 }; 6459 6460 /* CPUMAP map-value layout 6461 * 6462 * The struct data-layout of map-value is a configuration interface. 6463 * New members can only be added to the end of this structure. 6464 */ 6465 struct bpf_cpumap_val { 6466 __u32 qsize; /* queue size to remote target CPU */ 6467 union { 6468 int fd; /* prog fd on map write */ 6469 __u32 id; /* prog id on map read */ 6470 } bpf_prog; 6471 }; 6472 6473 enum sk_action { 6474 SK_DROP = 0, 6475 SK_PASS, 6476 }; 6477 6478 /* user accessible metadata for SK_MSG packet hook, new fields must 6479 * be added to the end of this structure 6480 */ 6481 struct sk_msg_md { 6482 __bpf_md_ptr(void *, data); 6483 __bpf_md_ptr(void *, data_end); 6484 6485 __u32 family; 6486 __u32 remote_ip4; /* Stored in network byte order */ 6487 __u32 local_ip4; /* Stored in network byte order */ 6488 __u32 remote_ip6[4]; /* Stored in network byte order */ 6489 __u32 local_ip6[4]; /* Stored in network byte order */ 6490 __u32 remote_port; /* Stored in network byte order */ 6491 __u32 local_port; /* stored in host byte order */ 6492 __u32 size; /* Total size of sk_msg */ 6493 6494 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6495 }; 6496 6497 struct sk_reuseport_md { 6498 /* 6499 * Start of directly accessible data. It begins from 6500 * the tcp/udp header. 6501 */ 6502 __bpf_md_ptr(void *, data); 6503 /* End of directly accessible data */ 6504 __bpf_md_ptr(void *, data_end); 6505 /* 6506 * Total length of packet (starting from the tcp/udp header). 6507 * Note that the directly accessible bytes (data_end - data) 6508 * could be less than this "len". Those bytes could be 6509 * indirectly read by a helper "bpf_skb_load_bytes()". 6510 */ 6511 __u32 len; 6512 /* 6513 * Eth protocol in the mac header (network byte order). e.g. 6514 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6515 */ 6516 __u32 eth_protocol; 6517 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6518 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6519 __u32 hash; /* A hash of the packet 4 tuples */ 6520 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6521 * new incoming connection request (e.g. selecting a listen sk for 6522 * the received SYN in the TCP case). reuse->sk is one of the sk 6523 * in the reuseport group. The bpf prog can use reuse->sk to learn 6524 * the local listening ip/port without looking into the skb. 6525 * 6526 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6527 * reuse->migrating_sk is the socket that needs to be migrated 6528 * to another listening socket. migrating_sk could be a fullsock 6529 * sk that is fully established or a reqsk that is in-the-middle 6530 * of 3-way handshake. 6531 */ 6532 __bpf_md_ptr(struct bpf_sock *, sk); 6533 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6534 }; 6535 6536 #define BPF_TAG_SIZE 8 6537 6538 struct bpf_prog_info { 6539 __u32 type; 6540 __u32 id; 6541 __u8 tag[BPF_TAG_SIZE]; 6542 __u32 jited_prog_len; 6543 __u32 xlated_prog_len; 6544 __aligned_u64 jited_prog_insns; 6545 __aligned_u64 xlated_prog_insns; 6546 __u64 load_time; /* ns since boottime */ 6547 __u32 created_by_uid; 6548 __u32 nr_map_ids; 6549 __aligned_u64 map_ids; 6550 char name[BPF_OBJ_NAME_LEN]; 6551 __u32 ifindex; 6552 __u32 gpl_compatible:1; 6553 __u32 :31; /* alignment pad */ 6554 __u64 netns_dev; 6555 __u64 netns_ino; 6556 __u32 nr_jited_ksyms; 6557 __u32 nr_jited_func_lens; 6558 __aligned_u64 jited_ksyms; 6559 __aligned_u64 jited_func_lens; 6560 __u32 btf_id; 6561 __u32 func_info_rec_size; 6562 __aligned_u64 func_info; 6563 __u32 nr_func_info; 6564 __u32 nr_line_info; 6565 __aligned_u64 line_info; 6566 __aligned_u64 jited_line_info; 6567 __u32 nr_jited_line_info; 6568 __u32 line_info_rec_size; 6569 __u32 jited_line_info_rec_size; 6570 __u32 nr_prog_tags; 6571 __aligned_u64 prog_tags; 6572 __u64 run_time_ns; 6573 __u64 run_cnt; 6574 __u64 recursion_misses; 6575 __u32 verified_insns; 6576 __u32 attach_btf_obj_id; 6577 __u32 attach_btf_id; 6578 } __attribute__((aligned(8))); 6579 6580 struct bpf_map_info { 6581 __u32 type; 6582 __u32 id; 6583 __u32 key_size; 6584 __u32 value_size; 6585 __u32 max_entries; 6586 __u32 map_flags; 6587 char name[BPF_OBJ_NAME_LEN]; 6588 __u32 ifindex; 6589 __u32 btf_vmlinux_value_type_id; 6590 __u64 netns_dev; 6591 __u64 netns_ino; 6592 __u32 btf_id; 6593 __u32 btf_key_type_id; 6594 __u32 btf_value_type_id; 6595 __u32 btf_vmlinux_id; 6596 __u64 map_extra; 6597 } __attribute__((aligned(8))); 6598 6599 struct bpf_btf_info { 6600 __aligned_u64 btf; 6601 __u32 btf_size; 6602 __u32 id; 6603 __aligned_u64 name; 6604 __u32 name_len; 6605 __u32 kernel_btf; 6606 } __attribute__((aligned(8))); 6607 6608 struct bpf_link_info { 6609 __u32 type; 6610 __u32 id; 6611 __u32 prog_id; 6612 union { 6613 struct { 6614 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6615 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6616 } raw_tracepoint; 6617 struct { 6618 __u32 attach_type; 6619 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6620 __u32 target_btf_id; /* BTF type id inside the object */ 6621 } tracing; 6622 struct { 6623 __u64 cgroup_id; 6624 __u32 attach_type; 6625 } cgroup; 6626 struct { 6627 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6628 __u32 target_name_len; /* in/out: target_name buffer len */ 6629 6630 /* If the iter specific field is 32 bits, it can be put 6631 * in the first or second union. Otherwise it should be 6632 * put in the second union. 6633 */ 6634 union { 6635 struct { 6636 __u32 map_id; 6637 } map; 6638 }; 6639 union { 6640 struct { 6641 __u64 cgroup_id; 6642 __u32 order; 6643 } cgroup; 6644 struct { 6645 __u32 tid; 6646 __u32 pid; 6647 } task; 6648 }; 6649 } iter; 6650 struct { 6651 __u32 netns_ino; 6652 __u32 attach_type; 6653 } netns; 6654 struct { 6655 __u32 ifindex; 6656 } xdp; 6657 struct { 6658 __u32 map_id; 6659 } struct_ops; 6660 struct { 6661 __u32 pf; 6662 __u32 hooknum; 6663 __s32 priority; 6664 __u32 flags; 6665 } netfilter; 6666 struct { 6667 __aligned_u64 addrs; 6668 __u32 count; /* in/out: kprobe_multi function count */ 6669 __u32 flags; 6670 __u64 missed; 6671 __aligned_u64 cookies; 6672 } kprobe_multi; 6673 struct { 6674 __aligned_u64 path; 6675 __aligned_u64 offsets; 6676 __aligned_u64 ref_ctr_offsets; 6677 __aligned_u64 cookies; 6678 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6679 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6680 __u32 flags; 6681 __u32 pid; 6682 } uprobe_multi; 6683 struct { 6684 __u32 type; /* enum bpf_perf_event_type */ 6685 __u32 :32; 6686 union { 6687 struct { 6688 __aligned_u64 file_name; /* in/out */ 6689 __u32 name_len; 6690 __u32 offset; /* offset from file_name */ 6691 __u64 cookie; 6692 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6693 struct { 6694 __aligned_u64 func_name; /* in/out */ 6695 __u32 name_len; 6696 __u32 offset; /* offset from func_name */ 6697 __u64 addr; 6698 __u64 missed; 6699 __u64 cookie; 6700 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6701 struct { 6702 __aligned_u64 tp_name; /* in/out */ 6703 __u32 name_len; 6704 __u32 :32; 6705 __u64 cookie; 6706 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6707 struct { 6708 __u64 config; 6709 __u32 type; 6710 __u32 :32; 6711 __u64 cookie; 6712 } event; /* BPF_PERF_EVENT_EVENT */ 6713 }; 6714 } perf_event; 6715 struct { 6716 __u32 ifindex; 6717 __u32 attach_type; 6718 } tcx; 6719 struct { 6720 __u32 ifindex; 6721 __u32 attach_type; 6722 } netkit; 6723 }; 6724 } __attribute__((aligned(8))); 6725 6726 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6727 * by user and intended to be used by socket (e.g. to bind to, depends on 6728 * attach type). 6729 */ 6730 struct bpf_sock_addr { 6731 __u32 user_family; /* Allows 4-byte read, but no write. */ 6732 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6733 * Stored in network byte order. 6734 */ 6735 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6736 * Stored in network byte order. 6737 */ 6738 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6739 * Stored in network byte order 6740 */ 6741 __u32 family; /* Allows 4-byte read, but no write */ 6742 __u32 type; /* Allows 4-byte read, but no write */ 6743 __u32 protocol; /* Allows 4-byte read, but no write */ 6744 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6745 * Stored in network byte order. 6746 */ 6747 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6748 * Stored in network byte order. 6749 */ 6750 __bpf_md_ptr(struct bpf_sock *, sk); 6751 }; 6752 6753 /* User bpf_sock_ops struct to access socket values and specify request ops 6754 * and their replies. 6755 * Some of this fields are in network (bigendian) byte order and may need 6756 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6757 * New fields can only be added at the end of this structure 6758 */ 6759 struct bpf_sock_ops { 6760 __u32 op; 6761 union { 6762 __u32 args[4]; /* Optionally passed to bpf program */ 6763 __u32 reply; /* Returned by bpf program */ 6764 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6765 }; 6766 __u32 family; 6767 __u32 remote_ip4; /* Stored in network byte order */ 6768 __u32 local_ip4; /* Stored in network byte order */ 6769 __u32 remote_ip6[4]; /* Stored in network byte order */ 6770 __u32 local_ip6[4]; /* Stored in network byte order */ 6771 __u32 remote_port; /* Stored in network byte order */ 6772 __u32 local_port; /* stored in host byte order */ 6773 __u32 is_fullsock; /* Some TCP fields are only valid if 6774 * there is a full socket. If not, the 6775 * fields read as zero. 6776 */ 6777 __u32 snd_cwnd; 6778 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6779 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6780 __u32 state; 6781 __u32 rtt_min; 6782 __u32 snd_ssthresh; 6783 __u32 rcv_nxt; 6784 __u32 snd_nxt; 6785 __u32 snd_una; 6786 __u32 mss_cache; 6787 __u32 ecn_flags; 6788 __u32 rate_delivered; 6789 __u32 rate_interval_us; 6790 __u32 packets_out; 6791 __u32 retrans_out; 6792 __u32 total_retrans; 6793 __u32 segs_in; 6794 __u32 data_segs_in; 6795 __u32 segs_out; 6796 __u32 data_segs_out; 6797 __u32 lost_out; 6798 __u32 sacked_out; 6799 __u32 sk_txhash; 6800 __u64 bytes_received; 6801 __u64 bytes_acked; 6802 __bpf_md_ptr(struct bpf_sock *, sk); 6803 /* [skb_data, skb_data_end) covers the whole TCP header. 6804 * 6805 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6806 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6807 * header has not been written. 6808 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6809 * been written so far. 6810 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6811 * the 3WHS. 6812 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6813 * the 3WHS. 6814 * 6815 * bpf_load_hdr_opt() can also be used to read a particular option. 6816 */ 6817 __bpf_md_ptr(void *, skb_data); 6818 __bpf_md_ptr(void *, skb_data_end); 6819 __u32 skb_len; /* The total length of a packet. 6820 * It includes the header, options, 6821 * and payload. 6822 */ 6823 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6824 * an easy way to check for tcp_flags 6825 * without parsing skb_data. 6826 * 6827 * In particular, the skb_tcp_flags 6828 * will still be available in 6829 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6830 * the outgoing header has not 6831 * been written yet. 6832 */ 6833 __u64 skb_hwtstamp; 6834 }; 6835 6836 /* Definitions for bpf_sock_ops_cb_flags */ 6837 enum { 6838 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6839 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6840 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6841 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6842 /* Call bpf for all received TCP headers. The bpf prog will be 6843 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6844 * 6845 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6846 * for the header option related helpers that will be useful 6847 * to the bpf programs. 6848 * 6849 * It could be used at the client/active side (i.e. connect() side) 6850 * when the server told it that the server was in syncookie 6851 * mode and required the active side to resend the bpf-written 6852 * options. The active side can keep writing the bpf-options until 6853 * it received a valid packet from the server side to confirm 6854 * the earlier packet (and options) has been received. The later 6855 * example patch is using it like this at the active side when the 6856 * server is in syncookie mode. 6857 * 6858 * The bpf prog will usually turn this off in the common cases. 6859 */ 6860 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6861 /* Call bpf when kernel has received a header option that 6862 * the kernel cannot handle. The bpf prog will be called under 6863 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6864 * 6865 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6866 * for the header option related helpers that will be useful 6867 * to the bpf programs. 6868 */ 6869 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6870 /* Call bpf when the kernel is writing header options for the 6871 * outgoing packet. The bpf prog will first be called 6872 * to reserve space in a skb under 6873 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6874 * the bpf prog will be called to write the header option(s) 6875 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6876 * 6877 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6878 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6879 * related helpers that will be useful to the bpf programs. 6880 * 6881 * The kernel gets its chance to reserve space and write 6882 * options first before the BPF program does. 6883 */ 6884 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6885 /* Mask of all currently supported cb flags */ 6886 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6887 }; 6888 6889 /* List of known BPF sock_ops operators. 6890 * New entries can only be added at the end 6891 */ 6892 enum { 6893 BPF_SOCK_OPS_VOID, 6894 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6895 * -1 if default value should be used 6896 */ 6897 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6898 * window (in packets) or -1 if default 6899 * value should be used 6900 */ 6901 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6902 * active connection is initialized 6903 */ 6904 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6905 * active connection is 6906 * established 6907 */ 6908 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6909 * passive connection is 6910 * established 6911 */ 6912 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6913 * needs ECN 6914 */ 6915 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6916 * based on the path and may be 6917 * dependent on the congestion control 6918 * algorithm. In general it indicates 6919 * a congestion threshold. RTTs above 6920 * this indicate congestion 6921 */ 6922 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6923 * Arg1: value of icsk_retransmits 6924 * Arg2: value of icsk_rto 6925 * Arg3: whether RTO has expired 6926 */ 6927 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6928 * Arg1: sequence number of 1st byte 6929 * Arg2: # segments 6930 * Arg3: return value of 6931 * tcp_transmit_skb (0 => success) 6932 */ 6933 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6934 * Arg1: old_state 6935 * Arg2: new_state 6936 */ 6937 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6938 * socket transition to LISTEN state. 6939 */ 6940 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6941 */ 6942 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6943 * It will be called to handle 6944 * the packets received at 6945 * an already established 6946 * connection. 6947 * 6948 * sock_ops->skb_data: 6949 * Referring to the received skb. 6950 * It covers the TCP header only. 6951 * 6952 * bpf_load_hdr_opt() can also 6953 * be used to search for a 6954 * particular option. 6955 */ 6956 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6957 * header option later in 6958 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6959 * Arg1: bool want_cookie. (in 6960 * writing SYNACK only) 6961 * 6962 * sock_ops->skb_data: 6963 * Not available because no header has 6964 * been written yet. 6965 * 6966 * sock_ops->skb_tcp_flags: 6967 * The tcp_flags of the 6968 * outgoing skb. (e.g. SYN, ACK, FIN). 6969 * 6970 * bpf_reserve_hdr_opt() should 6971 * be used to reserve space. 6972 */ 6973 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6974 * Arg1: bool want_cookie. (in 6975 * writing SYNACK only) 6976 * 6977 * sock_ops->skb_data: 6978 * Referring to the outgoing skb. 6979 * It covers the TCP header 6980 * that has already been written 6981 * by the kernel and the 6982 * earlier bpf-progs. 6983 * 6984 * sock_ops->skb_tcp_flags: 6985 * The tcp_flags of the outgoing 6986 * skb. (e.g. SYN, ACK, FIN). 6987 * 6988 * bpf_store_hdr_opt() should 6989 * be used to write the 6990 * option. 6991 * 6992 * bpf_load_hdr_opt() can also 6993 * be used to search for a 6994 * particular option that 6995 * has already been written 6996 * by the kernel or the 6997 * earlier bpf-progs. 6998 */ 6999 }; 7000 7001 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 7002 * changes between the TCP and BPF versions. Ideally this should never happen. 7003 * If it does, we need to add code to convert them before calling 7004 * the BPF sock_ops function. 7005 */ 7006 enum { 7007 BPF_TCP_ESTABLISHED = 1, 7008 BPF_TCP_SYN_SENT, 7009 BPF_TCP_SYN_RECV, 7010 BPF_TCP_FIN_WAIT1, 7011 BPF_TCP_FIN_WAIT2, 7012 BPF_TCP_TIME_WAIT, 7013 BPF_TCP_CLOSE, 7014 BPF_TCP_CLOSE_WAIT, 7015 BPF_TCP_LAST_ACK, 7016 BPF_TCP_LISTEN, 7017 BPF_TCP_CLOSING, /* Now a valid state */ 7018 BPF_TCP_NEW_SYN_RECV, 7019 BPF_TCP_BOUND_INACTIVE, 7020 7021 BPF_TCP_MAX_STATES /* Leave at the end! */ 7022 }; 7023 7024 enum { 7025 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 7026 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 7027 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 7028 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 7029 /* Copy the SYN pkt to optval 7030 * 7031 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 7032 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 7033 * to only getting from the saved_syn. It can either get the 7034 * syn packet from: 7035 * 7036 * 1. the just-received SYN packet (only available when writing the 7037 * SYNACK). It will be useful when it is not necessary to 7038 * save the SYN packet for latter use. It is also the only way 7039 * to get the SYN during syncookie mode because the syn 7040 * packet cannot be saved during syncookie. 7041 * 7042 * OR 7043 * 7044 * 2. the earlier saved syn which was done by 7045 * bpf_setsockopt(TCP_SAVE_SYN). 7046 * 7047 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7048 * SYN packet is obtained. 7049 * 7050 * If the bpf-prog does not need the IP[46] header, the 7051 * bpf-prog can avoid parsing the IP header by using 7052 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7053 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7054 * 7055 * >0: Total number of bytes copied 7056 * -ENOSPC: Not enough space in optval. Only optlen number of 7057 * bytes is copied. 7058 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7059 * is not saved by setsockopt(TCP_SAVE_SYN). 7060 */ 7061 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7062 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7063 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7064 }; 7065 7066 enum { 7067 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7068 }; 7069 7070 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7071 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7072 */ 7073 enum { 7074 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7075 * total option spaces 7076 * required for an established 7077 * sk in order to calculate the 7078 * MSS. No skb is actually 7079 * sent. 7080 */ 7081 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7082 * when sending a SYN. 7083 */ 7084 }; 7085 7086 struct bpf_perf_event_value { 7087 __u64 counter; 7088 __u64 enabled; 7089 __u64 running; 7090 }; 7091 7092 enum { 7093 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7094 BPF_DEVCG_ACC_READ = (1ULL << 1), 7095 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7096 }; 7097 7098 enum { 7099 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7100 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7101 }; 7102 7103 struct bpf_cgroup_dev_ctx { 7104 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7105 __u32 access_type; 7106 __u32 major; 7107 __u32 minor; 7108 }; 7109 7110 struct bpf_raw_tracepoint_args { 7111 __u64 args[0]; 7112 }; 7113 7114 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7115 * OUTPUT: Do lookup from egress perspective; default is ingress 7116 */ 7117 enum { 7118 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7119 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7120 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7121 BPF_FIB_LOOKUP_TBID = (1U << 3), 7122 BPF_FIB_LOOKUP_SRC = (1U << 4), 7123 }; 7124 7125 enum { 7126 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7127 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7128 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7129 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7130 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7131 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7132 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7133 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7134 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7135 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7136 }; 7137 7138 struct bpf_fib_lookup { 7139 /* input: network family for lookup (AF_INET, AF_INET6) 7140 * output: network family of egress nexthop 7141 */ 7142 __u8 family; 7143 7144 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7145 __u8 l4_protocol; 7146 __be16 sport; 7147 __be16 dport; 7148 7149 union { /* used for MTU check */ 7150 /* input to lookup */ 7151 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7152 7153 /* output: MTU value */ 7154 __u16 mtu_result; 7155 }; 7156 /* input: L3 device index for lookup 7157 * output: device index from FIB lookup 7158 */ 7159 __u32 ifindex; 7160 7161 union { 7162 /* inputs to lookup */ 7163 __u8 tos; /* AF_INET */ 7164 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7165 7166 /* output: metric of fib result (IPv4/IPv6 only) */ 7167 __u32 rt_metric; 7168 }; 7169 7170 /* input: source address to consider for lookup 7171 * output: source address result from lookup 7172 */ 7173 union { 7174 __be32 ipv4_src; 7175 __u32 ipv6_src[4]; /* in6_addr; network order */ 7176 }; 7177 7178 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7179 * network header. output: bpf_fib_lookup sets to gateway address 7180 * if FIB lookup returns gateway route 7181 */ 7182 union { 7183 __be32 ipv4_dst; 7184 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7185 }; 7186 7187 union { 7188 struct { 7189 /* output */ 7190 __be16 h_vlan_proto; 7191 __be16 h_vlan_TCI; 7192 }; 7193 /* input: when accompanied with the 7194 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7195 * specific routing table to use for the fib lookup. 7196 */ 7197 __u32 tbid; 7198 }; 7199 7200 __u8 smac[6]; /* ETH_ALEN */ 7201 __u8 dmac[6]; /* ETH_ALEN */ 7202 }; 7203 7204 struct bpf_redir_neigh { 7205 /* network family for lookup (AF_INET, AF_INET6) */ 7206 __u32 nh_family; 7207 /* network address of nexthop; skips fib lookup to find gateway */ 7208 union { 7209 __be32 ipv4_nh; 7210 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7211 }; 7212 }; 7213 7214 /* bpf_check_mtu flags*/ 7215 enum bpf_check_mtu_flags { 7216 BPF_MTU_CHK_SEGS = (1U << 0), 7217 }; 7218 7219 enum bpf_check_mtu_ret { 7220 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7221 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7222 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7223 }; 7224 7225 enum bpf_task_fd_type { 7226 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7227 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7228 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7229 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7230 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7231 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7232 }; 7233 7234 enum { 7235 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7236 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7237 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7238 }; 7239 7240 struct bpf_flow_keys { 7241 __u16 nhoff; 7242 __u16 thoff; 7243 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7244 __u8 is_frag; 7245 __u8 is_first_frag; 7246 __u8 is_encap; 7247 __u8 ip_proto; 7248 __be16 n_proto; 7249 __be16 sport; 7250 __be16 dport; 7251 union { 7252 struct { 7253 __be32 ipv4_src; 7254 __be32 ipv4_dst; 7255 }; 7256 struct { 7257 __u32 ipv6_src[4]; /* in6_addr; network order */ 7258 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7259 }; 7260 }; 7261 __u32 flags; 7262 __be32 flow_label; 7263 }; 7264 7265 struct bpf_func_info { 7266 __u32 insn_off; 7267 __u32 type_id; 7268 }; 7269 7270 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7271 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7272 7273 struct bpf_line_info { 7274 __u32 insn_off; 7275 __u32 file_name_off; 7276 __u32 line_off; 7277 __u32 line_col; 7278 }; 7279 7280 struct bpf_spin_lock { 7281 __u32 val; 7282 }; 7283 7284 struct bpf_timer { 7285 __u64 __opaque[2]; 7286 } __attribute__((aligned(8))); 7287 7288 struct bpf_dynptr { 7289 __u64 __opaque[2]; 7290 } __attribute__((aligned(8))); 7291 7292 struct bpf_list_head { 7293 __u64 __opaque[2]; 7294 } __attribute__((aligned(8))); 7295 7296 struct bpf_list_node { 7297 __u64 __opaque[3]; 7298 } __attribute__((aligned(8))); 7299 7300 struct bpf_rb_root { 7301 __u64 __opaque[2]; 7302 } __attribute__((aligned(8))); 7303 7304 struct bpf_rb_node { 7305 __u64 __opaque[4]; 7306 } __attribute__((aligned(8))); 7307 7308 struct bpf_refcount { 7309 __u32 __opaque[1]; 7310 } __attribute__((aligned(4))); 7311 7312 struct bpf_sysctl { 7313 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7314 * Allows 1,2,4-byte read, but no write. 7315 */ 7316 __u32 file_pos; /* Sysctl file position to read from, write to. 7317 * Allows 1,2,4-byte read an 4-byte write. 7318 */ 7319 }; 7320 7321 struct bpf_sockopt { 7322 __bpf_md_ptr(struct bpf_sock *, sk); 7323 __bpf_md_ptr(void *, optval); 7324 __bpf_md_ptr(void *, optval_end); 7325 7326 __s32 level; 7327 __s32 optname; 7328 __s32 optlen; 7329 __s32 retval; 7330 }; 7331 7332 struct bpf_pidns_info { 7333 __u32 pid; 7334 __u32 tgid; 7335 }; 7336 7337 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7338 struct bpf_sk_lookup { 7339 union { 7340 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7341 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7342 }; 7343 7344 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7345 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7346 __u32 remote_ip4; /* Network byte order */ 7347 __u32 remote_ip6[4]; /* Network byte order */ 7348 __be16 remote_port; /* Network byte order */ 7349 __u16 :16; /* Zero padding */ 7350 __u32 local_ip4; /* Network byte order */ 7351 __u32 local_ip6[4]; /* Network byte order */ 7352 __u32 local_port; /* Host byte order */ 7353 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7354 }; 7355 7356 /* 7357 * struct btf_ptr is used for typed pointer representation; the 7358 * type id is used to render the pointer data as the appropriate type 7359 * via the bpf_snprintf_btf() helper described above. A flags field - 7360 * potentially to specify additional details about the BTF pointer 7361 * (rather than its mode of display) - is included for future use. 7362 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7363 */ 7364 struct btf_ptr { 7365 void *ptr; 7366 __u32 type_id; 7367 __u32 flags; /* BTF ptr flags; unused at present. */ 7368 }; 7369 7370 /* 7371 * Flags to control bpf_snprintf_btf() behaviour. 7372 * - BTF_F_COMPACT: no formatting around type information 7373 * - BTF_F_NONAME: no struct/union member names/types 7374 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7375 * equivalent to %px. 7376 * - BTF_F_ZERO: show zero-valued struct/union members; they 7377 * are not displayed by default 7378 */ 7379 enum { 7380 BTF_F_COMPACT = (1ULL << 0), 7381 BTF_F_NONAME = (1ULL << 1), 7382 BTF_F_PTR_RAW = (1ULL << 2), 7383 BTF_F_ZERO = (1ULL << 3), 7384 }; 7385 7386 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7387 * has to be adjusted by relocations. It is emitted by llvm and passed to 7388 * libbpf and later to the kernel. 7389 */ 7390 enum bpf_core_relo_kind { 7391 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7392 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7393 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7394 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7395 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7396 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7397 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7398 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7399 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7400 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7401 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7402 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7403 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7404 }; 7405 7406 /* 7407 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7408 * and from libbpf to the kernel. 7409 * 7410 * CO-RE relocation captures the following data: 7411 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7412 * its insn->imm field to be relocated with actual field info; 7413 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7414 * type or field; 7415 * - access_str_off - offset into corresponding .BTF string section. String 7416 * interpretation depends on specific relocation kind: 7417 * - for field-based relocations, string encodes an accessed field using 7418 * a sequence of field and array indices, separated by colon (:). It's 7419 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7420 * arguments for identifying offset to a field. 7421 * - for type-based relocations, strings is expected to be just "0"; 7422 * - for enum value-based relocations, string contains an index of enum 7423 * value within its enum type; 7424 * - kind - one of enum bpf_core_relo_kind; 7425 * 7426 * Example: 7427 * struct sample { 7428 * int a; 7429 * struct { 7430 * int b[10]; 7431 * }; 7432 * }; 7433 * 7434 * struct sample *s = ...; 7435 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7436 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7437 * // b is field #0 inside anon struct, accessing elem #5) 7438 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7439 * 7440 * type_id for all relocs in this example will capture BTF type id of 7441 * `struct sample`. 7442 * 7443 * Such relocation is emitted when using __builtin_preserve_access_index() 7444 * Clang built-in, passing expression that captures field address, e.g.: 7445 * 7446 * bpf_probe_read(&dst, sizeof(dst), 7447 * __builtin_preserve_access_index(&src->a.b.c)); 7448 * 7449 * In this case Clang will emit field relocation recording necessary data to 7450 * be able to find offset of embedded `a.b.c` field within `src` struct. 7451 * 7452 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7453 */ 7454 struct bpf_core_relo { 7455 __u32 insn_off; 7456 __u32 type_id; 7457 __u32 access_str_off; 7458 enum bpf_core_relo_kind kind; 7459 }; 7460 7461 /* 7462 * Flags to control bpf_timer_start() behaviour. 7463 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7464 * relative to current time. 7465 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7466 */ 7467 enum { 7468 BPF_F_TIMER_ABS = (1ULL << 0), 7469 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7470 }; 7471 7472 /* BPF numbers iterator state */ 7473 struct bpf_iter_num { 7474 /* opaque iterator state; having __u64 here allows to preserve correct 7475 * alignment requirements in vmlinux.h, generated from BTF 7476 */ 7477 __u64 __opaque[1]; 7478 } __attribute__((aligned(8))); 7479 7480 #endif /* _UAPI__LINUX_BPF_H__ */ 7481